Cross-Species Radiomics: Evaluating the Generalizability of Intervertebral Disc MRI-based Radiomics Models between Humans and Experimental Monkeys

Cross-Species Radiomics: Evaluating the Generalizability of Intervertebral Disc MRI-based Radiomics Models between Humans and Experimental Monkeys | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Cross-Species Radiomics: Evaluating the Generalizability of Intervertebral Disc MRI-based Radiomics Models between Humans and Experimental Monkeys Jianmin Wang, Lei Guo, Jianfeng Li, Xiaodong Cao, Wei Du, Jiaxiang Zhou, and 9 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4486357/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 28 Nov, 2025 Read the published version in Scientific Reports → Version 1 posted 12 You are reading this latest preprint version Abstract Experimental monkeys serve as a bridge between basic research and clinical medicine. Accurately assessing the degree of intervertebral disc degeneration (IVDD) in experimental monkeys is crucial for further intervertebral disc related research in these animals. Radiomics promises significant enhancement in quantitative diagnostic precision for IVDD, while the cornerstone of constructing robust and efficient radiomics models (RMs) relies on access to large-scale sample data. In experimental monkey research, however, ethical restrictions and resource constraints typically limit sample sizes. This study addresses this challenge by comparing and analyzing the generalizability of intervertebral disc MRI-based radiomics models between humans and experimental monkeys. The findings reveal that 12.30% (438/3562) of the radiomics features demonstrate high reproducibility between the two species. Leveraging the sufficient human dataset, we built RMs and employed the experimental monkey dataset as a training set to validate the cross-species generalizability of these models. Notably, in the test phase, models constructed based on the inter-species reproducible features achieved AUC values ranging from 0.82 to 0.92, indicative of promising diagnostic performance. This study emphasizes the advantages of leveraging human data for the construction of RMs under conditions of constrained experimental monkey research. We innovatively propose and validate the potential for cross-species application of RMs. This study furnishes strong theoretical underpinnings and practical foundations for the broader application of radiomics in cross-species disease research. Biological sciences/Computational biology and bioinformatics/Machine learning Health sciences/Biomarkers/Diagnostic markers Biological sciences/Zoology/Animal physiology Radiomics Radiomics model Generalizability Machine learning Experimental monkey Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1. Introduction Experimental monkey species exhibit astonishing similarities to humans in locomotor behavior patterns, cell composition of the intervertebral disc (IVD), and the progression of intervertebral disc degeneration (IVDD), making them an ideal alternative animal model for studying human IVDD[ 1 ]. Furthermore, the experimental monkey model is the final step in preclinical development of drugs and vaccines, greatly improving the reliability of research results and providing a solid foundation for the eventual clinical translation. Experimental monkeys play an important role in basic and translational biomedical research, serving as a bridge between basic research and clinical medicine. In the research of experimental monkey IVDD, diagnostic errors in the degree of IVDD could seriously affect subsequent related research. Consequently, accurate assessment of the degree of IVDD holds great significance for in-depth research into the pathophysiology and therapeutic interventions of the condition. At present, the degree of IVDD is often determined using the Pfirrmann grading[ 2 ]. However, it has considerable subjectivity and inability to distinguish minor differences in IVDD[ 2 ]. Meanwhile, when graded by different observers, differences of opinion could also arise[ 3 ]. The pursuit of innovative techniques for accurate assessment of the degree of IVDD holds considerable significance in advancing the research and improving therapeutic strategies targeting IVDD. As a significant innovation in medical imaging analysis technology in recent years, radiomics could quantitatively extract and analyze high-throughput radiomics features (RFs) from medical images, providing richer and more accurate information to assist in clinical disease diagnosis and treatment outcome prediction[ 4 ]. Radiomics has shown great value in applications such as diseases diagnosis[ 5 ], prediction of clinical treatment outcomes[ 6 – 8 ], assessment of the pathological heterogeneity of whole tumor tissues[ 9 , 10 ], and gene expression prediction[ 11 ]. Therefore, radiomics technology should be adopted to extract high-throughput RFs from IVD imaging data, with the aim of more accurately assessing the degree of IVDD based on these high-throughput RFs. When constructing radiomics models (RMs), a common challenge is the limited sample size, which reduces the robustness and reliability of RMs. Large sample size could enhance the performance of RMs. In the study of experimental monkeys, due to ethical and resource constraints, the number of precious animals is limited, which cannot meet the sample size requirements of radiomics for study subjects, while these data could be obtained more easily in humans. There is a large amount of human IVD data available in clinical practice for constructing RMs. Therefore, we propose a hypothesis that RMs could be constructed based on human IVD data and then applied in experimental monkeys. However, it is unclear whether the RFs of the IVDs between experimental monkey and humans are reproducible and whether the resulting RMs could be applied across species. Cross-species application of the IVD RM has not previously been well established in the literature. In the present study, we analyzed the reproducibility of RFs in humans and cynomolgus monkeys using lumbar and lower thoracic IVDs as subjects. We also used the human dataset as a training set to construct RMs to predict the degree of disc degeneration and validated the interspecies generality of the RMs with the experimental monkey dataset. Our objective is to develop a methodology that compensates for the limitation in sample size when constructing RMs, thereby enabling accurate assessment of the degree of IVDD in experimental monkeys. This will pave the way for further experimentation and research concerning IVDD in experimental monkey model. A diagram of this workflow is illustrated in Fig. 1 . 2. Materials and methods 2.1. Study subjects This study included a prospective dataset from cynomolgus monkeys and a retrospective dataset from human volunteers. The experimental protocol was reviewed and approved by the ethics committee of the Institute of Zoology, Guangdong Academy of Sciences (No. G2Z20210103) and the ethics committee of the First Hospital of Sun Yat-sen University and the institutional review board (No. 2008-55). This study was conducted strictly in accordance with ARRIVE guidelines ( https://arriveguidelines.org ). All experiments were performed in accordance with relevant guidelines and regulations. Research involving human research participants was conducted in accordance with the Declaration of Helsinki. The written informed consent was obtained from all subjects before enrollment in the study. The experimental monkeys completed lumbar MRI in 2021. The human study subjects were volunteers who underwent MRI examinations of the lumbar spine at the First Affiliated Hospital of Sun Yat-sen University from July 2009 to November 2010. The inclusion criteria were as follows: ① Human volunteers who volunteer for research. ② Study subjects had no history of spinal surgery, trauma, and experimental spinal research. The exclusion criteria were as follows: Study subjects with poor image quality. 2.2. MRI protocol Human MRI data were acquired using a 1.5-Tesla MRI scanner (Philips, United States). The experimental monkey MRIs were all performed using a 3.0-Tesla MRI scanner (General Electric Company, United States). All scans were performed in the supine position. The MRI protocol for the lumbar spine consisted of a sagittal T1WI sequence and T2WI sequence (Table 1 , Fig. 2 ). The experimental monkey MRIs were preceded by anesthesia delivered by an experienced veterinarian who was also responsible for animal care. Table 1 Parameters for mr imaging of humans and experimental monkeys. Parameter Experimental Monkey Human T1WI T2WI T1WI T2WI Pulse sequence FSE FSE Spin echo Spin echo Repetition time (ms) 500 2000 448 3228 Echo time (ms) 8 102 8.0 100 Field of view (mm) 20×20 20×20 250×250 250×250 Pixel bandwidth (Hz) 390.6 325.5 320 348 Voxel size (mm) 0.8×0.8 0.8×0.8 0.9×1.25 0.9×1.25 Slice thickness (mm) 3 3 3 3 Interslice gap (mm) 0.6 0.6 0.3 0.3 Number of slices 7 7 15 15 Turbo factor 3 16 4 45 Acquisition time 70 100 208 160 Note: T1WI = T1-weighted imaging, T2WI = T2-weighted imaging. 2.3. Pfirrmann grading The degree of disc degeneration was determined according to the criteria described by Pfirrmann et al[ 2 ]. The grading process was performed independently by 2 experts (Zhiyu Z., and Jianmin W., with 25, 8 years of clinical work experience, respectively) performing 2 analyses each at least 1 week apart. Inconsistencies in the disc grading results were resolved later by discussion among the 2 experts. In the subsequent radiomics analysis, we considered grades Ⅰ-Ⅱ to represent normal discs and marked them as 0 and grades Ⅲ-Ⅴ to represent degenerated discs and marked them as 1. 2.4. Image data preprocessing and image segmentation. We used the N4ITK Bias Field Correction module ( https://www.slicer.org/wiki/Documentation/Nightly/Modules/N4ITKBiasFieldCorrection )[ 12 ] in 3D Slicer software ( https://www.slicer.org/ , version 4.11.20210226) to perform bias fields corrections. The regions of interest (ROIs) were manually segmented in 3D Slicer software in the median sagittal plane of the target disc and included the entire nucleus pulposus, annulus fibrosus, and endplate (Fig. 2 ). The ROIs were initially segmented on the T2WI and subsequently replicated on the T1WI by two specialists (Zhiyu Z., J.W.). One of the specialists (Zhiyu Z.) performed 2 analyses, at least 1 week apart; the ROIs outlined the 2nd time were selected for the RF analysis and modeling studies. 2.5. Extraction of radiomics features. PyRadiomics ( https://pyradiomics.readthedocs.io/en/latest/index.html , version 3.0.1)[ 13 ], an Image Biomarker Standardization Initiative (IBSI)[ 14 ] guideline-compliant program, was used to extract the RFs on the images from both the T1WI and T2WI sequences. The images were normalized before feature extraction[ 15 ], then resampled (resampled voxel size set to 1,1,1) with a binWidth of 25. ‘sitk.sitkBSpline’ was used as an interpolator. All features based on the original image and derived images were extracted. All the features extracted by PyRadiomics were grouped into feature set A. 2.6. Screening of reproducible radiomics features between humans and experimental monkeys. First, we used the independent-samples t test to screen for reproducible features between humans and experimental monkeys. The features with p < 0.05 were removed. We further screened for reproducible RFs between species by LASSO. In this study, we treated features with regression coefficients equal to 0 as those that contributed less to interspecies differences, meaning that they are reproducible features across species. Reproducible features between the species screened with the combination of the independent-samples t test and LASSO to form feature set B. 2.7. Intraobserver and interobserver agreement. The intraclass correlation coefficient (ICC) of the RFs of different species may be different between different observers and even within the same observer. Here, we used the Pingouin package (version-0.5.1) to calculate the ICC. Intra- and interobserver ICCs were calculated separately for both species. Features with ICCs > 0.75 in all tests were considered to have satisfactory agreement[ 16 ]. 2.8. Construction of the radiomics models and validation of model performance across species. To verify whether a RM could be applied with similar efficacy in both humans and experimental monkeys, we constructed RMs based on feature set A, and feature set B of humans. First, the RFs were standardized using z scores. Mutual information (MI) and LASSO were used for dimensionality reduction. We used the human data as a training set to construct RMs to assist in evaluating the degree of IVDD using Support Vector Machine (SVM), Decision Tree Classifier, Random Forest Classifier, Logistic Regression, and Naive Bayes Classifier, respectively. Then, the experimental monkey data were used as the test set to verify the generalizability of the RMs across species. Due to data imbalance, oversampling was performed to balance the data using the imbalanced-learn package (version-0.9.0). 2.9. Statistical Analysis. IBM SPSS Statistics for Windows v.26 (SPSS, Chicago), Python (version-3.7.13), and R (version-4.1.1) were used to perform the statistical analyses. The glmnet R package (version-4.1. 4) was used to perform LASSO analyses. Differences between groups were assessed by the t test (scipy package version 1.7.3) or chi square test. For all analyses, p < 0.05 was considered significant. Values are expressed as the mean ± standard deviation (SD). Model performance was evaluated by receiver operating characteristic (ROC) curve. 3. Results 3.1. Participant characteristics. The characteristics of the 720 enrolled IVDs are shown in Table 2 . In humans, a total of 90 volunteers underwent MRI and 1 case was excluded because the poor image quality. A total of 89 human volunteers (61 men, 28 women) with a mean age of 31.91 years ± 6.62 (SD) were included. MRI data of 575 IVDs were obtained from human volunteers. The Pfirrmann grading distribution of these IVDs was as follows: grade I-II: 436, grade III-V: 139. Sixteen experimental monkeys completed MRI and were included, with no excluded cases. The enrolled experimental animals had a mean age of 11.81 years ± 4.42 (SD) and a mean body weight was 7.94 kg ± 1.91 (SD). All experimental monkeys were males. MRI data of 145 IVDs were obtained from experimental monkeys. The Pfirrmann grading distribution of these IVDs was as follows: grade I-II: 100, grade III-V: 45. Table 2 Participant characteristics. Characteristic Total Training Set Test Set p Value Number of participants 105 89 16 Number of IVDs 720 575 145 Pfirrmann grades 0.091 I-II 536 436 100 III-V 184 139 45 Age (y)* 33.91 ± 9.80 31.91 ± 6.56 41.83 ± 15.12 # <0.001 # Sex <0.001 M 538 393 145 F 182 182 0 Segments <0.001 T8/9 1 - 1 T9/10 1 - 1 T10/11 3 1 2 T11/12 57 44 13 T12/L1 105 89 16 L1/2 105 89 16 L2/3 105 89 16 L3/4 105 89 16 L4/5 105 89 16 L5/6 & 101 85 16 L6/7 16 - 16 L7/S1 16 - 16 Note: 1) All data are the number of intervertebral discs or participants, except for age. 2) *: Data are means ± standard deviations. 