Classifying Intraspinal Tumors: A Multi-modal MRI and Machine Learning Approach 

Classifying Intraspinal Tumors: A Multi-modal MRI and Machine Learning Approach | 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 Classifying Intraspinal Tumors: A Multi-modal MRI and Machine Learning Approach Hu Liyun, Zhang Yuan, Lin Yuanshan, Ma Jie, Yuan Xiaofan This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4624787/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Objective (s):Develop a predictive model for benign and malignant intraspinal tumors using MRI radiomics combined with machine learning, enhancing diagnostic accuracy. Method s : Retrospective data from 132 patients with intraspinal tumors (2020-2023) were analyzed. MRI scans were processed with 3D Slicer and Pyradiomics to extract features. Features were reduced using the LASSO algorithm and recursive elimination, followed by model construction using logistic regression, SVM, DT, RF, and XGBoost. Results: 15 key features were identified post-screening. The XGBoost model showed the highest accuracy, with AUCs of 0.992 (training) and 0.542 (test). RF and DT models also performed well, with AUCs of 0.964/0.885 and 0.986/0.543 respectively. Conclusion(s): Integrating MRI radiomics with machine learning, particularly XGBoost, effectively differentiates intraspinal tumors, offering a non-invasive diagnostic tool that enhances early identification and treatment planning. Biological sciences/Cancer/Cancer imaging Biological sciences/Neuroscience/Neurogenesis Machine Learning Radiomics MRI Intraspinal Space-Occupying Predictive Models Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Intraspinal tumors can be classified into benign and malignant lesions based on their nature and clinical manifestations. The classification of tumors is crucial for the selection of treatment methods [ 1 – 4 ] . Although CT and MRI are the main diagnostic tools, MRI is recommended due to its high detection rate [ 5 – 6 ] . Histopathological examination is the "gold standard" for diagnosis, but it is an invasive procedure with the risk of injury and tumor spread. Traditional diagnosis of intraspinal tumors relies on the experience of physicians, which may lead to missed or misdiagnosed cases. Therefore, the development of an accurate, objective, and non-invasive diagnostic method is particularly important. Radiomics, first proposed by Lambin P et al. [ 7 ] in 2012, can extract quantitative features from medical images to support precise treatment decisions. Moreover, the further application of machine learning makes it possible to obtain valuable clinical information from imaging data. This study uses machine learning technology to analyze MRI radiomics features and establish a predictive and differential diagnosis model for intraspinal tumors. Currently, there are few systematic studies on the combination of radiomics and machine learning systems, and more focus is on traditional routine diagnosis and anatomical structure recognition [ 3 – 6 , 15 – 16 , 18 ] . This study aims to fill this gap, provide support for early diagnosis and precise treatment, and promote the application of radiomics in the diagnosis of intraspinal tumors. Materials and methods 1.1 Research Subjects This retrospective study collected data from patients with intraspinal space-occupying lesions treated at the First Affiliated Hospital of Kunming Medical University between January 2020 and October 2023. Informed consent was waived due to the retrospective nature. The study has been approved by The Department of Orthopedics, The First Affiliated Hospital of Kunming Medical University and The Ethics Committee of The First Affiliated Hospital of Kunming Medical University. All research operations were carried out in accordance with the relevant guidelines and regulations of the hospital department. Patient information was anonymized prior to analysis.Informed consent was waived due to the retrospective nature by ethics committee of the The Ethics Committee of The First Affiliated Hospital of Kunming Medical University. Inclusion criteria: Patients with confirmed pathological diagnoses of intraspinal space-occupying lesions, who underwent spinal MRI at our hospital, and had complete clinical and laboratory records. Exclusion criteria: Patients without intraspinal space-occupying lesions, those with a history of spinal surgery, incomplete data, or unclear pathology. Enrolled patients' data included MRI sagittal plane images, demographic details, lesion characteristics (benign or malignant), sensory and motor function status, and laboratory values (WBC, NEUT%, NEUT#, HGB, ALB, GLB). 1.2 MRI Instruments and Equipment A 3.0 T MRI scanner (Siemens Magnetom Trio, Germany) or a 1.5 T MRI scanner (Siemens Healthcare, Erlangen, Germany) was used, and an 8-channel spinal phased array coil was used to perform routine spinal MRI sequence scanning. 1.3 Image Acquisition All patients were scanned in the supine position, and scans of the three sequences of the conventional axial, sagittal, and coronal positions were performed respectively. The patients were scanned with the conventional sagittal T1-weighted imaging (T1WI) sequence, T2-weighted imaging (T2WI) sequence, and T2-weighted fat-suppressed imaging (T2WI-FS) sequence. The scanning sequences and parameters involved in this study are as follows: T1WI sequence: The sagittal field of view is 320mm×320mm, the slice thickness is 3.0-4.0mm, the slice spacing is 1.0mm, and the matrix is 256×256; T2WI sequence: The sagittal field of view is 320mm×320mm, the slice thickness is 3.0-4.0mm, the slice spacing is 1.0mm, and the matrix is 256×256; T2WI-FS sequence: The sagittal field of view is 320mm×320mm, the slice thickness is 3.0-4.0mm, the slice spacing is 1.0mm, and the matrix is 256×256. All sequences need to ensure that the lesion is completely displayed on the image. 