Differentiating between PCNSL GCB Subtype and Non-GCB Subtype using Radiomics: A Multicenter Study

Differentiating between PCNSL GCB Subtype and Non-GCB Subtype using Radiomics: A Multicenter Study | 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 Research Article Differentiating between PCNSL GCB Subtype and Non-GCB Subtype using Radiomics: A Multicenter Study Yelong Shen, Siyu Wu, Yanan Wu, Chao Cui, Haiou Li, Shuang Yang, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4505854/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 10 You are reading this latest preprint version Abstract Objectives To investigate the characteristics and pathological basis of MRI in germinal center B cell (GCB) and non-germinal center B cell (non-GCB) in PCNSL (primary central nervous system lymphoma). And to explore the predictive ability of MRI radiomics-based in differentiating the GCB and non-GCB of PCNSL. Methods This study retrospectively analyzed standard diagnostic MRI examinations in 24 immunocompetent patients (9 men; age 56.4 ± 15.1 years) with GCB and 56 immunocompetent patients (35 men; age 61.1 ± 9.3 years) with non-GCB. The radiomics features were extracted from ADC, DWI, and T1-CE images respectively, and the features were screened by machine learning algorithm and statistical method. Finally, radiomics models of seven different sequence permutations were constructed. The area under the receiver operating characteristic (ROC AUC) curve was used to evaluate the predictive performance of all models. Delong test was utilized to compare the differences among models. Results The GCB cases all showed diffusion restriction, which was observed in 80.36% of the non-GBM cases; p < 0.05. Grade 3 edema was rare in GCB cases (8.33%) and common in non-GCB cases (50.00%); p < 0.001. 62.50% of male patients were non-GCB and 37.50% of female patients were non-GCB; p < 0.05. Additionally, patients with the GCB subtype are younger than those with the non-GCB subtype; p < 0.05. The best prediction model in our study used a combination of ADC, DWI, and T1-CE achieving the highest AUC of 0.854. And there was a significant difference between the best-combined model and some of the other models. Conclusion The GCB subtype is commonly seen in women, with mild peritumoral edema in most cases and diffusion restriction in all cases; however, the non-GCB subtype is commonly seen in men, with severe peritumoral edema in most cases. Additionally, the radiomics model developed by all sequences combined had good performance in discriminating between GCB and non-GCB. Machine learning. Primary central nervous system lymphoma. Magnetic resonance imaging. Pathological classification. Radiomics Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Primary central nervous system lymphoma (PCNSL) is a rare malignant non-hodgkin's lymphoma (NHL), only 1% of lymphoma cases are PCNSL [ 1 ]. The incidence of PCNSL is relatively low with high aggressiveness and poor prognosis [ 2 ]. Diffuse large B-cell lymphoma (DLBCL) is the most common pathological type [ 1 ]. DLBCL in PCNSL was divided into germinal center B-cell type (GCB) and non-germinal center B-cell type (non-GCB) according to the expression of immune markers such as CD10, B-cell lymphoma oncogene-6 (BCL-6) and multiple myeloma oncogene-1 (MUM-1) [ 3 ]. Progression-free survival (PFS) and overall survival (OS) were significantly worse in non-GCB patients than in GCB patients [ 3 ]. In non-GCB patients, lenalidomide and/or ibrutinib will be added to the R-CHOP regimen to improve efficacy [ 4 ]. Hence, it is important to differentiate the GCB and non-GCB subtypes both for prognosis prediction and patient medication instruction [ 4 , 5 ]. In clinical practice, gene expression profiling (GEP) and immunohistochemistry are often used to classify the GCB and non-GCB subtypes, in which biopsy was needed [ 6 ]. However, the therapy of GCB patients often uses radiotherapy and chemotherapy regimens like rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisolone (R-CHOP) [ 7 ]. Surgical operation is not the first choice method of treatment. It is best to avoid a biopsy. A noninvasive method that can predict GCB and non-GCB subtypes is needed. Radiomics is an emerging field in quantitative imaging that uses advanced