3) #: The age of experimental monkeys has been converted into the age equivalent to that of humans in a ratio of 1:3.5. 4) &: In humans, it is L5/S1. 5) IVD: intervertebral disc. 3.2. Analysis of radiomics features. In this study, 3562 features (feature set A) were extracted from the MR images of 575 human discs and 145 experimental monkey discs (Fig. 3 a). The distributions of these features in humans and experimental monkeys did not show species differences. The number of features extracted was the same for both the T1-weighted imaging (T1WI) and T2-weighted imaging (T2WI) sequences (Supplementary Fig. 1a). The 3562 features were distributed in 10 Image Types, including original image and 9 derived images (Fig. 3 b). The 10 Image Types were: original, wavelet, log, square, squareroot, logarithm, exponential, gradient, LocalBinaryPattern2D (lbp-2D), and LocalBinaryPattern3D (lbp-3D). The wavelet Image Type also included 8 subprojects (HHH, HHL, HLH, HLL, LHH, LHL, LLH, LLL), while the lbp-3D Image Type included 3 subprojects (k, m1, m2). Figure 3 b shows the number of features in each Image Type. The features extracted by Pyradiomics were distributed into 7 Feature Classes, namely, First Order Features (firstorder), Shape Features, Gray Level Co-occurrence Matrix (glcm) Features, Gray Level Size Zone Matrix (glszm) Features, Gray Level Run Length Matrix (glrlm) Features, Neighbouring Gray Tone Difference Matrix (ngtdm) Features, and Gray Level Dependence Matrix (gldm) Features. Figure 3 c shows the number of features in each Feature Class. Among them, the Shape features were extracted only from the original image; the remaining Feature Class features were extracted from both the original image and derived images. Of all the Feature Classes, the smallest was the Shape class, with 28 features, and the largest was glcm, with 912 features. A total of 214 features were extracted from the original image, and their distributions were identical in both the T1WI and T2WI sequences (Supplementary Fig. 1b). 3.3. Reproducible radiomics features between humans and experimental monkeys. A total of 559 features were obtained after removing features with p < 0.05 by t test. Furthermore, using least absolute shrinkage and selection operator (LASSO), 438 (feature set B) of these 559 features were found to be reproducible between the species (Supplementary Fig. 2a). There were 183 features extracted from the T1WI and 255 features extracted from the T2WI in feature set B (Supplementary Fig. 2d). Of the Image Types that were represented in feature set B, the Image Types with the largest number of features was wavelet-HLH (50 of 438, 11.42%), followed by exponential (39 of 438, 8.90%) (Fig. 4 a, Supplementary Fig. 2b). Among the seven Feature Classes, the Feature Class with the highest number of features was glszm (125 of 438, 28.54%), followed by glrlm (84 of 438, 19.18%) (Fig. 4 a, Supplementary Fig. 2c). In feature set B, exponential reproducible features from the T2WI and wavelet-HLH reproducible features from the T1WI accounted for the highest percentages, 33.33% (31 of 93) and 31.18% (29 of 93), respectively. Analysis of the reproducible feature percentages in terms of Feature Class shows the highest value for the glszm features from the T2WIs, at 24.67% (75 of 304) (Fig. 4 b, c). 3.4. Intraclass correlation coefficient analysis between humans and experimental monkeys. Intra- and interobserver stability of the RFs between humans and experimental monkeys was determined using intraclass correlation coefficient (ICC) analysis. Figure 5 (a, b) shows the intra- and interobserver ICC values for the 2 species in feature set A, and feature set B. In total, 766 (feature set A1), and 67 features (feature set B1) selected from each feature set that had intra- and interobserver ICCs > 0.75 for both species were used for follow-up studies (Fig. 5 c, d). The proportions of human features with intra- and interobserver ICCs > 0.75 in feature set A, and feature set B were higher than the corresponding proportions of experimental monkey features, except intraobserver in feature set B (Fig. 5 e). 3.5. Effect of reproducible radiomics features screening on dimensionality reduction. Nine (feature set A2), and 7 (feature set B2) RFs were obtained after filtering feature set A1, and feature set B1, respectively. The features and corresponding weights occupied of the filtered features are detailed in Fig. 6 (a, b). The features obtained by dimensionality reduction in two feature sets were mainly from the T2WI sequences, with percentages of 77.78% (7 of 9), and 71.43% (5 of 7), respectively (Supplementary Fig. 4a). The Image Types with the highest number of features in feature set A2 was square (3 of 9 [33.33%]), while wavelet-LHL (2 of 7 [28.57%]), wavelet-LLL (2 of 7 [28.57%]), and original (2 of 7 [28.57%]) were the largest in feature set B2 (Supplementary Fig. 4b, c). The Feature Classes with the largest number of features in feature set A2 and feature set B2 were glcm (3 of 9 [33.33%]) and firstorder (3 of 7 [42.86%]) (Supplementary Fig. 4b, c). In feature set A2 and feature set B2, the differences in feature values were compared between humans and experimental monkeys. In feature set A2, all feature values were significantly different between humans and experimental monkeys ( p < 0.001) (Fig. 6 c). However, this difference was not statistically significant in feature set B2 ( p ≥ 0.05) (Fig. 6 d). In feature set A2, the differences in feature values between species may impact the performance of the RM. 3.6. Validation of radiomics models’ performance across species. The data from the human were used as the training set, and Support Vector Machine (SVM), Decision Tree Classifier, Random Forest Classifier, Logistic Regression, and Naive Bayes Classifier were used to construct RMs to evaluate IVDD. In the training set, the areas under the curve (AUCs) of the five models constructed based on feature set A2 were 0.96, 0.99, 1.00, 0.96, and 0.95, respectively, while the AUCs of the five models constructed based on feature set B2 were 0.93, 0.99, 1.00, 0.92, and 0.89, respectively (Fig. 7 a, c). Data from experimental monkeys were used as the test set to validate the models’ performance across species. In the test set, the AUCs of the five models constructed based on feature set A2 were 0.70, 0.51, 0.96, 0.95, and 0.96, respectively, while the AUCs of the five models constructed based on feature set B2 were 0.89, 0.82, 0.88, 0.85, and 0.92, respectively (Fig. 7 b, d). In the test set, the AUC values of the RMs constructed based on feature set A2 were higher than that of the RMs constructed based on feature set B2, except for the SVM and Decision Tree Classifier. However, the sensitivity of the first four RMs (SVM, Decision Tree Classifier, Random Forest Classifier, and Logistic Regression) constructed based on feature set A2 performed poorly in identifying degenerative IVDs in experimental monkeys, with 0.36, 0.02, 0.02, and 0.20, respectively (Table 3 ). In contrast, the sensitivity of the above RMs constructed based on feature set B2 in identifying degenerative IVDs in experimental monkeys was 0.96, 0.82, 0.82, and 0.93, respectively (Table 3 ). Table 3 The performance comparison of different models. Model Feature Set AUC Accuracy Precision Sensitivity 0 1 0 1 Support Vector Machine A2 0.70 0.79 0.77 0.89 0.98 0.36 B2 0.89 0.80 0.97 0.61 0.73 0.96 Decision Tree Classifier A2 0.51 0.70 0.69 1.00 1.00 0.02 B2 0.82 0.74 0.90 0.56 0.71 0.82 Random Forest Classifier A2 0.96 0.70 0.69 1.00 1.00 0.02 B2 0.88 0.83 0.91 0.69 0.83 0.82 Logistic Regression A2 0.95 0.75 0.74 1.00 1.00 0.20 B2 0.85 0.72 0.96 0.53 0.63 0.93 Naive Bayes Classifier A2 0.96 0.90 0.92 0.86 0.94 0.82 B2 0.92 0.85 0.99 0.68 0.79 0.98 Note: 0 = Healthy intervertebral discs, 1 = Degenerated intervertebral discs. The above results show that the RMs could be applied across species, and the use of the t test combined with the LASSO method to screen reproducible features between species could improve the performance of the model. 4. Discussion The accurate assessment of the degree of IVDD in experimental monkeys is crucial for further research concerning IVDD in experimental monkeys. Radiomics is an emerging field that extracts high-dimensional quantitative features from medical images, showing promising prospects in enhancing disease representation and diagnosis. When constructing RMs, a large number of datasets are needed to optimize model performance. The shortage of sample size poses significant challenges to the development of robust and reliable RMs. To this end, this study explored the generalizability of radiomics models between humans and experimental monkeys, with a focus on IVD. Here, we analyzed the reproducibility of radiomics features between humans and experimental monkey and found that a total of 12.30% (438/3562) of radiomics features were reproducible between species. Subsequently, we constructed radiomics models based on the human dataset and used the data from the experimental monkey dataset as a testing set to verify the generalizability of the model between species. In the test set, the AUCs of the models constructed based on inter species reproducible features reached 0.82–0.92. This study provides a theoretical basis for the cross-species application of radiomics. Through RFs, it is possible to interpret disease features and understand potential pathological and physiological processes. The biological significance of RFs lies in their ability to quantitatively express macroscopic and microscopic tissue features that are invisible to the naked eye. By combining these features with advanced analytical techniques, it is possible to discover new biomarkers, improve disease classification, and guide personalized treatment strategies, ultimately promoting our understanding of disease mechanisms. The RFs extracted by PyRadiomics include First Order Statistics (first order), Shape based (3D), Shape based (2D), Gray Level Co occurrence Matrix (glcm), Gray Level Run Length Matrix (glrlm), Gray Level Size Zone Matrix (glszm), Neighboring Gray Tone Difference Matrix (ngtdm), and Gray Level Dependence Matrix (gldm). By decomposing images into numerous RFs, radiomics could provide multi parameter characterization of the process of degenerative diseases, potentially capturing subtle changes overlooked by traditional visual analysis. This fine-grained evaluation could enhance understanding of IVDD progression and improve diagnostic precision. For example, texture analysis features could reflect the characteristics of organizational microstructure. GLCM represents the statistical patternss of image texture and intensity, where features such as contrast or uniformity could provide a deeper understanding of the randomness and regularity of image grayscale. These indicators could reflect changes in cell density, fibrosis or necrosis, which are known biological indicators of disease progression or response to treatment. Due to common sample size limitations in experimental monkey research, constructing high-precision radiomics models (RM) directly on experimental monkeys faces challenges. Therefore, we have adopted an innovative strategy: first, we use human rich imaging data to construct radiomics models, thanks to the relatively abundant sample resources in human research. Subsequently, we attempted to apply these models to crab eating monkeys to test their cross species applicability and reproducibility of radiomics features. This research design cleverly bypasses the challenge of sample size. Through comparative analysis, we aim to reveal which RFs are consistent between two species and which features may be influenced by species specificity, laying the foundation for future cross species medical research and disease understanding, and providing strong support for the development of new diagnostic and treatment methods for diseases such as intervertebral disc degeneration. Feature reproducibility plays an important role in radiomics research[ 17 – 19 ]. We investigated the reproducibility of RFs between human and experimental monkey and found that a number of these features were indeed reproducible. We screened 438 features (feature set B) that were reproducible between species from 3562 features by t -test combined with LASSO's method. The number of T2WI features was greater than the number of T1WI features in feature set B (255:183). In feature set B, the Image Types with the highest number of features were wavelet-HLH (50 of 438, 11.42%), and the Feature Classes with the highest number of features were glszm (125 of 438, 28.54%). RMs constructed based on reproducible RFs could theoretically be applied across species. To verify this speculation, we used the data from the human as the training set and the experimental monkeys’ data as the test set to evaluate IVDD. In the RMs constructed based on feature sets A, and B, the AUCs in the test set were 0.70, 0.51, 0.96, 0.95, 0.96, and 0.89, 0.82, 0.88, 0.85, 0.92, respectively. This suggests that the RMs constructed based on the human’s dataset could be applied in experimental monkeys. At the same time, we also found that some RMs performed poorly in the sensitivity of identifying degenerative IVDs, with values of 0.36, 0.02, 0.02, and 0.20, respectively. However, after removing the non-reproducible features between species, the sensitivity reaches 0.82–0.96. So, the use of the t test combined with the LASSO method to screen reproducible features between species could improve the performance of the model. When screening for reproducible RFs between species using independent samples t -test, RFs with p < 0.05 could be considered to be definitely different between species. However, the opposite is not necessarily true; that is, RFs with p ≥ 0.05 do not necessarily differ between species. LASSO, a regression analysis method used to simultaneously perform feature selection and regularization, was first proposed by Robert Tibshirani in 1996[ 20 ]. By forcing the sum of the absolute values of the regression coefficients to be less than a fixed value, LASSO forces some regression coefficients to be zero, thus effectively selects simpler models that do not include the covariates corresponding to these regression coefficients. That is, the covariates whose regression coefficients become zero following LASSO play a smaller role in the prediction of