1.4 Image Processing and Data Analysis The workflow of this study is shown in Figure 1. Among them, part A involves the collection and collation of relevant image data, as well as the manual segmentation of the lesion area in the MRI image using the 3D Slicer software. Part B is feature extraction, and the Pyradiomics package of Python 3.7 is used to extract clinical, laboratory and radiomics features. Part C is feature dimension reduction and screening. Feature dimension reduction is carried out through the lasso algorithm, and features are screened using the recursive elimination method. Finally, in part D, we use machine learning algorithms to establish diagnosis and prediction models based on the screened features, and evaluate the performance of the classifier in this study by calculating the AUC, sensitivity, specificity, accuracy, and also the precision, recall rate, F1 value and other indicators of the model on the training set and validation set data. 1.5 Image Preprocessing The N4ITK bias field was corrected using the SimpleITK package of Python 3.7 [8] , and the image preprocessing settings were performed. The voxel size was set to 1x1x1mm3 for resampling using the sitkNearestNeighbor interpolation. 1.6 Lesion Segmentation Processing Collect the spinal MRI plain sagittal T2WI, T2WI-FS, and T1WI images of all enrolled patients, and use all the lesion areas of the sagittal T2WI, T2WI-FS, and T1WI images as the region of interest (ROI), and use the 3D slicer software package version 5.0.3 (http://www.slicer.org/) to perform manual outlining and segmentation layer by layer. See Figure 2. The outlining and segmentation of the ROI are performed by two experienced researchers and cross-checked. When there is a disagreement, discuss with the third researcher to jointly decide. Note: Figure 2A is the sagittal image of T2WI-FS without ROI outlining, Figure 2B is the sagittal image of T2WI without ROI outlining, Figure 2C is the sagittal image of T1WI without ROI outlining, Figure 2D is the sagittal image of T2WI-FS with ROI outlining, Figure 2E is the sagittal image of T2WI with ROI outlining, and Figure 2F is the sagittal image of T1WI with ROI outlining. 1.7 Extraction of Radiomics Features After completing the ROI outlining and segmentation of all the spinal intradural space-occupying images on the sagittal T2WI, T2WI-FS, and T1WI, the Pyradiomics package (https://pyradiomics.readthedocs.io/) is used to extract radiomics features from the ROI. 1.8 Feature Dimensionality Reduction, Screening and Feature Standardization In this study, we employed the lasso regression algorithm for initial feature selection, utilizing L1 regularization to penalize the loss function and eliminate irrelevant features by setting their coefficients to zero [9]. Subsequently, we applied recursive elimination to identify the top 15 most significant features. The selected features were standardized using z-scores to enhance model training accuracy. The processed data were then integrated into our machine learning model for analysis. 1.9 Establishment and Prediction of the Model This study constructs a radiomics model using features extracted from spinal intradural space-occupying lesions' MRI images, combined with clinical and laboratory data. The model employs a five-fold cross-validation method to randomly divide the data into training and test sets. We utilized five machine learning algorithms for model construction: Logistic Regression (LR) for probability estimation. Support Vector Machine (SVM) for maximizing the margin between data classes. Decision Tree (DT) for hierarchical feature-based classification. Random Forest (RF) for ensemble classification, enhancing accuracy. Extreme Gradient Boosting (XGBoost) for optimized prediction by node splitting. 2.0 Statistical Analysis Statistical analysis was performed using Python 3.7, with measurement data presented as mean ± standard deviation. Model prediction thresholds were set at 0.5 and adjusted based on sample proportions. The diagnostic performance of each model for spinal intradural space-occupying lesions was evaluated using ROC curves, calculating AUC, sensitivity, specificity, accuracy, precision, recall, and F1 score. Additionally, the "spearman" rank correlation coefficient was determined for feature interactions, and a heatmap was generated for visualization. Discussion The harmfulness of intraspinal tumors is a medical issue that has received much attention. Some intraspinal space-occupying lesions, such as meningiomas or schwannomas, although most of them are benign, there is still a small part that may undergo malignant transformation. Malignant transformation may lead to an accelerated growth rate and enhanced invasiveness of the tumor, increasing the complexity of treatment and the uncertainty of prognosis [ 10 – 14 ] . Therefore, it is of great significance to improve the detection and diagnosis of intraspinal tumor lesions. Traditionally, CT and MRI are commonly used to detect intraspinal space-occupying lesions, but combined with clinical experience and previous studies, most scholars believe that MRI is more practical [ 15 – 16 ] . In this study, an imagingomics model was constructed through the routine MRI sequence to classify intraspinal space-occupying lesions, but there are still some shortcomings. Firstly, this study is a single-center and retrospective study, and the selection of cases will inevitably have biases. Secondly, the ROI in this study is manually delineated and segmented by hand, which is inevitably affected by the subjective factors of the delineator. In the future, with the improvement of machine learning unsupervised, weakly supervised and self-supervised algorithms and the application of large artificial intelligence models, ROI delineation is expected to shift to the automatic delineation by machines to avoid human interference. Moreover, this study lacks external validation and may have biases and errors. Finally, the number of research samples in this study is limited, and the