imaging features to objectively and quantitatively describe tumor phenotypes [ 8 , 9 ]. Radiomics and machine learning can provide a basis for treatment decisions by modeling and analyzing radiomics features in medical images, which can also be used to predict the prognosis of diseases [ 8 ]. It is recommended to use apparent diffusion coefficient (ADC), diffusion-weighted imaging (DWI), and T1 contrast-enhanced (T1-CE) sequences to diagnose PCNSL [ 10 ]. Then multi-mode fault-guided biopsy or minimally invasive surgery is to achieve a pathological diagnosis. Previous studies have described the MRI signs of PCNSL, but the imaging features of GCB and non-GCB patients have not been summarized and analyzed [ 11 ]. In this study, the MRI characteristics will be summarized. ADC, DWI, and T1-CE are used to differentiate the GCB and non-GCB subtypes using radiomics. Materials and Methods Patients This study included 80 PCNSL patients with complete ADC, DWI, and T1-CE MRI image data, clinic-pathological features, and follow-up information. The imaging data of 145 patients were retrospectively collected from the picture archiving and communication systems (PACS) of Shandong Provincial Hospital Affiliated to Shandong First Medical University, Qilu Hospital of Shandong University, Shandong Cancer Hospital and Institute, the Affiliated Hospital of Qingdao University and Shandong Provincial Qianfoshan Hospital. Excluding lesions cannot accurately measure cases. Inclusion criteria include: 1) GCB or non-GCB was confirmed by immunohistochemical examination; 2) Complete MRI examination data with DWI (b = 1000), ADC, and T1-CE images; 3) No history of congenital or acquired immune deficiency; 4) No history of organ transplantation; 5) No other lymphoma; 6) No treatment was performed before MRI examination; 7) Patient information is complete. Finally, eighty patients (44 men; 5–80 years, mean age 59.5 years) met the inclusion criteria. Radiographic Data and Image Segmentation All images were acquired during routine clinical examination using a 3.0T MRI scanner with a standard Siemens 20-channel head coil. DWI was performed using a single-shot spin echo (SE) echo plane sequence with the following parameters: Echo time (TE)/repetition time (TR) = 74.6/5300 ms, 90° flip angle, 38 transverse sections, slice thickness = 5 mm, field of view (FOV) = 240 mm. A diffusion sensitization gradient is applied sequentially in the x, y and z directions with a b-factor of 1000 s/mm 2 .The ADC is automatically calculated and displayed as a corresponding ADC plot on the operator console of the MR scanner. Enhanced T1-weighted 3D gradient echo sequence (GRE) imaging was obtained using the following parameters: TE/TR = 15.87/1884 ms, flip angle 90°, 1-mm section thickness and 240 mm FOV. A standard dose (0.1 mmol/kg body weight) of gadoteric acid (Gd-DOTA, Dotarem; Laboratoire Guerbet, Aulnay-sous-Bois, France) was administered intravenously. Without knowing the pathology results, these images were analysed by two experienced radiologists. Signs analyzed include lesion sites, type of lesions, type of enhancement, degree of diffusion, peritumor edema, signs of bleeding and butterfly pattern. Types of lesions include solitary demarcated lesion, solitary infiltrative type, diffuse infiltrative type and multiple infiltrative lesion. Solitary or multiple infiltrative lesions were defined as follows (at least one criterion): no clear demarcation; non-enhancing parts of the tumor other than the enhancing parts; invasion of the meninges, ependyma or cranial nerves. Inclusion criteria for the diffuse infiltrative type were as follows: both white and grey matter of the brain were involved; tumor spread along the white matter tract and affected more than 2/3 of the unilateral cerebral hemisphere and/or both supratentorial and infratentorial regions were affected. Based on comparison with conventional T1-weighted fat-suppressed TSE sequence, the degree of enhancement was assessed by two diagnostic radiologists, the results include homogeneous enhancement, nonhomogeneous enhancement, non-enhancing areas indicating necrotic tissue and diffuse infiltrative. Grade the degree of peri-tumoral edema: grade 0: no peri-tumoral edema; grade 1: width of edema zone ≤ 2cm; grade 2: width of