the results. In this study, we chose covariates with regression coefficients of 0 as characteristics that are reproducible across species. The influence of the reproducibility of RFs may be present in the various steps of radiomics, in addition to its interspecies nature[ 21 – 23 ]. However, compared to other steps of radiomics analysis, ROIs segmentation is often a manual and subjective process. Although automatic or semiautomatic ROIs segmentation methods are available[ 24 ], manual segmentation of ROIs remains the gold standard; this could lead to errors when observers segment ROIs of different species. This study analyzed the effect of the two species, human and experimental monkey, on ICC, and found that experimental monkeys had a slightly lower number of features than humans with intraobserver and interobserver ICCs > 0.75, except intraobserver in feature set B. Therefore, the effect of species on the ROIs segmentation needs to be considered when considering cross-species applications of RMs. Cross-species applications of RMs generally involve data from different machine sources, so the effect of the image acquisition process is an unavoidable but necessary consideration[ 25 ], an issue we also addressed in this study. However, it should be noted that the above results were obtained on different machines and with different scanning parameters, which increases the chance of feature instability but still resulted in a model with good performance. Compared with previous studies, this study is innovative in comparing the reproducibility of RFs between humans and experimental monkeys, providing a theoretical basis for the cross-species application of RM. With the in-depth research and application of radiomics technology, the cross-species application of RMs has great application value. A few previous studies have used animals as subjects when generating RMs. For example, given the invasive, time-consuming, and expensive nature of lung tumor biopsy and its associated complications, Hannah Able[ 26 ] used dogs as subjects and found that the CT RFs had prognostic utility for lung tumors. Anton S. Becker[ 27 ] studied liver metastases using radiomics in mice before and on days 4, 8, 12, 16, and 20 after injection of MC-38 tumor cells, and the analysis revealed that textural features could quantify liver metastases. However, to our knowledge, whether these findings and RMs could be applied to humans has not previously been well established in the literature. Our study has some limitations. Compared with experimental monkeys, the IVD tissue composition of other laboratory animals such as mice, rats, and rabbits, differs more from the tissue composition of humans. These differences could have an impact on the reproducibility of the RFs between species. Therefore, further study is needed to better understand this reproducibility among other species in the future. 5. Conclusion In conclusion, this study revealed that the MRI radiomics features of intervertebral discs exhibit reproducibility across both humans and experimental monkeys, and the corresponding radiomics model could be used interchangeably between the two species. Use of the t test combined with the LASSO method to screen reproducible features between species could improve the performance of the radiomics models. This study thereby furnishes a theoretical framework supporting the cross-species transferability of radiomics models, specifically between humans and experimental monkeys. Abbreviations AUC area under the curve ICC intraclass correlation coefficient IVD intervertebral disc IVDD intervertebral disc degeneration LASSO least absolute shrinkage and selection operator RF radiomics feature RM radiomics model ROC receiver operating characteristic ROI region of interest T1WI T1-weighted imaging T2WI T2-weighted imaging. Declarations Author contributions Jianmin Wang: Conceptualization, Investigation, Writing - original draft. Lei Guo: Conceptualization, Data curation, Project administration, Writing - review & editing. Jianfeng Li: Methodology, Software, Writing - review & editing. Xiaodong Cao: Resources, Software. Wei Du: Supervision, Visualization. Jiaxiang Zhou: Investigation. Haizhen Li: Data curation, Validation. Junhong Li: Investigation. Zhengya Zhu: Methodology. Tao Tang: Validation. Xianlong Li: Visualization. Zhiyu Zhou: Funding acquisition, Investigation. Zhiguo Liu: Project administration, Supervision, Writing - review & editing. Yongming Xi: Conceptualization, Resources, Supervision, Writing - review & editing. Manman Gao: Funding acquisition, Supervision, Visualization. All authors have read and agreed to the published version of the manuscript. Acknowledgements This work was financially supported by the National Natural Science Foundation of China (Nos. U22A20162, 31900583, 32071351, 81772400, 82102604, 81960395), foundation of Shenzhen Committee for Science and Technology Innovation (No. JCYJ202205300150417038), the Beijing Municipal Health Commission (Nos. BMHC-2021-6, BMHC-2019-9, BMHC-2018-4, PXM2020_026275_000002), Key Clinical Specialty Discipline Construction Program of Fuzhou, Fujian, P.R.C (No. 20220104), AO CMF CPP on Bone Regeneration (No. AOCMF-21-04S, supported by AO Foundation, AO CMF. AO CMF is a clinical division of the AO Foundation - an independent medically-guided not-for-profit organization), Academic Affairs Office of Sun Yat-sen University (Nos. 20242043, 20242118, 20242144, 20242162). Code availability Some of the core code generated or used during research is available in repositories or online: https://github.com/wangjm1224/radiomics.git Data availability The raw demographic and MRI data are protected and are not publicly available due to hospital regulations, even all the identification has been removed. Data generated or analyzed during the study are available from the corresponding author by request. Declaration of interest The authors declare no competing interests. References Wang J, Zhu P, Pan X, Yang J, Wang S, et al. Correlation between motor behavior and age-related intervertebral disc degeneration in cynomolgus monkeys. Jor Spine 2022; 5 (1): e1183. https://doi.org/10.1002/jsp2.1183 . Pfirrmann C W, Metzdorf A, Zanetti M, Hodler J, Boos N. Magnetic resonance classification of lumbar intervertebral disc degeneration. Spine 2001; 26 (17): 1873–8. https://doi.org/10.1097/00007632-200109010-00011 . Griffith J F, Wang Y J, Antonio G E, Choi K C, Yu A, et al. Modified pfirrmann grading system for lumbar intervertebral disc degeneration. Spine (Philadelphia, Pa. 1976) 2007; 32 (24): E708-12. https://doi.org/10.1097/BRS.0b013e31815a59a0 . Lambin P, Rios-Velazquez E, Leijenaar R, Carvalho S, van Stiphout R G P M, et al. Radiomics: extracting more information from medical images using advanced feature analysis. Eur. J. Cancer 2012; 48 (4): 441–6. https://doi.org/10.1016/j.ejca.2011.11.036 . Liu Z, Wang S, Dong D, Wei J, Fang C, et al. The applications of radiomics in precision diagnosis and treatment of oncology: opportunities and challenges. Theranostics 2019; 9 (5): 1303–22. https://doi.org/10.7150/thno.30309 . Pedersen C F, Andersen M Ø, Carreon L Y, Eiskjær S. Applied machine learning for spine surgeons: predicting outcome for patients undergoing treatment for lumbar disc herniation using pro data. Glob. Spine J. 2022; 12 (5): 866–76. https://doi.org/10.1177/2192568220967643 . Zhang M Z, Ou Yang H Q, Jiang L, Wang C J, Liu J F, et al. Optimal machine learning methods for radiomic prediction models: clinical application for preoperative t2*-weighted images of cervical spondylotic myelopathy. Jor Spine 2021; 4 (4): e1178. https://doi.org/10.1002/jsp2.1178 . Liu Z, Meng X, Zhang H, Li Z, Liu J, et al. Predicting distant metastasis and chemotherapy benefit in locally advanced rectal cancer. Nat. Commun. 2020; 11 (1): 4308. https://doi.org/10.1038/s41467-020-18162-9 . Feng Z, Li H, Liu Q, Duan J, Zhou W, et al. Ct radiomics to predict macrotrabecular-massive subtype and immune status in hepatocellular carcinoma. Radiology 2022: 221291. https://doi.org/10.1148/radiol.221291 . Mayerhoefer M E, Materka A, Langs G, Häggström I, Szczypiński P, et al. Introduction to radiomics. J. Nucl. Med. 2020; 61 (4): 488–95. https://doi.org/10.2967/jnumed.118.222893 . Park Y W, Han K, Ahn S S, Bae S, Choi Y S, et al. Prediction of idh1-mutation and 1p/19q-codeletion status using preoperative mr imaging phenotypes in lower grade gliomas. Am. J. Neuroradiol. 2018; 39 (1): 37–42. https://doi.org/10.3174/ajnr.A5421 . Tustison N J, Avants B B, Cook P A, Zheng Y, Egan A, et al. N4itk: improved n3 bias correction. Ieee Trans. Med. Imaging 2010; 29 (6): 1310–20. https://doi.org/10.1109/TMI.2010.2046908 . van Griethuysen J J M, Fedorov A, Parmar C, Hosny A, Aucoin N, et al. Computational radiomics system to decode the radiographic phenotype. Cancer Res. 2017; 77 (21): e104-7. https://doi.org/10.1158/0008-5472.CAN-17-0339 . Zwanenburg A, Vallières M, Abdalah M A, Aerts H J W L, Andrearczyk V, et al. The image biomarker standardization initiative: standardized quantitative radiomics for high-throughput image-based phenotyping. Radiology 2020; 295 (2): 328–38. https://doi.org/10.1148/radiol.2020191145 . Scalco E, Belfatto A, Mastropietro A, Rancati T, Avuzzi B, et al. T2w-mri signal normalization affects radiomics features reproducibility. Med. Phys. 2020; 47 (4): 1680–91. https://doi.org/10.1002/mp.14038 . Shafiq Ul Hassan M, Zhang G G, Latifi K, Ullah G, Hunt D C, et al. Intrinsic dependencies of ct radiomic features on voxel size and number of gray levels. Med. Phys. 2017; 44 (3): 1050–62. https://doi.org/10.1002/mp.12123 . Park J E, Park S Y, Kim H J, Kim H S. Reproducibility and generalizability in radiomics modeling: possible strategies in radiologic and statistical perspectives. Korean J. Radiol. 2019; 20 (7): 1124. https://doi.org/10.3348/kjr.2018.0070 . Lee J, Steinmann A, Ding Y, Lee H, Owens C, et al. Radiomics feature robustness as measured using an mri phantom. Sci Rep 2021; 11 (1): 3973. https://doi.org/10.1038/s41598-021-83593-3 . Berenguer R, Pastor-Juan M D R, Canales-Vázquez J, Castro-García M, Villas M V, et al. Radiomics of ct features may be nonreproducible and redundant: influence of ct acquisition parameters. Radiology 2018; 288 (2): 172361–415. https://doi.org/10.1148/radiol.2018172361 . Tibshirani R. Regression shrinkage and selection via the lasso. J. R. Stat. Soc. Ser. B-Stat. Methodol. 1996; 58 (1): 267 – 88. https://doi.org/10.1111/j.2517-6161.1996.tb02080.x . Ford J, Dogan N, Young L, Yang F. Quantitative radiomics: impact of pulse sequence parameter selection on mri-based textural features of the brain. Contrast Media Mol. Imaging 2018; 2018: 1–9. https://doi.org/10.1155/2018/1729071 . Schurink N W, van Kranen S R, Roberti S, van Griethuysen J J M, Bogveradze N, et al. Sources of variation in multicenter rectal mri data and their effect on radiomics feature reproducibility. Eur. Radiol. 2022; 32 (3): 1506–16. https://doi.org/10.1007/s00330-021-08251-8 . Traverso A, Wee L, Dekker A, Gillies R. Repeatability and reproducibility of radiomic features: a systematic review. Int. J. Radiat. Oncol. Biol. Phys. 2018; 102 (4): 1143–58. https://doi.org/10.1016/j.ijrobp.2018.05.053 . Zheng H, Sun Y, Kong D, Yin M, Chen J, et al. Deep learning-based high-accuracy quantitation for lumbar intervertebral disc degeneration from mri. Nat. Commun. 2022; 13 (1): 841. https://doi.org/10.1038/s41467-022-28387-5 . Wang H, Zhou Y, Wang X, Zhang Y, Ma C, et al. Reproducibility and repeatability of cbct-derived radiomics features. Front. Oncol. 2021; 11: 773512. https://doi.org/10.3389/fonc.2021.773512 . Able H, Wolf-Ringwall A, Rendahl A, Ober C P, Seelig D M, et al. Computed tomography radiomic features hold prognostic utility for canine lung tumors: an analytical study. Plos One 2021; 16 (8): e256139. https://doi.org/10.1371/journal.pone.0256139 . Becker A S, Schneider M A, Wurnig M C, Wagner M, Clavien P A, et al. Radiomics of liver mri predict metastases in mice. Eur. Radiol. Exp. 2018; 2 (1): 11. https://doi.org/10.1186/s41747-018-0044-7 . Additional Declarations No competing interests reported. Supplementary Files SupplementaryInformation.docx Cite Share Download PDF Status: Published Journal Publication published 28 Nov, 2025 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Revision requested 01 Oct, 2025 Reviews received at journal 24 Sep, 2025 Reviewers agreed at journal 10 Sep, 2025 Reviewers agreed at journal 08 Feb, 2025 Reviewers agreed at journal 02 Oct, 2024 Reviews received at journal 20 Sep, 2024 Reviewers agreed at journal 16 Sep, 2024 Reviewers invited by journal 06 Aug, 2024 Editor assigned by journal 30 Jul, 2024 Editor invited by journal 24 Jun, 2024 Submission checks completed at journal 24 Jun, 2024 First submitted to journal 27 May, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4486357","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":326415271,"identity":"bdbef9ba-49f6-48a2-afcd-fcd6a25258a6","order_by":0,"name":"Jianmin Wang","email":"","orcid":"","institution":"Yantaishan Hospital, Binzhou Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jianmin","middleName":"","lastName":"Wang","suffix":""},{"id":326415272,"identity":"96c94bf6-8753-4fc5-bead-cfaacd1c21e8","order_by":1,"name":"Lei Guo","email":"","orcid":"","institution":"Yantaishan Hospital, Binzhou Medical University","correspondingAuthor":false,"prefix":"","firstName":"Lei","middleName":"","lastName":"Guo","suffix":""},{"id":326415273,"identity":"68aabae2-46d1-4e84-be3a-ed35ba70f9a1","order_by":2,"name":"Jianfeng Li","email":"","orcid":"","institution":"Sun Yat-sen University","correspondingAuthor":false,"prefix":"","firstName":"Jianfeng","middleName":"","lastName":"Li","suffix":""},{"id":326415274,"identity":"2bfe826f-58ae-4032-be6a-950e9f605684","order_by":3,"name":"Xiaodong Cao","email":"","orcid":"","institution":"South China University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Xiaodong","middleName":"","lastName":"Cao","suffix":""},{"id":326415275,"identity":"e84ae587-339f-44c0-a94d-469f580911a8","order_by":4,"name":"Wei Du","email":"","orcid":"","institution":"Yantaishan Hospital, Binzhou Medical University","correspondingAuthor":false,"prefix":"","firstName":"Wei","middleName":"","lastName":"Du","suffix":""},{"id":326415277,"identity":"b0d38ec9-07c5-4ba5-be68-594cbe51306c","order_by":5,"name":"Jiaxiang Zhou","email":"","orcid":"","institution":"Sun Yat-sen University","correspondingAuthor":false,"prefix":"","firstName":"Jiaxiang","middleName":"","lastName":"Zhou","suffix":""},{"id":326415281,"identity":"6fa47718-cbad-4314-9a8a-6e8719df57ea","order_by":6,"name":"Haizhen Li","email":"","orcid":"","institution":"Sun Yat-sen