inclusion of each part of the sample is insufficient, and the representativeness of the population is not good, and the conclusion may have biases. To sum up, it should be admitted that the machine learning based on MRI imagingomics in this study has relatively high application advantages in the application of differential diagnosis of intraspinal space-occupying lesions, and the features extracted through imagingomics can also play an important role in the differential diagnosis of intraspinal space-occupying lesions. Among the machine learning algorithm models, especially the machine learning algorithm constructed by XGBoost, it shows higher robustness in many machine learning competitions and practical applications, and at the same time reduces the risk of overfitting. It can provide a relatively reliable diagnostic basis for the prediction and prognosis assessment of intraspinal space-occupying lesions, and has a good application prospect. The progress of imagingomics combined with machine learning in analyzing diseases marks that medical diagnosis is developing towards a more efficient and more accurate direction. In the future, with the continuous progress of technology and the increasing enrichment of medical data, the machine learning based on MRI imagingomics will play a more critical role in improving the diagnostic accuracy of intraspinal space-occupying lesions and optimizing treatment decisions. Conclusions The features of MRI-based radiomics can be used as predictors of the pathology of intraspinal tumor lesions in patients. Among all the algorithm models, the XGBoost model has strong robustness, high accuracy, and excellent performance, which can provide a relatively reliable diagnostic basis for the prediction and prognosis assessment of intraspinal space-occupying lesions. The combination of radiomics and machine learning analysis has broader advantages than traditional analysis methods, and it can conduct early quantitative analysis to assist in diagnosis and treatment, and it is a non-invasive analysis method. Declarations Statement of conflict of interest All authors declare that there is no conflict of interest in this study, and they have not received any funds, grants or other support during the writing of this manuscript. Funding: There is no funding source. Availability of data and material ： The data used in this study are not publicly available due to privacy concerns. However, they can be made available upon reasonable request to the corresponding author. Code availability ： The code used in this study is not publicly available due to data privacy. However, it can be made available upon reasonable request to the corresponding author. Declaration of AI Usage ： I have polished and modified my article by using artificial intelligence websites such as "https://www.doubao.com/" and "https://kimi.moonshot.cn/". Acknowledgements: Thanks to Zhang Yuan for his knowledge of the research. Author Contribution Hu Liyun collected, analyzed, and initially drafted the paper; Zhang Yuan polished and revised the paper; Lin Yuanshan organized and initially analyzed the paper data; Ma Jie and Yuan Xiaofan participated in the collection of the paper data. References Ahn DK, Park HS, Choi DJ, Kim KS, Kim TW, Park SY. The surgical treatment for spinal intradural extramedullary tumors. Clin Orthop Surg. 2009;1(3):165–72. Li Tiandong, Wang Guoliang, Bai Hongmin, et al. Microsurgical resection of intraspinal meningiomas [J]. Chinese Journal of Minimally Invasive Neurosurgery, 2021, 26(01): 24–27. Wang Jin, Yang Tongtao, Qian Jixian, et al. The clinical diagnosis and surgical effect of intraspinal tumors [J]. Modern Oncology Medicine, 2016, 24(06): 964–967. Li Tiandong, Wang Guoliang, Bai Hongmin, et al. The diagnosis and microsurgical treatment of intraspinal schwannomas [J]. Chinese Journal of Minimally Invasive Neurosurgery, 2021, 26(03): 123–126. Yang Yingde. To observe the application effect of MRI images in the diagnosis of intraspinal space-occupying lesions [J]. China Foreign Medical Treatment, 2016, 35(35): 180–181, 184. Chen Xiangming. Application research of MRI images in the diagnosis of intraspinal space-occupying lesions [J]. Imaging Research and Medical Applications, 2019, 3(09): 235–236. Lambin P, Rios-Velazquez E, Leijenaar R, et al. Radiomics:extracting more information from medical images using advanced feature analysis[J]. Eur J Cancer, 2012, 48(4): 441–446. Tustison NJ, Avants BB, Cook PA, Zheng Y, Egan A, Yushkevich PA, Gee JC. N4ITK: improved N3 bias correction. IEEE Trans Med Imaging. 2010;29(6):1310–20. Daghir-Wojtkowiak E, Wiczling P, Bocian S, et al. Least absolute shrinkage and selection operator and dimensionality reduction techniques in quantitative structure retention relationship modeling of retention in hydrophilic interaction liquid chromatography[J]. J Chromatogr A, 2015, 1403: 54–62. Fehlings MG, Tetreault L, Nater A, et al. The Aging of the Global Population: The Changing Epidemiology of Disease and Spinal Disorders. Neurosurgery. 2015;77 Suppl 4:S1-S5. Fourney DR. Management of Intramedullary Spinal Cord Tumors. Prog Neurol Surg. 2018;32:102–110. Barbagallo GMV, Branca G, Certo F, Albanese V. Intradural Extramedullary Spinal Tumors: A Review of Literature [published correction appears in Asian J Neurosurg. 2020 Jan-Mar;15(1):223]. Asian J Neurosurg. 