edema zone > 2cm and not exceeding 50% of the maximum transverse diameter of the cerebral hemisphere; grade 3: width of the edema zone exceeding 50% of the maximum transverse diameter of the cerebral hemisphere. In case of disagreement between the two physicians, they will discuss it with the senior physician to reach an agreement. Feature Extraction We resampled and normalized the ADC, DWI and T1-CE images before the radiomics features were extracted, and nine methods of imaging transformation were performed to protect the data from heterogeneity bias. Features were extracted from the ROIs of the original images. Radiomics features were extracted from the ROI of each of the three sequences separately via the Deepwise Multimodal Research Platform (version 2.0, https://keyan.deepwise.com ). Feature Selection Inter-observer and intra-observer repeatability were assessed to evaluate the stability of the extracted radiomics features. Fifteen cases were randomly selected to create two identical DWI image labeling tasks. ROIs were segmented independently by observer 1 and observer 2 (both with 10 years of neuro-MRI diagnostic experience). The radiomic features obtained were used for inter-observer ICC analysis. Observer 1 re-segmented the above 15 cases at monthly intervals, the first and second extracted radiomic features were used for intra-observer ICC analysis; finally, Observer 1 segmented the ROI of all the remaining cases. Feature selection is a key step in the process of model building. First, unstable radiomics features were removed based on the results of intra- and inter-observer ICC analysis. A feature correlation analysis with a threshold of 0.7 was used to screen out redundant features. F-test was conducted for further feature screening Model construction and evaluating Support vector machine is a linear classifier with a maximum interval defined on the feature space, and its algorithm is an optimization algorithm for solving convex quadratic programming. Previous studies suggest that SVM is effective in high-dimensional feature space, and its sparse feature can suppress data noise [ 12 ]. As a result, SVM was used in this study. The dataset of 80 samples was randomly divided in a ratio of 7:3. 56 samples were used as a training set to train the model and 24 samples were used as testing set to validate the performance of the model. Based on the features of single-sequence images (DWI, T1-CE, ADC) and multiple-sequence images (ADC + DWI, ADC + T1-CE, DWI + T1-CE, ADC + DWI + T1-CE), seven radiomics models were established using the SVM algorithm, respectively. Figure 1 describes the workflow of the current study. The performance of these models was evaluated by calculating AUC, accuracy, sensitivity and specificity. In order to analyze the clinical efficacy of the best machine learning model, decision curve was plotted in this study. Furthermore, Delong test was applied to calculate whether the models were statistically different. Statistical Analysis The chi-square test was performed to compare type of lesions, type of enhancement, degree of diffusion, peritumor edema, butterfly pattern, signs of bleeding and gender between patients with GCB subtype and those with non-GCB subtype. Independent samples t-test was utilized to compare the ages of patients across the two pathology types. Delong test was applied to calculate the statistical differences among the different models by medcalc (version 20.014, https://www.medcalc.org ). Statistical analysis and plots were conducted by the Deepwise Multimodal Research Platform, R statistical software application (version 4.2.2, http://www.Rproject.org ) and SPSS 25.0. P values < 0.05 was regarded as statistically significant. Results Patient Characteristics Among the 80 cases, 24 (9 men; age 56.4 ± 15.1 years) and 56 (35 men; age 61.1 ± 9.3 years) were in the GCB group and the non-GCB group, respectively. As shown in Table 1 , there were statistical differences between GCB patients and non-GCB patients in terms of both gender and age (both p < 0.05). Additionally, we found that GCB subtype mostly located in the frontal lobe, temporal lobe and lateral ventricle, non-GCB subtype mostly located in the frontal lobe, temporal lobe and corpus callosum, as shown in Table 2 . Moreover, diffusion