University","correspondingAuthor":false,"prefix":"","firstName":"Haizhen","middleName":"","lastName":"Li","suffix":""},{"id":326415282,"identity":"5f899a31-8ba0-4743-a6c9-a0cab40a27de","order_by":7,"name":"Junhong Li","email":"","orcid":"","institution":"Affiliated Hospital of Yunnan University","correspondingAuthor":false,"prefix":"","firstName":"Junhong","middleName":"","lastName":"Li","suffix":""},{"id":326415283,"identity":"9e7ba39b-8e37-48e2-841f-7f0199528121","order_by":8,"name":"Zhengya Zhu","email":"","orcid":"","institution":"The Affiliated Hospital of Xuzhou Medical University","correspondingAuthor":false,"prefix":"","firstName":"Zhengya","middleName":"","lastName":"Zhu","suffix":""},{"id":326415284,"identity":"cd467c1d-ca05-4296-90ee-35d2e2fed21d","order_by":9,"name":"Tao Tang","email":"","orcid":"","institution":"The Second Affiliated Hospital of Anhui Medical University","correspondingAuthor":false,"prefix":"","firstName":"Tao","middleName":"","lastName":"Tang","suffix":""},{"id":326415285,"identity":"b49598dc-504e-4a6d-9ac2-e16e01c09ac7","order_by":10,"name":"Xianlong Li","email":"","orcid":"","institution":"Sun Yat-sen University","correspondingAuthor":false,"prefix":"","firstName":"Xianlong","middleName":"","lastName":"Li","suffix":""},{"id":326415286,"identity":"58ff7212-a810-43dd-befa-4a132684a88b","order_by":11,"name":"Zhiyu Zhou","email":"","orcid":"","institution":"Sun Yat-sen University","correspondingAuthor":false,"prefix":"","firstName":"Zhiyu","middleName":"","lastName":"Zhou","suffix":""},{"id":326415287,"identity":"529ebf21-741d-4c13-a43c-0b8249b56115","order_by":12,"name":"Zhiguo Liu","email":"","orcid":"","institution":"CentIaI HospitaI Affiliated to Shandong First Medical University","correspondingAuthor":false,"prefix":"","firstName":"Zhiguo","middleName":"","lastName":"Liu","suffix":""},{"id":326415288,"identity":"5dc8ef2a-1700-44a6-8521-e6dde0e76128","order_by":13,"name":"Yongming Xi","email":"","orcid":"","institution":"The Affiliated Hospital of Qingdao University","correspondingAuthor":false,"prefix":"","firstName":"Yongming","middleName":"","lastName":"Xi","suffix":""},{"id":326415289,"identity":"cb7e8196-2a05-4694-bca0-4482ced3d01b","order_by":14,"name":"Manman Gao","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABFUlEQVRIiWNgGAWjYBADOQYGxsYDH2BcHiK0GAO1NBycQYqWxAYgcRiuEp8WeffDRzf83LEtfW374YYDPAx1ifNnJDA+eNvGIG+OQ4vhmbS0m71nbuduO5PYcECC4XDihhsJzIZz2xgMdzbg0NKQY3aDtw2o5QBQiwHDgcQNEgls0rxtDAkGB3Bo6X9jdvNv2+10s/MPGw4kQBzG/hufFnmJHLPbQFsSzG4AbTnAwJzYcCOBjRmfFgOJZ2m3ZdtuG2678bDhYIPBYeMNZx42S845J2G4AZct/cnHbr5tuy1vdj794eM/FXWy89uTD354U2Yjj9MWVHEDBscGYJwCWRLY1YNsaUATsMepdBSMglEwCkYsAADESWprDvhOgQAAAABJRU5ErkJggg==","orcid":"","institution":"Fuzhou Second Hospital","correspondingAuthor":true,"prefix":"","firstName":"Manman","middleName":"","lastName":"Gao","suffix":""}],"badges":[],"createdAt":"2024-05-27 16:51:52","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4486357/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4486357/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-025-29167-z","type":"published","date":"2025-11-28T15:57:45+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":60618450,"identity":"704f20d6-ceaf-42cf-9f77-6933e4ec64b1","added_by":"auto","created_at":"2024-07-18 20:36:55","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":729806,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA flowchart of the study process. \u003c/strong\u003eMRI data of 720 IVDs are obtained from humans and experimental monkeys. Extract radiomics features, and then \u003cem\u003et\u003c/em\u003e-test combined with LASSO method is used to select features with reproducibility between humans and experimental monkeys. Analyze the distribution characteristics of reproducible features and compare the differences in ICC values between humans and experimental monkeys. Radiomics models are constructed based on the human dataset and the dataset from the experimental monkey is used as a test set to verify the generalizability of the model between species. ICC = intraclass correlation coefficient, IVD = intervertebral disc, ROI = region of interest, T1WI = T1-weighted imaging, T2WI = T2-weighted imaging.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-4486357/v1/cf7e4e7795f7ec013007c70a.png"},{"id":60618451,"identity":"d3916c64-bde2-46a9-8a3a-6b1461a9be83","added_by":"auto","created_at":"2024-07-18 20:36:55","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":816584,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMRI cases of human and experimental monkey.\u003c/strong\u003e (\u003cstrong\u003ea\u003c/strong\u003e-\u003cstrong\u003ec\u003c/strong\u003e) Images in a 32-year-old man human. (\u003cstrong\u003ed\u003c/strong\u003e-\u003cstrong\u003ef\u003c/strong\u003e) Images in a 20-year-old male experimental monkey. (\u003cstrong\u003ea\u003c/strong\u003e) and (\u003cstrong\u003ed\u003c/strong\u003e): T1-weighted imaging. (\u003cstrong\u003eb\u003c/strong\u003e) and (\u003cstrong\u003ee\u003c/strong\u003e): T2-weighted imaging. (\u003cstrong\u003ec\u003c/strong\u003e) and (\u003cstrong\u003ef\u003c/strong\u003e): segmented intervertebral disc regions of interest.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-4486357/v1/4c2a580c164c36e7411b4402.png"},{"id":60619397,"identity":"935bc321-1284-45d0-ad0f-19a147f6d40a","added_by":"auto","created_at":"2024-07-18 20:44:55","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1892136,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDistribution of radiomics features.\u003c/strong\u003e (\u003cstrong\u003ea\u003c/strong\u003e) Heatmap showing all 3562 radiomics features for the 720 human and experimental monkey IVDs. (\u003cstrong\u003eb\u003c/strong\u003e) The radiomics features are distributed across 10 major and 19 minor Image Types. (\u003cstrong\u003ec\u003c/strong\u003e) The radiomics features are distributed into seven Feature Classes. (n=3562) IVD = intervertebral disc, T1WI = T1-weighted imaging, T2WI = T2-weighted imaging.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-4486357/v1/c789baa172b68bdacd5f647a.png"},{"id":60620568,"identity":"6caf4423-6d27-4cab-acc4-b2a35fbe2ffe","added_by":"auto","created_at":"2024-07-18 20:52:55","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":590009,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDistribution pattern of reproducible features between species screened by \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003et\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e test combined with LASSO.\u003c/strong\u003e (\u003cstrong\u003ea\u003c/strong\u003e) Stacked histograms show the distribution of feature set B in terms of Image Type and Feature Class. (\u003cstrong\u003eb-c\u003c/strong\u003e) Proportion of reproducible features in T1WI (\u003cstrong\u003eb\u003c/strong\u003e) and T2WI (\u003cstrong\u003ec\u003c/strong\u003e). Proportion of Reproducible Features = The number of features in each classification of feature set B / The number of features in each classification of feature set A. n=438 in (\u003cstrong\u003ea\u003c/strong\u003e), n=183 in (\u003cstrong\u003eb\u003c/strong\u003e), n=255 in (\u003cstrong\u003ec\u003c/strong\u003e). LASSO = least absolute shrinkage and selection operator, T1WI = T1-weighted imaging, T2WI = T2-weighted imaging.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-4486357/v1/0d5b77fb5b93d196add0a5eb.png"},{"id":60618458,"identity":"b819a66b-7854-4e19-a87c-2a3a20a625b6","added_by":"auto","created_at":"2024-07-18 20:36:55","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":707165,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eICC analysis between humans and experimental monkeys.\u003c/strong\u003e The ICC values of all features in feature set A (\u003cstrong\u003ea\u003c/strong\u003e) and feature set B (\u003cstrong\u003eb\u003c/strong\u003e). Venn diagrams of features with high intra- and interobserver ICCs in the feature set A (\u003cstrong\u003ec\u003c/strong\u003e) and feature set B (\u003cstrong\u003ed\u003c/strong\u003e). (\u003cstrong\u003ee\u003c/strong\u003e) Comparison of the number of features with ICC\u0026gt;0.75 between humans and experimental monkeys. ICC = intraclass correlation coefficient.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-4486357/v1/0bbe018cdaff634da8e3d594.png"},{"id":60619400,"identity":"1482f7a4-e22f-4bc4-9d07-061c10f899f5","added_by":"auto","created_at":"2024-07-18 20:44:55","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":327049,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eEffect of reproducible radiomics features screening on dimensionality reduction. \u003c/strong\u003eThe features and feature coefficients in feature set A2 (\u003cstrong\u003ea\u003c/strong\u003e) and feature set B2 (\u003cstrong\u003eb\u003c/strong\u003e). Comparison of feature values between species for each feature in feature set A2 (\u003cstrong\u003ec\u003c/strong\u003e) and feature set B2 (\u003cstrong\u003ed\u003c/strong\u003e). ns, \u003cem\u003ep\u003c/em\u003e ≥ 0.05; ** \u003cem\u003ep\u003c/em\u003e ＜ 0.001; mean ± standard deviation (SD); n=720. The feature values on the vertical axis have been standardized. The horizontal axis is the feature number. 1132: T1_wavelet-LHL_firstorder_90Percentile; 1160: T1_wavelet-LHL_glcm_Idmn; 1221: T1_wavelet-LHL_ngtdm_Complexity; 1518: T1_wavelet-HHL_firstorder_TotalEnergy; 1808: T2_original_firstorder_RobustMeanAbsoluteDeviation; 1846: T2_original_gldm_LargeDependenceHighGrayLevelEmphasis; 1884: T2_original_ngtdm_Busyness; 2054: T2_gradient_glszm_GrayLevelNonUniformity; 2661: T2_square_glcm_Idm; 2690: T2_square_glrlm_GrayLevelNonUniformityNormalized; 2719: T2_square_glszm_ZonePercentage; 2734: T2_squareroot_firstorder_Mean; 2912: T2_wavelet-LHL_firstorder_10Percentile; 3501: T2_wavelet-LLL_glcm_Imc1; 3502: T2_wavelet-LLL_glcm_Imc2; 3553: T2_wavelet-LLL_glszm_SmallAreaHighGrayLevelEmphasis.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-4486357/v1/485d002aa700475011ce6810.png"},{"id":60619398,"identity":"45072cc3-093f-4316-bf32-b0242d4858f6","added_by":"auto","created_at":"2024-07-18 20:44:55","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":423263,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePerformance measurement of radiomics models.\u003c/strong\u003e (\u003cstrong\u003ea\u003c/strong\u003e-\u003cstrong\u003eb\u003c/strong\u003e) ROC curves of the radiomics models constructed based on feature set A2. (\u003cstrong\u003ec\u003c/strong\u003e-\u003cstrong\u003ed\u003c/strong\u003e) ROC curves of the radiomics models constructed based on feature set B2. (\u003cstrong\u003ea\u003c/strong\u003e) and (\u003cstrong\u003ec\u003c/strong\u003e): training set. (\u003cstrong\u003eb\u003c/strong\u003e) and (\u003cstrong\u003ed\u003c/strong\u003e): test set. AUC = area under the curve, ROC = receiver operating characteristic.\u003c/p\u003e","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-4486357/v1/0019d0775801652ea80b8659.png"},{"id":97179327,"identity":"0d9c3063-d2ee-4a64-9d94-445d1b000b9d","added_by":"auto","created_at":"2025-12-01 16:14:41","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":8654247,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4486357/v1/c0b14a9c-25dd-4858-9fa1-55f0f4dff06f.pdf"},{"id":60618453,"identity":"3b2f2ec9-2b02-4f9c-aa4a-e7a338172857","added_by":"auto","created_at":"2024-07-18 20:36:55","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":720251,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryInformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-4486357/v1/1a96f37fc23438266c11f6aa.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Cross-Species Radiomics: Evaluating the Generalizability of Intervertebral Disc MRI-based Radiomics Models between Humans and Experimental Monkeys","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eExperimental monkey species exhibit astonishing similarities to humans in locomotor behavior patterns, cell composition of the intervertebral disc (IVD), and the progression of intervertebral disc degeneration (IVDD), making them an ideal alternative animal model for studying human IVDD[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Furthermore, the experimental monkey model is the final step in preclinical development of drugs and vaccines, greatly improving the reliability of research results and providing a solid foundation for the eventual clinical translation. Experimental monkeys play an important role in basic and translational biomedical research, serving as a bridge between basic research and clinical medicine.\u003c/p\u003e \u003cp\u003eIn the research of experimental monkey IVDD, diagnostic errors in the degree of IVDD could seriously affect subsequent related research. Consequently, accurate assessment of the degree of IVDD holds great significance for in-depth research into the pathophysiology and therapeutic interventions of the condition. At present, the degree of IVDD is often determined using the Pfirrmann grading[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. However, it has considerable subjectivity and inability to distinguish minor differences in IVDD[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Meanwhile, when graded by different observers, differences of opinion could also arise[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. The pursuit of innovative techniques for accurate assessment of the degree of IVDD holds considerable significance in advancing the research and improving therapeutic strategies targeting IVDD.\u003c/p\u003e \u003cp\u003eAs a significant innovation in medical imaging analysis technology in recent years, radiomics could quantitatively extract and analyze high-throughput radiomics features (RFs) from medical images, providing richer and more accurate information to assist in clinical disease diagnosis and treatment outcome prediction[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Radiomics has shown great value in applications such as diseases diagnosis[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e], prediction of clinical treatment outcomes[\u003cspan additionalcitationids=\"CR7\" citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e], assessment of the pathological heterogeneity of whole tumor tissues[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e], and gene expression prediction[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Therefore, radiomics technology should be adopted to extract high-throughput RFs from IVD imaging data, with the aim of more accurately assessing the degree of IVDD based on these high-throughput RFs.