2019;14(3):586–594. Batzdorf U. Spinal Cord Compression: Mechanisms and Diagnosis. San Rafael (CA): Morgan & Claypool Life Sciences; 2010. Chapter 10. Lee YS, Kim SW, Son BC, Lee JY, Joo SP. Surgical Outcome and Prognosis of Intramedullary Spinal Cord Tumors: Comparison between Astrocytoma and Ependymoma. J Kor Neurosurg Soc. 2009;46(3):206–210. Liu Ting. Analysis of the clinical application effect of MRI images in the diagnosis of intraspinal space-occupying lesions [J]. Guide of China Medicine, 2018, 16(25): 116–117. Gao Dahai. To observe the application effect of MRI images in the diagnosis of intraspinal space-occupying lesions [J]. Guide of China Medicine, 2017, 15(25): 178. Zhao Lanfeng, Wang Zhenge, Ma Guohua, et al. MRI differential diagnosis of intraspinal atypical schwannoma and atypical meningioma [J]. Modern Oncology Medicine, 2023, 31(05): 912–915. Zhao Lanfeng, Wang Zhenge. The ratio of tumor-subcutaneous fat signal intensity on enhanced MRI for differentiating intraspinal schwannoma from meningioma [J]. Chinese Journal of Medical Imaging Technology, 2022, 38(01): 49–52. Table Table 6 is available in the Supplementary Files section. Additional Declarations No competing interests reported. Supplementary Files Table6ROCcurvesandlearningcurvesoffivedifferentalgorithmmodels.docx Cite Share Download PDF Status: Posted Version 1 posted 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-4624787","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":326710065,"identity":"97d037b2-eca8-496c-8732-52fc4982d06c","order_by":0,"name":"Hu Liyun","email":"","orcid":"","institution":"The First Affiliated Hospital of Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"Hu","middleName":"","lastName":"Liyun","suffix":""},{"id":326710066,"identity":"1a25ceaa-2145-448d-b7df-337e90d22b57","order_by":1,"name":"Zhang Yuan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5ElEQVRIiWNgGAWjYDACCRiDvQ3KOEC0Fp5jJGuRSCNSi/zs5mcPv7YdljeXfJa66WYbgxzfjQTGzwV4tDDOOWZuLNt22HDn7LRjt3PbGIwlbyQwS8/Ao4VZIsFMWrLtMOOG2+ltIC2JG24ksDHz4NHCJpH+DaTFfsPN42At9QS18EjkmEl+bDsMNJwN7LAEA0JaJCRyyqQZzqUnbziTlnY755yE4cwzD5ul8WmRn5G+TfJHmbXthuPHzG7nlNnI8x1PPvgZnxYQYOZlQ9gKxIwNBDQAlfz4Q1DNKBgFo2AUjGQAAH6DTx+yGm1ZAAAAAElFTkSuQmCC","orcid":"","institution":"The First Affiliated Hospital of Kunming Medical University","correspondingAuthor":true,"prefix":"","firstName":"Zhang","middleName":"","lastName":"Yuan","suffix":""},{"id":326710067,"identity":"d3d8ce81-9780-4c07-9023-e829fff10527","order_by":2,"name":"Lin Yuanshan","email":"","orcid":"","institution":"The First Affiliated Hospital of Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"Lin","middleName":"","lastName":"Yuanshan","suffix":""},{"id":326710068,"identity":"e86b4fa6-698e-4adf-8090-a6117dd72a38","order_by":3,"name":"Ma Jie","email":"","orcid":"","institution":"The First Affiliated Hospital of Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"Ma","middleName":"","lastName":"Jie","suffix":""},{"id":326710069,"identity":"bcfab738-d4fe-4448-94c1-7d8650d50f6d","order_by":4,"name":"Yuan Xiaofan","email":"","orcid":"","institution":"The First Affiliated Hospital of Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yuan","middleName":"","lastName":"Xiaofan","suffix":""}],"badges":[],"createdAt":"2024-06-23 10:09:51","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4624787/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4624787/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":60712474,"identity":"06733553-04aa-4f4c-bfce-49c85dcad804","added_by":"auto","created_at":"2024-07-19 20:25:54","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":171952,"visible":true,"origin":"","legend":"\u003cp\u003eThe workflow chart of this study.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-4624787/v1/49e955609c3cb3bba6ee1161.png"},{"id":60711343,"identity":"3d177d22-819c-4bd6-9115-ccc074a90e35","added_by":"auto","created_at":"2024-07-19 20:17:54","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":457485,"visible":true,"origin":"","legend":"\u003cp\u003eComparison before and after manual outlining of the region of interest on MR T2WI-FS, T2WI, and T1WI.\u003c/p\u003e\n\u003cp\u003eNote: Figure 2A is the sagittal image of T2WI-FS without ROI outlining, Figure 2B is the sagittal image of T2WI without ROI outlining, Figure 2C is the sagittal image of T1WI without ROI outlining, Figure 2D is the sagittal image of T2WI-FS with ROI outlining, Figure 2E is the sagittal image of T2WI with ROI outlining, and Figure 2F is the sagittal image of T1WI with ROI outlining.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-4624787/v1/74918116812591dd29aed478.png"},{"id":60712473,"identity":"8461e8dd-5601-4a8b-bf47-c051c420f082","added_by":"auto","created_at":"2024-07-19 20:25:54","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":114206,"visible":true,"origin":"","legend":"\u003cp\u003eGeneral process of feature screening when differentiating the benign and malignant of spinal intradural space-occupying tumor.\u003c/p\u003e\n\u003cp\u003eNote: In Figures 3, A is based on the MSE path, the horizontal axis represents the log value of Lambda, and the vertical axis represents the mean square error; B is based on the Lasso path, the horizontal axis represents the log value of Lambda, and the vertical axis represents the feature coefficients under different fitting methods; C uses the lasso algorithm and the recursive elimination method to select the 15 features with the largest weight. The horizontal axis represents the screened features, and the vertical axis represents the feature weights.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-4624787/v1/e3094aa8493436ffb49b944f.png"},{"id":60711341,"identity":"32491aeb-d2da-4898-9b6d-8fb2dd4f94bd","added_by":"auto","created_at":"2024-07-19 20:17:54","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":157701,"visible":true,"origin":"","legend":"\u003cp\u003eSpearman's rank correlation coefficient heatmap of features in differentiating the benign and malignant of the spinal intradural space-occupying tumor.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-4624787/v1/fef6b73da2046f6ead7028a6.png"},{"id":92163695,"identity":"8c392815-71f2-4852-a9bc-83f060ba0ab9","added_by":"auto","created_at":"2025-09-25 10:32:26","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1461414,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4624787/v1/76d75a1b-6026-4641-96de-2e3e5bdda682.pdf"},{"id":60711345,"identity":"ff197250-c836-4810-98a3-82e2848c4368","added_by":"auto","created_at":"2024-07-19 20:17:54","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":255734,"visible":true,"origin":"","legend":"","description":"","filename":"Table6ROCcurvesandlearningcurvesoffivedifferentalgorithmmodels.docx","url":"https://assets-eu.researchsquare.com/files/rs-4624787/v1/d3808eb4498194d661a3b5be.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Classifying Intraspinal Tumors: A Multi-modal MRI and Machine Learning Approach ","fulltext":[{"header":"Introduction","content":"\u003cp\u003eIntraspinal tumors can be classified into benign and malignant lesions based on their nature and clinical manifestations. The classification of tumors is crucial for the selection of treatment methods \u003csup\u003e[\u003cspan additionalcitationids=\"CR2 CR3\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/sup\u003e. Although CT and MRI are the main diagnostic tools, MRI is recommended due to its high detection rate \u003csup\u003e[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/sup\u003e. Histopathological examination is the \"gold standard\" for diagnosis, but it is an invasive procedure with the risk of injury and tumor spread. Traditional diagnosis of intraspinal tumors relies on the experience of physicians, which may lead to missed or misdiagnosed cases. Therefore, the development of an accurate, objective, and non-invasive diagnostic method is particularly important. Radiomics, first proposed by Lambin P et al. \u003csup\u003e[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e in 2012, can extract quantitative features from medical images to support precise treatment decisions. Moreover, the further application of machine learning makes it possible to obtain valuable clinical information from imaging data. This study uses machine learning technology to analyze MRI radiomics features and establish a predictive and differential diagnosis model for intraspinal tumors. Currently, there are few systematic studies on the combination of radiomics and machine learning systems, and more focus is on traditional routine diagnosis and anatomical structure recognition \u003csup\u003e[\u003cspan additionalcitationids=\"CR4 CR5\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]\u003c/sup\u003e. This study aims to fill this gap, provide support for early diagnosis and precise treatment, and promote the application of radiomics in the diagnosis of intraspinal tumors.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cp\u003e\u003cstrong\u003e1.1 Research Subjects\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThis retrospective study collected data from patients with intraspinal space-occupying lesions treated at the First Affiliated Hospital of Kunming Medical University between January 2020 and October 2023. Informed consent was waived due to the retrospective nature. The study has been approved by The Department of Orthopedics, The First Affiliated Hospital of Kunming Medical University and The Ethics Committee of The First Affiliated Hospital of Kunming Medical University. All research operations were carried out in accordance with the relevant guidelines and regulations of the hospital department. Patient information was anonymized prior to analysis.Informed consent was waived due to the retrospective nature by ethics committee of the The Ethics Committee of The First Affiliated Hospital of Kunming Medical University.\u003c/p\u003e\n\u003cp\u003eInclusion criteria: Patients with confirmed pathological diagnoses of intraspinal space-occupying lesions, who underwent spinal MRI at our hospital, and had complete clinical and laboratory records.\u003c/p\u003e\n\u003cp\u003eExclusion criteria: Patients without intraspinal space-occupying lesions, those with a history of spinal surgery, incomplete data, or unclear pathology.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eEnrolled patients\u0026apos; data included MRI sagittal plane images, demographic details, lesion characteristics (benign or malignant), sensory and motor function status, and laboratory values (WBC, NEUT%, NEUT#, HGB, ALB, GLB).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e1.2 MRI Instruments and Equipment\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA 3.0 T MRI scanner (Siemens Magnetom Trio, Germany) or a 1.5 T MRI scanner (Siemens Healthcare, Erlangen, Germany) was used, and an 8-channel spinal phased array coil was used to perform routine spinal MRI sequence scanning.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e1.3 Image Acquisition\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll patients were scanned in the supine position, and scans of the three sequences of the conventional axial, sagittal, and coronal positions were performed respectively. The patients were scanned with the conventional sagittal T1-weighted imaging (T1WI) sequence, T2-weighted imaging (T2WI) sequence, and T2-weighted fat-suppressed imaging (T2WI-FS) sequence. The scanning sequences and parameters involved in this study are as follows: T1WI sequence: The sagittal field of view is 320mm\u0026times;320mm, the slice thickness is 3.0-4.0mm, the slice spacing is 1.0mm, and the matrix is 256\u0026times;256; T2WI sequence: The sagittal field of view is 320mm\u0026times;320mm, the slice thickness is 3.0-4.0mm, the slice spacing is 1.0mm, and the matrix is 256\u0026times;256; T2WI-FS sequence: The sagittal field of view is 320mm\u0026times;320mm, the slice thickness is 3.0-4.0mm, the slice spacing is 1.0mm, and the matrix is 256\u0026times;256. All sequences need to ensure that the lesion is completely displayed on the image.