\u003c/p\u003e \u003cp\u003eWhen constructing radiomics models (RMs), a common challenge is the limited sample size, which reduces the robustness and reliability of RMs. Large sample size could enhance the performance of RMs. In the study of experimental monkeys, due to ethical and resource constraints, the number of precious animals is limited, which cannot meet the sample size requirements of radiomics for study subjects, while these data could be obtained more easily in humans. There is a large amount of human IVD data available in clinical practice for constructing RMs. Therefore, we propose a hypothesis that RMs could be constructed based on human IVD data and then applied in experimental monkeys. However, it is unclear whether the RFs of the IVDs between experimental monkey and humans are reproducible and whether the resulting RMs could be applied across species. Cross-species application of the IVD RM has not previously been well established in the literature.\u003c/p\u003e \u003cp\u003eIn the present study, we analyzed the reproducibility of RFs in humans and cynomolgus monkeys using lumbar and lower thoracic IVDs as subjects. We also used the human dataset as a training set to construct RMs to predict the degree of disc degeneration and validated the interspecies generality of the RMs with the experimental monkey dataset. Our objective is to develop a methodology that compensates for the limitation in sample size when constructing RMs, thereby enabling accurate assessment of the degree of IVDD in experimental monkeys. This will pave the way for further experimentation and research concerning IVDD in experimental monkey model. A diagram of this workflow is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"2. Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Study subjects\u003c/h2\u003e \u003cp\u003eThis study included a prospective dataset from cynomolgus monkeys and a retrospective dataset from human volunteers. The experimental protocol was reviewed and approved by the ethics committee of the Institute of Zoology, Guangdong Academy of Sciences (No. G2Z20210103) and the ethics committee of the First Hospital of Sun Yat-sen University and the institutional review board (No. 2008-55). This study was conducted strictly in accordance with ARRIVE guidelines (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://arriveguidelines.org\u003c/span\u003e\u003cspan address=\"https://arriveguidelines.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). All experiments were performed in accordance with relevant guidelines and regulations. Research involving human research participants was conducted in accordance with the Declaration of Helsinki. The written informed consent was obtained from all subjects before enrollment in the study.\u003c/p\u003e \u003cp\u003eThe experimental monkeys completed lumbar MRI in 2021. The human study subjects were volunteers who underwent MRI examinations of the lumbar spine at the First Affiliated Hospital of Sun Yat-sen University from July 2009 to November 2010. The inclusion criteria were as follows: ① Human volunteers who volunteer for research. ② Study subjects had no history of spinal surgery, trauma, and experimental spinal research. The exclusion criteria were as follows: Study subjects with poor image quality.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2. MRI protocol\u003c/h2\u003e \u003cp\u003eHuman MRI data were acquired using a 1.5-Tesla MRI scanner (Philips, United States). The experimental monkey MRIs were all performed using a 3.0-Tesla MRI scanner (General Electric Company, United States). All scans were performed in the supine position. The MRI protocol for the lumbar spine consisted of a sagittal T1WI sequence and T2WI sequence (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). The experimental monkey MRIs were preceded by anesthesia delivered by an experienced veterinarian who was also responsible for animal care.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eParameters for mr imaging of humans and experimental monkeys.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eParameter\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eExperimental Monkey\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003eHuman\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eT1WI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eT2WI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eT1WI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eT2WI\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePulse sequence\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFSE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFSE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSpin echo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eSpin echo\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRepetition time (ms)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e448\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3228\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEcho time (ms)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e102\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e8.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e100\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eField of view (mm)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e20\u0026times;20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20\u0026times;20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e250\u0026times;250\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e250\u0026times;250\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePixel bandwidth (Hz)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e390.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e325.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e320\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e348\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVoxel size (mm)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.8\u0026times;0.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.8\u0026times;0.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.9\u0026times;1.25\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.9\u0026times;1.25\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSlice thickness (mm)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInterslice gap (mm)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.3\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNumber of slices\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTurbo factor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e45\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAcquisition time\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e208\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e160\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"6\"\u003eNote: T1WI\u0026thinsp;=\u0026thinsp;T1-weighted imaging, T2WI\u0026thinsp;=\u0026thinsp;T2-weighted imaging.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3. Pfirrmann grading\u003c/h2\u003e \u003cp\u003eThe degree of disc degeneration was determined according to the criteria described by Pfirrmann et al[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. The grading process was performed independently by 2 experts (Zhiyu Z., and Jianmin W., with 25, 8 years of clinical work experience, respectively) performing 2 analyses each at least 1 week apart. Inconsistencies in the disc grading results were resolved later by discussion among the 2 experts. In the subsequent radiomics analysis, we considered grades Ⅰ-Ⅱ to represent normal discs and marked them as 0 and grades Ⅲ-Ⅴ to represent degenerated discs and marked them as 1.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4. Image data preprocessing and image segmentation.\u003c/h2\u003e \u003cp\u003eWe used the N4ITK Bias Field Correction module (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.slicer.org/wiki/Documentation/Nightly/Modules/N4ITKBiasFieldCorrection\u003c/span\u003e\u003cspan address=\"https://www.slicer.org/wiki/Documentation/Nightly/Modules/N4ITKBiasFieldCorrection\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e)[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] in 3D Slicer software (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.slicer.org/\u003c/span\u003e\u003cspan address=\"https://www.slicer.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, version 4.11.20210226) to perform bias fields corrections.\u003c/p\u003e \u003cp\u003eThe regions of interest (ROIs) were manually segmented in 3D Slicer software in the median sagittal plane of the target disc and included the entire nucleus pulposus, annulus fibrosus, and endplate (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). The ROIs were initially segmented on the T2WI and subsequently replicated on the T1WI by two specialists (Zhiyu Z., J.W.). One of the specialists (Zhiyu Z.) performed 2 analyses, at least 1 week apart; the ROIs outlined the 2nd time were selected for the RF analysis and modeling studies.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.5. Extraction of radiomics features.\u003c/h2\u003e \u003cp\u003ePyRadiomics (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://pyradiomics.readthedocs.io/en/latest/index.html\u003c/span\u003e\u003cspan address=\"https://pyradiomics.readthedocs.io/en/latest/index.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, version 3.0.1)[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e], an Image Biomarker Standardization Initiative (IBSI)[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e] guideline-compliant program, was used to extract the RFs on the images from both the T1WI and T2WI sequences. The images were normalized before feature extraction[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e], then resampled (resampled voxel size set to 1,1,1) with a binWidth of 25. \u0026lsquo;sitk.sitkBSpline\u0026rsquo; was used as an interpolator. All features based on the original image and derived images were extracted. All the features extracted by PyRadiomics were grouped into feature set A.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e2.6. Screening of reproducible radiomics features between humans and experimental monkeys.\u003c/h2\u003e \u003cp\u003eFirst, we used the independent-samples \u003cem\u003et\u003c/em\u003e test to screen for reproducible features between humans and experimental monkeys. The features with \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were removed. We further screened for reproducible RFs between species by LASSO. In this study, we treated features with regression coefficients equal to 0 as those that contributed less to interspecies differences, meaning that they are reproducible features across species. Reproducible features between the species screened with the combination of the independent-samples \u003cem\u003et\u003c/em\u003e test and LASSO to form feature set B.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e2.7. Intraobserver and interobserver agreement.\u003c/h2\u003e \u003cp\u003eThe intraclass correlation coefficient (ICC) of the RFs of different species may be different between different observers and even within the same observer. Here, we used the Pingouin package (version-0.5.1) to calculate the ICC. Intra- and interobserver ICCs were calculated separately for both species. Features with ICCs\u0026thinsp;\u0026gt;\u0026thinsp;0.75 in all tests were considered to have satisfactory agreement[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e2.8. Construction of the radiomics models and validation of model performance across species.\u003c/h2\u003e \u003cp\u003eTo verify whether a RM could be applied with similar efficacy in both humans and experimental monkeys, we constructed RMs based on feature set A, and feature set B of humans. First, the RFs were standardized using z scores. Mutual information (MI) and LASSO were used for dimensionality reduction. We used the human data as a training set to construct RMs to assist in evaluating the degree of IVDD using Support Vector Machine (SVM), Decision Tree Classifier, Random Forest Classifier, Logistic Regression, and Naive Bayes Classifier, respectively. Then, the experimental monkey data were used as the test set to verify the generalizability of the RMs across species. Due to data imbalance, oversampling was performed to balance the data using the imbalanced-learn package (version-0.9.0).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e2.9. Statistical Analysis.\u003c/h2\u003e \u003cp\u003eIBM SPSS Statistics for Windows v.26 (SPSS, Chicago), Python (version-3.7.13), and R (version-4.1.1) were used to perform the statistical analyses. The glmnet R package (version-4.1. 4) was used to perform LASSO analyses. Differences between groups were assessed by the \u003cem\u003et\u003c/em\u003e test (scipy package version 1.7.3) or chi square test. For all analyses, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered significant. Values are expressed as the mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (SD). Model performance was evaluated by receiver operating characteristic (ROC) curve.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e3.1. Participant characteristics.\u003c/h2\u003e \u003cp\u003eThe characteristics of the 720 enrolled IVDs are shown in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. In humans, a total of 90 volunteers underwent MRI and 1 case was excluded because the poor image quality. A total of 89 human volunteers (61 men, 28 women) with a mean age of 31.91 years\u0026thinsp;\u0026plusmn;\u0026thinsp;6.62 (SD) were included. MRI data of 575 IVDs were obtained from human volunteers. The Pfirrmann grading distribution of these IVDs was as follows: grade I-II: 436, grade III-V: 139. Sixteen experimental monkeys completed MRI and were included, with no excluded cases. The enrolled experimental animals had a mean age of 11.81 years\u0026thinsp;\u0026plusmn;\u0026thinsp;4.42 (SD) and a mean body weight was 7.94 kg\u0026thinsp;\u0026plusmn;\u0026thinsp;1.91 (SD). All experimental monkeys were males. MRI data of 145 IVDs were obtained from experimental monkeys. The Pfirrmann grading distribution of these IVDs was as follows: grade I-II: 100, grade III-V: 45.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eParticipant characteristics.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCharacteristic\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTraining Set\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTest Set\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e Value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNumber of participants\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e105\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNumber of IVDs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e720\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e575\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e145\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePfirrmann grades\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.091\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eI-II\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e536\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e436\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIII-V\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e184\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e139\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e45\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge (y)*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e33.91\u0026thinsp;\u0026plusmn;\u0026thinsp;9.