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e1.4 Image Processing and Data Analysis\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe workflow of this study is shown in Figure 1. Among them, part A involves the collection and collation of relevant image data, as well as the manual segmentation of the lesion area in the MRI image using the 3D Slicer software. Part B is feature extraction, and the Pyradiomics package of Python 3.7 is used to extract clinical, laboratory and radiomics features. Part C is feature dimension reduction and screening. Feature dimension reduction is carried out through the lasso algorithm, and features are screened using the recursive elimination method. Finally, in part D, we use machine learning algorithms to establish diagnosis and prediction models based on the screened features, and evaluate the performance of the classifier in this study by calculating the AUC, sensitivity, specificity, accuracy, and also the precision, recall rate, F1 value and other indicators of the model on the training set and validation set data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e1.5 Image Preprocessing\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe N4ITK bias field was corrected using the SimpleITK package of Python 3.7\u003csup\u003e[8]\u003c/sup\u003e, and the image preprocessing settings were performed. The voxel size was set to 1x1x1mm3 for resampling using the sitkNearestNeighbor interpolation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e1.6 Lesion Segmentation Processing\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCollect the spinal MRI plain sagittal T2WI, T2WI-FS, and T1WI images of all enrolled patients, and use all the lesion areas of the sagittal T2WI, T2WI-FS, and T1WI images as the region of interest (ROI), and use the 3D slicer software package version 5.0.3 (http://www.slicer.org/) to perform manual outlining and segmentation layer by layer. See Figure 2. The outlining and segmentation of the ROI are performed by two experienced researchers and cross-checked. When there is a disagreement, discuss with the third researcher to jointly decide.\u003c/p\u003e\n\u003cp\u003eNote: Figure 2A is the sagittal image of T2WI-FS without ROI outlining, Figure 2B is the sagittal image of T2WI without ROI outlining, Figure 2C is the sagittal image of T1WI without ROI outlining, Figure 2D is the sagittal image of T2WI-FS with ROI outlining, Figure 2E is the sagittal image of T2WI with ROI outlining, and Figure 2F is the sagittal image of T1WI with ROI outlining.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e1.7 Extraction of Radiomics Features\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAfter completing the ROI outlining and segmentation of all the spinal intradural space-occupying images on the sagittal T2WI, T2WI-FS, and T1WI, the Pyradiomics package (https://pyradiomics.readthedocs.io/) is used to extract radiomics features from the ROI.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e1.8 Feature Dimensionality Reduction, Screening and Feature Standardization\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn this study, we employed the lasso regression algorithm for initial feature selection, utilizing L1 regularization to penalize the loss function and eliminate irrelevant features by setting their coefficients to zero [9]. Subsequently, we applied recursive elimination to identify the top 15 most significant features. The selected features were standardized using z-scores to enhance model training accuracy. The processed data were then integrated into our machine learning model for analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e1.9 Establishment and Prediction of the Model\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study constructs a radiomics model using features extracted from spinal intradural space-occupying lesions\u0026apos; MRI images, combined with clinical and laboratory data. The model employs a five-fold cross-validation method to randomly divide the data into training and test sets.\u003c/p\u003e\n\u003cp\u003eWe utilized five machine learning algorithms for model construction:\u003c/p\u003e\n\u003col start=\"1\" type=\"1\"\u003e\n \u003cli\u003eLogistic Regression (LR) for probability estimation.\u003c/li\u003e\n \u003cli\u003eSupport Vector Machine (SVM) for maximizing the margin between data classes.\u003c/li\u003e\n \u003cli\u003eDecision Tree (DT) for hierarchical feature-based classification.\u003c/li\u003e\n \u003cli\u003eRandom Forest (RF) for ensemble classification, enhancing accuracy.\u003c/li\u003e\n \u003cli\u003eExtreme Gradient Boosting (XGBoost) for optimized prediction by node splitting.\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003e\u003cstrong\u003e2.0 Statistical Analysis\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eStatistical analysis was performed using Python 3.7, with measurement data presented as mean \u0026plusmn; standard deviation. Model prediction thresholds were set at 0.5 and adjusted based on sample proportions. The diagnostic performance of each model for spinal intradural space-occupying lesions was evaluated using ROC curves, calculating AUC, sensitivity, specificity, accuracy, precision, recall, and F1 score. Additionally, the \u0026quot;spearman\u0026quot; rank correlation coefficient was determined for feature interactions, and a heatmap was generated for visualization.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe harmfulness of intraspinal tumors is a medical issue that has received much attention. Some intraspinal space-occupying lesions, such as meningiomas or schwannomas, although most of them are benign, there is still a small part that may undergo malignant transformation. Malignant transformation may lead to an accelerated growth rate and enhanced invasiveness of the tumor, increasing the complexity of treatment and the uncertainty of prognosis \u003csup\u003e[\u003cspan additionalcitationids=\"CR11 CR12 CR13\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/sup\u003e. Therefore, it is of great significance to improve the detection and diagnosis of intraspinal tumor lesions. Traditionally, CT and MRI are commonly used to detect intraspinal space-occupying lesions, but combined with clinical experience and previous studies, most scholars believe that MRI is more practical \u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn this study, an imagingomics model was constructed through the routine MRI sequence to classify intraspinal space-occupying lesions, but there are still some shortcomings. Firstly, this study is a single-center and retrospective study, and the selection of cases will inevitably have biases. Secondly, the ROI in this study is manually delineated and segmented by hand, which is inevitably affected by the subjective factors of the delineator. In the future, with the improvement of machine learning unsupervised, weakly supervised and self-supervised algorithms and the application of large artificial intelligence models, ROI delineation is expected to shift to the automatic delineation by machines to avoid human interference. Moreover, this study lacks external validation and may have biases and errors. Finally, the number of research samples in this study is limited, and the inclusion of each part of the sample is insufficient, and the representativeness of the population is not good, and the conclusion may have biases.\u003c/p\u003e \u003cp\u003eTo sum up, it should be admitted that the machine learning based on MRI imagingomics in this study has relatively high application advantages in the application of differential diagnosis of intraspinal space-occupying lesions, and the features extracted through imagingomics can also play an important role in the differential diagnosis of intraspinal space-occupying lesions. Among the machine learning algorithm models, especially the machine learning algorithm constructed by XGBoost, it shows higher robustness in many machine learning competitions and practical applications, and at the same time reduces the risk of overfitting. It can provide a relatively reliable diagnostic basis for the prediction and prognosis assessment of intraspinal space-occupying lesions, and has a good application prospect. The progress of imagingomics combined with machine learning in analyzing diseases marks that medical diagnosis is developing towards a more efficient and more accurate direction. In the future, with the continuous progress of technology and the increasing enrichment of medical data, the machine learning based on MRI imagingomics will play a more critical role in improving the diagnostic accuracy of intraspinal space-occupying lesions and optimizing treatment decisions.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eThe features of MRI-based radiomics can be used as predictors of the pathology of intraspinal tumor lesions in patients.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eAmong all the algorithm models, the XGBoost model has strong robustness, high accuracy, and excellent performance, which can provide a relatively reliable diagnostic basis for the prediction and prognosis assessment of intraspinal space-occupying lesions.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eThe combination of radiomics and machine learning analysis has broader advantages than traditional analysis methods, and it can conduct early quantitative analysis to assist in diagnosis and treatment, and it is a non-invasive analysis method.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003e\u003cem\u003eStatement of conflict of interest\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors declare that there is no conflict of interest in this study, and they have not received any funds, grants or other support during the writing of this manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eFunding:\u0026nbsp;\u003c/em\u003e\u003c/strong\u003eThere is no funding source.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eAvailability of data and material\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u003cem\u003e：\u003c/em\u003e\u003c/strong\u003eThe data used in this study are not publicly available due to privacy concerns. However, they can be made available upon reasonable request to the corresponding author.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eCode availability\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u003cem\u003e：\u003c/em\u003e\u003c/strong\u003eThe code used in this study is not publicly available due to data privacy. However, it can be made available upon reasonable request to the corresponding author.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eDeclaration of AI Usage\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u003cem\u003e：\u003c/em\u003e\u003c/strong\u003eI have polished and modified my article by using artificial intelligence websites such as \u0026quot;https://www.doubao.com/\u0026quot; and \u0026quot;https://kimi.moonshot.cn/\u0026quot;.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eAcknowledgements:\u0026nbsp;\u003c/em\u003e\u003c/strong\u003eThanks to Zhang Yuan for his knowledge of the research.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eHu Liyun collected, analyzed, and initially drafted the paper; Zhang Yuan polished and revised the paper; Lin Yuanshan organized and initially analyzed the paper data; Ma Jie and Yuan Xiaofan participated in the collection of the paper data.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAhn DK, Park HS, Choi DJ, Kim KS, Kim TW, Park SY. The surgical treatment for spinal intradural extramedullary tumors. Clin Orthop Surg. 2009;1(3):165\u0026ndash;72.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi Tiandong, Wang Guoliang, Bai Hongmin, et al. Microsurgical resection of intraspinal meningiomas [J]. Chinese Journal of Minimally Invasive Neurosurgery, 2021, 26(01): 24\u0026ndash;27.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang Jin, Yang Tongtao, Qian Jixian, et al. The clinical diagnosis and surgical effect of intraspinal tumors [J]. Modern Oncology Medicine, 2016, 24(06): 964\u0026ndash;967.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi Tiandong, Wang Guoliang, Bai Hongmin, et al. The diagnosis and microsurgical treatment of intraspinal schwannomas [J]. Chinese Journal of Minimally Invasive Neurosurgery, 2021, 26(03): 123\u0026ndash;126.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang Yingde. To observe the application effect of MRI images in the diagnosis of intraspinal space-occupying lesions [J]. China Foreign Medical Treatment, 2016, 35(35): 180\u0026ndash;181, 184.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen Xiangming. Application research of MRI images in the diagnosis of intraspinal space-occupying lesions [J]. Imaging Research and Medical Applications, 2019, 3(09): 235\u0026ndash;236.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLambin P, Rios-Velazquez E, Leijenaar R, et al. Radiomics:extracting more information from medical images using advanced feature analysis[J]. Eur J Cancer, 2012, 48(4): 441\u0026ndash;446.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTustison NJ, Avants BB, Cook PA, Zheng Y, Egan A, Yushkevich PA, Gee JC. N4ITK: improved N3 bias correction. IEEE Trans Med Imaging. 2010;29(6):1310\u0026ndash;20.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDaghir-Wojtkowiak E, Wiczling P, Bocian S, et al. Least absolute shrinkage and selection operator and dimensionality reduction techniques in quantitative structure retention relationship modeling of retention in hydrophilic interaction liquid chromatography[J]. J Chromatogr A, 2015, 1403: 54\u0026ndash;62.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFehlings MG, Tetreault L, Nater A, et al. The Aging of the Global Population: The Changing Epidemiology of Disease and Spinal Disorders. Neurosurgery. 2015;77 Suppl 4:S1-S5.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFourney DR. Management of Intramedullary Spinal Cord Tumors. Prog Neurol Surg. 2018;32:102\u0026ndash;110.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarbagallo GMV, Branca G, Certo F, Albanese V. Intradural Extramedullary Spinal Tumors: A Review of Literature [published correction appears in Asian J Neurosurg. 2020 Jan-Mar;15(1):223]. Asian J Neurosurg. 2019;14(3):586\u0026ndash;594.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBatzdorf U. Spinal Cord Compression: Mechanisms and Diagnosis. San Rafael (CA): Morgan \u0026amp; Claypool Life Sciences; 2010. Chapter 10.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee YS, Kim SW, Son BC, Lee JY, Joo SP. Surgical Outcome and Prognosis of Intramedullary Spinal Cord Tumors: Comparison between Astrocytoma and Ependymoma. J Kor Neurosurg Soc. 2009;46(3):206\u0026ndash;210.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu Ting. Analysis of the clinical application effect of MRI images in the diagnosis of intraspinal space-occupying lesions [J]. Guide of China Medicine, 2018, 16(25): 116\u0026ndash;117.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGao Dahai. To observe the application effect of MRI images in the diagnosis of intraspinal space-occupying lesions [J]. Guide of China Medicine, 2017, 15(25): 178.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao Lanfeng, Wang Zhenge, Ma Guohua, et al. MRI differential diagnosis of intraspinal atypical schwannoma and atypical meningioma [J]. Modern Oncology Medicine, 2023, 31(05): 912\u0026ndash;915.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao Lanfeng, Wang Zhenge. The ratio of tumor-subcutaneous fat signal intensity on enhanced MRI for differentiating intraspinal schwannoma from meningioma [J]. Chinese Journal of Medical Imaging Technology, 2022, 38(01): 49\u0026ndash;52.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Table","content":"\u003cp\u003eTable 6 is available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Machine Learning,Radiomics,MRI,Intraspinal Space-Occupying,Predictive Models","lastPublishedDoi":"10.21203/rs.3.rs-4624787/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4624787/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eObjective \u003c/strong\u003e(s):Develop a predictive model for benign and malignant intraspinal tumors using MRI radiomics combined with machine learning, enhancing diagnostic accuracy.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethod\u003c/strong\u003es\u003cstrong\u003e:\u003c/strong\u003e Retrospective data from 132 patients with intraspinal tumors (2020-2023) were analyzed. MRI scans were processed with 3D Slicer and Pyradiomics to extract features. Features were reduced using the LASSO algorithm and recursive elimination, followed by model construction using logistic regression, SVM, DT, RF, and XGBoost.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults:\u003c/strong\u003e 15 key features were identified post-screening. The XGBoost model showed the highest accuracy, with AUCs of 0.992 (training) and 0.542 (test). RF and DT models also performed well, with AUCs of 0.964/0.885 and 0.986/0.543 respectively.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion(s):\u003c/strong\u003e Integrating MRI radiomics with machine learning, particularly XGBoost, effectively differentiates intraspinal tumors, offering a non-invasive diagnostic tool that enhances early identification and treatment planning.\u003c/p\u003e","manuscriptTitle":"Classifying Intraspinal Tumors: A Multi-modal MRI and Machine Learning Approach ","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-07-19 20:17:49","doi":"10.21203/rs.3.rs-4624787/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"a6514f0c-02f1-4a1c-83f7-67a93909a470","owner":[],"postedDate":"July 19th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":34568132,"name":"Biological sciences/Cancer/Cancer imaging"},{"id":34568133,"name":"Biological sciences/Neuroscience/Neurogenesis"}],"tags":[],"updatedAt":"2025-09-25T10:24:07+00:00","versionOfRecord":[],"versionCreatedAt":"2024-07-19 20:17:49","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4624787","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4624787","identity":"rs-4624787","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}