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e31.91\u0026thinsp;\u0026plusmn;\u0026thinsp;6.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e41.83\u0026thinsp;\u0026plusmn;\u0026thinsp;15.12\u003csup\u003e#\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;0.001\u003csup\u003e#\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSex\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e538\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e393\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e145\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e182\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e182\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSegments\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT8/9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT9/10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT10/11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT11/12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e57\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e44\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT12/L1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e105\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eL1/2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e105\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eL2/3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e105\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eL3/4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e105\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eL4/5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e105\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eL5/6 \u003csup\u003e\u0026amp;\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e101\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eL6/7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eL7/S1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"5\"\u003eNote: 1) All data are the number of intervertebral discs or participants, except for age. 2) *: Data are means\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviations. 3) #: The age of experimental monkeys has been converted into the age equivalent to that of humans in a ratio of 1:3.5. 4) \u0026amp;: In humans, it is L5/S1. 5) IVD: intervertebral disc.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e3.2. Analysis of radiomics features.\u003c/h2\u003e \u003cp\u003eIn this study, 3562 features (feature set A) were extracted from the MR images of 575 human discs and 145 experimental monkey discs (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea). The distributions of these features in humans and experimental monkeys did not show species differences. The number of features extracted was the same for both the T1-weighted imaging (T1WI) and T2-weighted imaging (T2WI) sequences (Supplementary Fig.\u0026nbsp;1a). The 3562 features were distributed in 10 Image Types, including original image and 9 derived images (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb). The 10 Image Types were: original, wavelet, log, square, squareroot, logarithm, exponential, gradient, LocalBinaryPattern2D (lbp-2D), and LocalBinaryPattern3D (lbp-3D). The wavelet Image Type also included 8 subprojects (HHH, HHL, HLH, HLL, LHH, LHL, LLH, LLL), while the lbp-3D Image Type included 3 subprojects (k, m1, m2). Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb shows the number of features in each Image Type. The features extracted by Pyradiomics were distributed into 7 Feature Classes, namely, First Order Features (firstorder), Shape Features, Gray Level Co-occurrence Matrix (glcm) Features, Gray Level Size Zone Matrix (glszm) Features, Gray Level Run Length Matrix (glrlm) Features, Neighbouring Gray Tone Difference Matrix (ngtdm) Features, and Gray Level Dependence Matrix (gldm) Features. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ec shows the number of features in each Feature Class. Among them, the Shape features were extracted only from the original image; the remaining Feature Class features were extracted from both the original image and derived images. Of all the Feature Classes, the smallest was the Shape class, with 28 features, and the largest was glcm, with 912 features. A total of 214 features were extracted from the original image, and their distributions were identical in both the T1WI and T2WI sequences (Supplementary Fig.\u0026nbsp;1b).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e3.3. Reproducible radiomics features between humans and experimental monkeys.\u003c/h2\u003e \u003cp\u003eA total of 559 features were obtained after removing features with \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 by \u003cem\u003et\u003c/em\u003e test. Furthermore, using least absolute shrinkage and selection operator (LASSO), 438 (feature set B) of these 559 features were found to be reproducible between the species (Supplementary Fig.\u0026nbsp;2a). There were 183 features extracted from the T1WI and 255 features extracted from the T2WI in feature set B (Supplementary Fig.\u0026nbsp;2d). Of the Image Types that were represented in feature set B, the Image Types with the largest number of features was wavelet-HLH (50 of 438, 11.42%), followed by exponential (39 of 438, 8.90%) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea, Supplementary Fig.\u0026nbsp;2b). Among the seven Feature Classes, the Feature Class with the highest number of features was glszm (125 of 438, 28.54%), followed by glrlm (84 of 438, 19.18%) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea, Supplementary Fig.\u0026nbsp;2c).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn feature set B, exponential reproducible features from the T2WI and wavelet-HLH reproducible features from the T1WI accounted for the highest percentages, 33.33% (31 of 93) and 31.18% (29 of 93), respectively. Analysis of the reproducible feature percentages in terms of Feature Class shows the highest value for the glszm features from the T2WIs, at 24.67% (75 of 304) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb, c).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e3.4. Intraclass correlation coefficient analysis between humans and experimental monkeys.\u003c/h2\u003e \u003cp\u003eIntra- and interobserver stability of the RFs between humans and experimental monkeys was determined using intraclass correlation coefficient (ICC) analysis. Figure\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e (a, b) shows the intra- and interobserver ICC values for the 2 species in feature set A, and feature set B. In total, 766 (feature set A1), and 67 features (feature set B1) selected from each feature set that had intra- and interobserver ICCs\u0026thinsp;\u0026gt;\u0026thinsp;0.75 for both species were used for follow-up studies (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ec, d). The proportions of human features with intra- and interobserver ICCs\u0026thinsp;\u0026gt;\u0026thinsp;0.75 in feature set A, and feature set B were higher than the corresponding proportions of experimental monkey features, except intraobserver in feature set B (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ee).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e3.5. Effect of reproducible radiomics features screening on dimensionality reduction.\u003c/h2\u003e \u003cp\u003eNine (feature set A2), and 7 (feature set B2) RFs were obtained after filtering feature set A1, and feature set B1, respectively. The features and corresponding weights occupied of the filtered features are detailed in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e (a, b). The features obtained by dimensionality reduction in two feature sets were mainly from the T2WI sequences, with percentages of 77.78% (7 of 9), and 71.43% (5 of 7), respectively (Supplementary Fig.\u0026nbsp;4a). The Image Types with the highest number of features in feature set A2 was square (3 of 9 [33.33%]), while wavelet-LHL (2 of 7 [28.57%]), wavelet-LLL (2 of 7 [28.57%]), and original (2 of 7 [28.57%]) were the largest in feature set B2 (Supplementary Fig.\u0026nbsp;4b, c). The Feature Classes with the largest number of features in feature set A2 and feature set B2 were glcm (3 of 9 [33.33%]) and firstorder (3 of 7 [42.86%]) (Supplementary Fig.\u0026nbsp;4b, c).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn feature set A2 and feature set B2, the differences in feature values were compared between humans and experimental monkeys. In feature set A2, all feature values were significantly different between humans and experimental monkeys (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ec). However, this difference was not statistically significant in feature set B2 (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026ge;\u0026thinsp;0.05) (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ed). In feature set A2, the differences in feature values between species may impact the performance of the RM.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e3.6. Validation of radiomics models\u0026rsquo; performance across species.\u003c/h2\u003e \u003cp\u003eThe data from the human were used as the training set, and Support Vector Machine (SVM), Decision Tree Classifier, Random Forest Classifier, Logistic Regression, and Naive Bayes Classifier were used to construct RMs to evaluate IVDD. In the training set, the areas under the curve (AUCs) of the five models constructed based on feature set A2 were 0.96, 0.99, 1.00, 0.96, and 0.95, respectively, while the AUCs of the five models constructed based on feature set B2 were 0.93, 0.99, 1.00, 0.92, and 0.89, respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ea, c). Data from experimental monkeys were used as the test set to validate the models\u0026rsquo; performance across species. In the test set, the AUCs of the five models constructed based on feature set A2 were 0.70, 0.51, 0.96, 0.95, and 0.96, respectively, while the AUCs of the five models constructed based on feature set B2 were 0.89, 0.82, 0.88, 0.85, and 0.92, respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eb, d).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn the test set, the AUC values of the RMs constructed based on feature set A2 were higher than that of the RMs constructed based on feature set B2, except for the SVM and Decision Tree Classifier. However, the sensitivity of the first four RMs (SVM, Decision Tree Classifier, Random Forest Classifier, and Logistic Regression) constructed based on feature set A2 performed poorly in identifying degenerative IVDs in experimental monkeys, with 0.36, 0.02, 0.02, and 0.20, respectively (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). In contrast, the sensitivity of the above RMs constructed based on feature set B2 in identifying degenerative IVDs in experimental monkeys was 0.96, 0.82, 0.82, and 0.93, respectively (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe performance comparison of different models.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eFeature Set\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003ePrecision\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003eSensitivity\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eSupport Vector Machine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.79\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.36\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eB2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.61\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.73\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eDecision Tree Classifier\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.51\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.69\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.02\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eB2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.71\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.82\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eRandom Forest Classifier\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.69\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.02\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eB2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.91\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.69\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.82\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eLogistic Regression\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.20\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eB2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.72\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.53\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.63\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.93\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eNaive Bayes Classifier\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.82\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eB2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.68\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.79\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.98\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"9\"\u003eNote: 0\u0026thinsp;=\u0026thinsp;Healthy intervertebral discs, 1\u0026thinsp;=\u0026thinsp;Degenerated intervertebral discs.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe above results show that the RMs could be applied across species, and the use of the \u003cem\u003et\u003c/em\u003e test combined with the LASSO method to screen reproducible features between species could improve the performance of the model.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eThe accurate assessment of the degree of IVDD in experimental monkeys is crucial for further research concerning IVDD in experimental monkeys. Radiomics is an emerging field that extracts high-dimensional quantitative features from medical images, showing promising prospects in enhancing disease representation and diagnosis. When constructing RMs, a large number of datasets are needed to optimize model performance. The shortage of sample size poses significant challenges to the development of robust and reliable RMs. To this end, this study explored the generalizability of radiomics models between humans and experimental monkeys, with a focus on IVD. Here, we analyzed the reproducibility of radiomics features between humans and experimental monkey and found that a total of 12.30% (438/3562) of radiomics features were reproducible between species. Subsequently, we constructed radiomics models based on the human dataset and used the data from the experimental monkey dataset as a testing set to verify the generalizability of the model between species. In the test set, the AUCs of the models constructed based on inter species reproducible features reached 0.82\u0026ndash;0.92. This study provides a theoretical basis for the cross-species application of radiomics.\u003c/p\u003e \u003cp\u003eThrough RFs, it is possible to interpret disease features and understand potential pathological and physiological processes. The biological significance of RFs lies in their ability to quantitatively express macroscopic and microscopic tissue features that are invisible to the naked eye. By combining these features with advanced analytical techniques, it is possible to discover new biomarkers, improve disease classification, and guide personalized treatment strategies, ultimately promoting our understanding of disease mechanisms. The RFs extracted by PyRadiomics include First Order Statistics (first order), Shape based (3D), Shape based (2D), Gray Level Co occurrence Matrix (glcm), Gray Level Run Length Matrix (glrlm), Gray Level Size Zone Matrix (glszm), Neighboring Gray Tone Difference Matrix (ngtdm), and Gray Level Dependence Matrix (gldm).\u003c/p\u003e \u003cp\u003eBy decomposing images into numerous RFs, radiomics could provide multi parameter characterization of the process of degenerative diseases, potentially capturing subtle changes overlooked by traditional visual analysis. This fine-grained evaluation could enhance understanding of IVDD progression and improve diagnostic precision. For example, texture analysis features could reflect the characteristics of organizational microstructure. GLCM represents the statistical patternss of image texture and intensity, where features such as contrast or uniformity could provide a deeper understanding of the randomness and regularity of image grayscale. These indicators could reflect changes in cell density, fibrosis or necrosis, which are known biological indicators of disease progression or response to treatment.\u003c/p\u003e \u003cp\u003eDue to common sample size limitations in experimental monkey research, constructing high-precision radiomics models (RM) directly on experimental monkeys faces challenges. Therefore, we have adopted an innovative strategy: first, we use human rich imaging data to construct radiomics models, thanks to the relatively abundant sample resources in human research. Subsequently, we attempted to apply these models to crab eating monkeys to test their cross species applicability and reproducibility of radiomics features. This research design cleverly bypasses the challenge of sample size. Through comparative analysis, we aim to reveal which RFs are consistent between two species and which features may be influenced by species specificity, laying the foundation for future cross species medical research and disease understanding, and providing strong support for the development of new diagnostic and treatment methods for diseases such as intervertebral disc degeneration.\u003c/p\u003e \u003cp\u003eFeature reproducibility plays an important role in radiomics research[\u003cspan additionalcitationids=\"CR18\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. We investigated the reproducibility of RFs between human and experimental monkey and found that a number of these features were indeed reproducible. We screened 438 features (feature set B) that were reproducible between species from 3562 features by \u003cem\u003et\u003c/em\u003e-test combined with LASSO's method. The number of T2WI features was greater than the number of T1WI features in feature set B (255:183). In feature set B, the Image Types with the highest number of features were wavelet-HLH (50 of 438, 11.42%), and the Feature Classes with the highest number of features were glszm (125 of 438, 28.54%).\u003c/p\u003e \u003cp\u003eRMs constructed based on reproducible RFs could theoretically be applied across species. To verify this speculation, we used the data from the human as the training set and the experimental monkeys\u0026rsquo; data as the test set to evaluate IVDD. In the RMs constructed based on feature sets A, and B, the AUCs in the test set were 0.70, 0.51, 0.96, 0.95, 0.96, and 0.89, 0.82, 0.88, 0.85, 0.92, respectively. This suggests that the RMs constructed based on the human\u0026rsquo;s dataset could be applied in experimental monkeys. At the same time, we also found that some RMs performed poorly in the sensitivity of identifying degenerative IVDs, with values of 0.36, 0.02, 0.02, and 0.20, respectively. However, after removing the non-reproducible features between species, the sensitivity reaches 0.82\u0026ndash;0.96. So, the use of the \u003cem\u003et\u003c/em\u003e test combined with the LASSO method to screen reproducible features between species could improve the performance of the model.\u003c/p\u003e \u003cp\u003eWhen screening for reproducible RFs between species using independent samples \u003cem\u003et\u003c/em\u003e-test, RFs with \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 could be considered to be definitely different between species. However, the opposite is not necessarily true; that is, RFs with \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026ge;\u0026thinsp;0.05 do not necessarily differ between species. LASSO, a regression analysis method used to simultaneously perform feature selection and regularization, was first proposed by Robert Tibshirani in 1996[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. By forcing the sum of the absolute values of the regression coefficients to be less than a fixed value, LASSO forces some regression coefficients to be zero, thus effectively selects simpler models that do not include the covariates corresponding to these regression coefficients. That is, the covariates whose regression coefficients become zero following LASSO play a smaller role in the prediction of the results. In this study, we chose covariates with regression coefficients of 0 as characteristics that are reproducible across species.\u003c/p\u003e \u003cp\u003eThe influence of the reproducibility of RFs may be present in the various steps of radiomics, in addition to its interspecies nature[\u003cspan additionalcitationids=\"CR22\" citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. However, compared to other steps of radiomics analysis, ROIs segmentation is often a manual and subjective process. Although automatic or semiautomatic ROIs segmentation methods are available[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e], manual segmentation of ROIs remains the gold standard; this could lead to errors when observers segment ROIs of different species. This study analyzed the effect of the two species, human and experimental monkey, on ICC, and found that experimental monkeys had a slightly lower number of features than humans with intraobserver and interobserver ICCs\u0026thinsp;\u0026gt;\u0026thinsp;0.75, except intraobserver in feature set B. Therefore, the effect of species on the ROIs segmentation needs to be considered when considering cross-species applications of RMs.\u003c/p\u003e \u003cp\u003eCross-species applications of RMs generally involve data from different machine sources, so the effect of the image acquisition process is an unavoidable but necessary consideration[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e], an issue we also addressed in this study. However, it should be noted that the above results were obtained on different machines and with different scanning parameters, which increases the chance of feature instability but still resulted in a model with good performance.\u003c/p\u003e \u003cp\u003eCompared with previous studies, this study is innovative in comparing the reproducibility of RFs between humans and experimental monkeys, providing a theoretical basis for the cross-species application of RM. With the in-depth research and application of radiomics technology, the cross-species application of RMs has great application value. A few previous studies have used animals as subjects when generating RMs. For example, given the invasive, time-consuming, and expensive nature of lung tumor biopsy and its associated complications, Hannah Able[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e] used dogs as subjects and found that the CT RFs had prognostic utility for lung tumors. Anton S. Becker[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e] studied liver metastases using radiomics in mice before and on days 4, 8, 12, 16, and 20 after injection of MC-38 tumor cells, and the analysis revealed that textural features could quantify liver metastases. However, to our knowledge, whether these findings and RMs could be applied to humans has not previously been well established in the literature.\u003c/p\u003e \u003cp\u003eOur study has some limitations. Compared with experimental monkeys, the IVD tissue composition of other laboratory animals such as mice, rats, and rabbits, differs more from the tissue composition of humans. These differences could have an impact on the reproducibility of the RFs between species. Therefore, further study is needed to better understand this reproducibility among other species in the future.\u003c/p\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eIn conclusion, this study revealed that the MRI radiomics features of intervertebral discs exhibit reproducibility across both humans and experimental monkeys, and the corresponding radiomics model could be used interchangeably between the two species. Use of the \u003cem\u003et\u003c/em\u003e test combined with the LASSO method to screen reproducible features between species could improve the performance of the radiomics models. This study thereby furnishes a theoretical framework supporting the cross-species transferability of radiomics models, specifically between humans and experimental monkeys.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAUC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003earea under the curve\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eICC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eintraclass correlation coefficient\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIVD\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eintervertebral disc\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIVDD\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eintervertebral disc degeneration\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLASSO\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eleast absolute shrinkage and selection operator\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eRF\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eradiomics feature\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eRM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eradiomics model\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eROC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ereceiver operating characteristic\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eROI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eregion of interest\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eT1WI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eT1-weighted imaging\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eT2WI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eT2-weighted imaging.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003eAuthor contributions\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eJianmin Wang:\u0026nbsp;\u003c/strong\u003eConceptualization, Investigation, Writing - original draft. \u003cstrong\u003eLei Guo:\u0026nbsp;\u003c/strong\u003eConceptualization, Data curation, Project administration, Writing - review \u0026amp; editing. \u003cstrong\u003eJianfeng Li:\u0026nbsp;\u003c/strong\u003eMethodology, Software, Writing - review \u0026amp; editing. \u003cstrong\u003eXiaodong Cao:\u0026nbsp;\u003c/strong\u003eResources, Software. \u003cstrong\u003eWei Du:\u0026nbsp;\u003c/strong\u003eSupervision, Visualization. \u003cstrong\u003eJiaxiang Zhou:\u0026nbsp;\u003c/strong\u003eInvestigation. \u003cstrong\u003eHaizhen Li:\u0026nbsp;\u003c/strong\u003eData curation, Validation. \u003cstrong\u003eJunhong Li:\u0026nbsp;\u003c/strong\u003eInvestigation. \u003cstrong\u003eZhengya Zhu:\u0026nbsp;\u003c/strong\u003eMethodology. \u003cstrong\u003eTao Tang:\u0026nbsp;\u003c/strong\u003eValidation. \u003cstrong\u003eXianlong Li:\u0026nbsp;\u003c/strong\u003eVisualization. \u003cstrong\u003eZhiyu Zhou:\u0026nbsp;\u003c/strong\u003eFunding acquisition, Investigation. \u003cstrong\u003eZhiguo Liu:\u0026nbsp;\u003c/strong\u003eProject administration, Supervision, Writing - review \u0026amp; editing. \u003cstrong\u003eYongming Xi:\u0026nbsp;\u003c/strong\u003eConceptualization, Resources, Supervision, Writing - review \u0026amp; editing. \u003cstrong\u003eManman Gao:\u0026nbsp;\u003c/strong\u003eFunding acquisition, Supervision, Visualization. All authors have read and agreed to the published version of the manuscript.\u003c/p\u003e\n\u003cp\u003eAcknowledgements\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThis work was financially supported by the National Natural Science Foundation of China (Nos. U22A20162, 31900583, 32071351, 81772400, 82102604, 81960395), foundation of Shenzhen Committee for Science and Technology Innovation (No. JCYJ202205300150417038), the Beijing Municipal Health Commission (Nos. BMHC-2021-6, BMHC-2019-9, BMHC-2018-4, PXM2020_026275_000002), Key Clinical Specialty Discipline Construction Program of Fuzhou, Fujian, P.R.C (No. 20220104), AO CMF CPP on Bone Regeneration (No. AOCMF-21-04S, supported by AO Foundation, AO CMF. AO CMF is a clinical division of the AO Foundation - an independent medically-guided not-for-profit organization), Academic Affairs Office of Sun Yat-sen University (Nos. 20242043, 20242118, 20242144, 20242162).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCode availability\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSome of the core code generated or used during research is available in repositories or\u003c/p\u003e\n\u003cp\u003eonline: https://github.com/wangjm1224/radiomics.git\u003c/p\u003e\n\u003cp\u003eData availability\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe raw demographic and MRI data are protected and are not publicly available due to hospital regulations, even all the identification has been removed. Data generated or analyzed during the study are available from the corresponding author by request.\u003c/p\u003e\n\u003cp\u003eDeclaration of interest\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eWang J, Zhu P, Pan X, Yang J, Wang S, et al. Correlation between motor behavior and age-related intervertebral disc degeneration in cynomolgus monkeys. Jor Spine 2022; 5 (1): e1183. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/jsp2.1183\u003c/span\u003e\u003cspan address=\"10.1002/jsp2.1183\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePfirrmann C W, Metzdorf A, Zanetti M, Hodler J, Boos N. Magnetic resonance classification of lumbar intervertebral disc degeneration. Spine 2001; 26 (17): 1873\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1097/00007632-200109010-00011\u003c/span\u003e\u003cspan address=\"10.1097/00007632-200109010-00011\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGriffith J F, Wang Y J, Antonio G E, Choi K C, Yu A, et al. Modified pfirrmann grading system for lumbar intervertebral disc degeneration. Spine (Philadelphia, Pa. 1976) 2007; 32 (24): E708-12. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1097/BRS.0b013e31815a59a0\u003c/span\u003e\u003cspan address=\"10.1097/BRS.0b013e31815a59a0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLambin P, Rios-Velazquez E, Leijenaar R, Carvalho S, van Stiphout R G P M, et al. Radiomics: extracting more information from medical images using advanced feature analysis. Eur. J. Cancer 2012; 48 (4): 441\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ejca.2011.11.036\u003c/span\u003e\u003cspan address=\"10.1016/j.ejca.2011.11.036\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu Z, Wang S, Dong D, Wei J, Fang C, et al. The applications of radiomics in precision diagnosis and treatment of oncology: opportunities and challenges. Theranostics 2019; 9 (5): 1303\u0026ndash;22. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.7150/thno.30309\u003c/span\u003e\u003cspan address=\"10.7150/thno.30309\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePedersen C F, Andersen M \u0026Oslash;, Carreon L Y, Eiskj\u0026aelig;r S. Applied machine learning for spine surgeons: predicting outcome for patients undergoing treatment for lumbar disc herniation using pro data. Glob. Spine J. 2022; 12 (5): 866\u0026ndash;76. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/2192568220967643\u003c/span\u003e\u003cspan address=\"10.1177/2192568220967643\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang M Z, Ou Yang H Q, Jiang L, Wang C J, Liu J F, et al. Optimal machine learning methods for radiomic prediction models: clinical application for preoperative t2*-weighted images of cervical spondylotic myelopathy. Jor Spine 2021; 4 (4): e1178. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/jsp2.1178\u003c/span\u003e\u003cspan address=\"10.1002/jsp2.1178\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu Z, Meng X, Zhang H, Li Z, Liu J, et al. Predicting distant metastasis and chemotherapy benefit in locally advanced rectal cancer. Nat. Commun. 2020; 11 (1): 4308. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41467-020-18162-9\u003c/span\u003e\u003cspan address=\"10.1038/s41467-020-18162-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeng Z, Li H, Liu Q, Duan J, Zhou W, et al. Ct radiomics to predict macrotrabecular-massive subtype and immune status in hepatocellular carcinoma. Radiology 2022: 221291. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1148/radiol.221291\u003c/span\u003e\u003cspan address=\"10.1148/radiol.221291\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMayerhoefer M E, Materka A, Langs G, H\u0026auml;ggstr\u0026ouml;m I, Szczypiński P, et al. Introduction to radiomics. J. Nucl. Med. 2020; 61 (4): 488\u0026ndash;95. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2967/jnumed.118.222893\u003c/span\u003e\u003cspan address=\"10.2967/jnumed.118.222893\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePark Y W, Han K, Ahn S S, Bae S, Choi Y S, et al. Prediction of idh1-mutation and 1p/19q-codeletion status using preoperative mr imaging phenotypes in lower grade gliomas. Am. J. Neuroradiol. 2018; 39 (1): 37\u0026ndash;42. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3174/ajnr.A5421\u003c/span\u003e\u003cspan address=\"10.3174/ajnr.A5421\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTustison N J, Avants B B, Cook P A, Zheng Y, Egan A, et al. N4itk: improved n3 bias correction. Ieee Trans. Med. Imaging 2010; 29 (6): 1310\u0026ndash;20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/TMI.2010.2046908\u003c/span\u003e\u003cspan address=\"10.1109/TMI.2010.2046908\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evan Griethuysen J J M, Fedorov A, Parmar C, Hosny A, Aucoin N, et al. Computational radiomics system to decode the radiographic phenotype. Cancer Res. 2017; 77 (21): e104-7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1158/0008-5472.CAN-17-0339\u003c/span\u003e\u003cspan address=\"10.1158/0008-5472.CAN-17-0339\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZwanenburg A, Valli\u0026egrave;res M, Abdalah M A, Aerts H J W L, Andrearczyk V, et al. The image biomarker standardization initiative: standardized quantitative radiomics for high-throughput image-based phenotyping. Radiology 2020; 295 (2): 328\u0026ndash;38. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1148/radiol.2020191145\u003c/span\u003e\u003cspan address=\"10.1148/radiol.2020191145\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eScalco E, Belfatto A, Mastropietro A, Rancati T, Avuzzi B, et al. T2w-mri signal normalization affects radiomics features reproducibility. Med. Phys. 2020; 47 (4): 1680\u0026ndash;91. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/mp.14038\u003c/span\u003e\u003cspan address=\"10.1002/mp.14038\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShafiq Ul Hassan M, Zhang G G, Latifi K, Ullah G, Hunt D C, et al. Intrinsic dependencies of ct radiomic features on voxel size and number of gray levels. Med. Phys. 2017; 44 (3): 1050\u0026ndash;62. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/mp.12123\u003c/span\u003e\u003cspan address=\"10.1002/mp.12123\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePark J E, Park S Y, Kim H J, Kim H S. Reproducibility and generalizability in radiomics modeling: possible strategies in radiologic and statistical perspectives. Korean J. Radiol. 2019; 20 (7): 1124. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3348/kjr.2018.0070\u003c/span\u003e\u003cspan address=\"10.3348/kjr.2018.0070\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee J, Steinmann A, Ding Y, Lee H, Owens C, et al. Radiomics feature robustness as measured using an mri phantom. Sci Rep 2021; 11 (1): 3973. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-021-83593-3\u003c/span\u003e\u003cspan address=\"10.1038/s41598-021-83593-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBerenguer R, Pastor-Juan M D R, Canales-V\u0026aacute;zquez J, Castro-Garc\u0026iacute;a M, Villas M V, et al. Radiomics of ct features may be nonreproducible and redundant: influence of ct acquisition parameters. Radiology 2018; 288 (2): 172361\u0026ndash;415. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1148/radiol.2018172361\u003c/span\u003e\u003cspan address=\"10.1148/radiol.2018172361\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTibshirani R. Regression shrinkage and selection via the lasso. J. R. Stat. Soc. Ser. B-Stat. Methodol. 1996; 58 (1): 267\u0026thinsp;\u0026ndash;\u0026thinsp;88. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/j.2517-6161.1996.tb02080.x\u003c/span\u003e\u003cspan address=\"10.1111/j.2517-6161.1996.tb02080.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFord J, Dogan N, Young L, Yang F. Quantitative radiomics: impact of pulse sequence parameter selection on mri-based textural features of the brain. Contrast Media Mol. Imaging 2018; 2018: 1\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1155/2018/1729071\u003c/span\u003e\u003cspan address=\"10.1155/2018/1729071\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchurink N W, van Kranen S R, Roberti S, van Griethuysen J J M, Bogveradze N, et al. Sources of variation in multicenter rectal mri data and their effect on radiomics feature reproducibility. Eur. Radiol. 2022; 32 (3): 1506\u0026ndash;16. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00330-021-08251-8\u003c/span\u003e\u003cspan address=\"10.1007/s00330-021-08251-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTraverso A, Wee L, Dekker A, Gillies R. Repeatability and reproducibility of radiomic features: a systematic review. Int. J. Radiat. Oncol. Biol. Phys. 2018; 102 (4): 1143\u0026ndash;58. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ijrobp.2018.05.053\u003c/span\u003e\u003cspan address=\"10.1016/j.ijrobp.2018.05.053\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZheng H, Sun Y, Kong D, Yin M, Chen J, et al. Deep learning-based high-accuracy quantitation for lumbar intervertebral disc degeneration from mri. Nat. Commun. 2022; 13 (1): 841. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41467-022-28387-5\u003c/span\u003e\u003cspan address=\"10.1038/s41467-022-28387-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang H, Zhou Y, Wang X, Zhang Y, Ma C, et al. Reproducibility and repeatability of cbct-derived radiomics features. Front. Oncol. 2021; 11: 773512. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fonc.2021.773512\u003c/span\u003e\u003cspan address=\"10.3389/fonc.2021.773512\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAble H, Wolf-Ringwall A, Rendahl A, Ober C P, Seelig D M, et al. Computed tomography radiomic features hold prognostic utility for canine lung tumors: an analytical study. Plos One 2021; 16 (8): e256139. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pone.0256139\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0256139\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBecker A S, Schneider M A, Wurnig M C, Wagner M, Clavien P A, et al. Radiomics of liver mri predict metastases in mice. Eur. Radiol. Exp. 2018; 2 (1): 11. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s41747-018-0044-7\u003c/span\u003e\u003cspan address=\"10.1186/s41747-018-0044-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Radiomics, Radiomics model, Generalizability, Machine learning, Experimental monkey","lastPublishedDoi":"10.21203/rs.3.rs-4486357/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4486357/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eExperimental monkeys serve as a bridge between basic research and clinical medicine. Accurately assessing the degree of intervertebral disc degeneration (IVDD) in experimental monkeys is crucial for further intervertebral disc related research in these animals. Radiomics promises significant enhancement in quantitative diagnostic precision for IVDD, while the cornerstone of constructing robust and efficient radiomics models (RMs) relies on access to large-scale sample data. In experimental monkey research, however, ethical restrictions and resource constraints typically limit sample sizes. This study addresses this challenge by comparing and analyzing the generalizability of intervertebral disc MRI-based radiomics models between humans and experimental monkeys. The findings reveal that 12.30% (438/3562) of the radiomics features demonstrate high reproducibility between the two species. Leveraging the sufficient human dataset, we built RMs and employed the experimental monkey dataset as a training set to validate the cross-species generalizability of these models. Notably, in the test phase, models constructed based on the inter-species reproducible features achieved AUC values ranging from 0.82 to 0.92, indicative of promising diagnostic performance. This study emphasizes the advantages of leveraging human data for the construction of RMs under conditions of constrained experimental monkey research. We innovatively propose and validate the potential for cross-species application of RMs. This study furnishes strong theoretical underpinnings and practical foundations for the broader application of radiomics in cross-species disease research.\u003c/p\u003e","manuscriptTitle":"Cross-Species Radiomics: Evaluating the Generalizability of Intervertebral Disc MRI-based Radiomics Models between Humans and Experimental Monkeys","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-07-18 20:36:50","doi":"10.21203/rs.3.rs-4486357/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-10-01T11:22:28+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-24T17:38:58+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"150444176108582229195934841718795533544","date":"2025-09-10T16:14:52+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"205487229645141903057172813357681897333","date":"2025-02-08T14:44:44+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"165153257257886285874567781408142082449","date":"2024-10-02T09:52:54+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-09-20T13:23:33+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"307486784130320180011444279241480251554","date":"2024-09-16T14:33:14+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-08-06T07:24:35+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-07-30T06:25:08+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2024-06-24T18:27:56+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-06-24T05:48:57+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2024-05-27T16:50:33+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"0ced8b71-aae0-49e3-bd4a-d939fe589dd3","owner":[],"postedDate":"July 18th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":34541314,"name":"Biological sciences/Computational biology and bioinformatics/Machine learning"},{"id":34541315,"name":"Health sciences/Biomarkers/Diagnostic markers"},{"id":34541316,"name":"Biological sciences/Zoology/Animal physiology"}],"tags":[],"updatedAt":"2025-12-01T16:09:39+00:00","versionOfRecord":{"articleIdentity":"rs-4486357","link":"https://doi.org/10.1038/s41598-025-29167-z","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2025-11-28 15:57:45","publishedOnDateReadable":"November 28th, 2025"},"versionCreatedAt":"2024-07-18 20:36:50","video":"","vorDoi":"10.1038/s41598-025-29167-z","vorDoiUrl":"https://doi.org/10.1038/s41598-025-29167-z","workflowStages":[]},"version":"v1","identity":"rs-4486357","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4486357","identity":"rs-4486357","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}