restriction was observed in all GCB subtypes, however there were 19.64% of non-GCB subtypes that performed with no diffusion restriction (p < 0.05). There were also statistical differences in the tumor edema between the two histological subtypes (p < 0.05). In order to facilitate clinical application, this study defined grade 0 and 1 edema as mild edema and grade 2 and 3 edema as severe edema. We found that most GCB subtype was mild edema (62.5%) and most non-GCB subtype was severe edema (66.07%), with a statistically significant difference between them (p = 0.018). According to Table 1 and Table 3 , the main factors in decision making process between GCB and non-GCB are gender and age of the patients, presence or absence of diffusion restriction and the degree of edema. Figure 2 illustrates MRI images and pathology of GCB subtype lymphoma and non-GCB subtype lymphoma. Table 1 Characteristics of GCB and non-GCB patients. Patient characteristics GCB subtype non-GCB subtype p value Gender Male 37.50% 62.50% 0.039 Female 62.50% 37.50% 0.039 Age 0.029 Table 2 Lesion sites of GCB and non-GCB patients. Lesion sites GCB subtype non-GCB subtype Frontal lobe 9 16 Temporal lobe 12 20 Parietal lobe 2 3 Occipital lobe 1 9 Cerebellum 2 6 Corpus callosum 3 13 Basal ganglia 2 5 Lateral ventricle 5 8 Third ventricle 1 1 Table 3 MRI findings of GCB and non-GCB. MRI finding GCB non-GCB p -value Type of lesions solitary demarcated lesion 33.33% 33.93% 0.959 solitary infiltrative type 20.83% 7.14% 0.165 diffuse infiltrative type 0% 1.79% 1.000 multiple infiltrative lesion 45.83% 57.14% 0.353 Type of enhancement homogeneous 33.33% 30.36% 0.068 nonhomogeneous 45.83% 39.29% 0.586 presence of necrosis 20.83% 28.57% 0.471 diffuse infiltrative 0% 1.79% 1.000 DWI free diffusion 100.00% 80.36% 0.047 restricted diffusion in some part of the solid 0% 19.64% 0.047 Tumor edema grade 0 12.50% 5.36% 0.517 grade 1 50.00% 28.57% 0.066 grade 2 29.17% 16.07% 0.300 grade 3 8.33% 50.00% <0.001 mild 62.50% 33.93% 0.018 severe 37.50% 66.07% 0.018 Butterfly pattern 12.50% 21.43% 0.532 Signs of bleeding 20.83% 28.57% 0.471 Feature Selection and Establishment of Radiomics Models A total of 1906,1906 and 1743 radiomics features were extracted from each patient 's ADC, DWI and T1-CE images, respectively. After inter-observer and intra-observer ICC analysis, ADC, DWI and T1-CE retained 1497,1447 and 1530 features, respectively. Finally, after feature reduction, 9 features of ADC images, 10 features of DWI images and 9 features of T1-CE features were retained. Models performance The models that established by features from individual MRI sequence achieved an AUC from 0.753 to 0.976 on training set and from 0.641 to 0.746 on testing set. Based on features from a combination of sequences, the models achieved an AUC from 0.830 to 0.879 on training set and from 0.725 to 0.831 on testing set. And based on features combined all different sequences, the model had the best performance with an AUC of 0.902, accuracy of 0.813, specificity of 0.871, sensitivity of 0.797 on training set and an AUC of 0.854, accuracy of 0.790, specificity of 0.539, sensitivity of 0.857 on testing set. The results were shown in supplementary material. Furthermore, there were statistical significance of differences between ADC + DWI + T1-CE model and the ADC model (p = 0.005), DWI model (p = 0.018) and T1-CE model (p = 0.003) on the training set, respectively. And there were statistical differences between ADC + DWI + T1-CE model and the ADC model (p = 0.048) and DWI model (p = 0.043) on testing set, respectively, as shown in Table 4 . Table 4 Delong test results of radiomics models based on different MRI sequences. Results with p value < 0.05 were marked coarse. p value of the training set p value of the testing set ADC VS. DWI 0.839 0.744 ADC VS. T1-CE <0.001 0.502 DWI VS. T1-CE <0.001 0.406 ADC VS. ADC + DWI <0.001 0.410 ADC VS. ADC + T1-CE <0.001 0.128 ADC VS. DWI + T1-CE 0.001 0.665 DWI VS. ADC + DWI <0.001 0.359 DWI VS. ADC + T1-CE 0.001 0.082 DWI VS. DWI + T1-CE 0.002 0.500 T1-CE VS. ADC + T1-CE 0.029 0.301 T1-CE VS. DWI + T1-CE 0.022 0.901 T1-CE VS. ADC + DWI 0.148 0.867 all VS. T1-CE 0.003 0.104 all VS. ADC 0.005 0.048 all VS. DWI 0.018 0.043 all VS. ADC + T1-CE 0.152 0.738 all VS. DWI + T1-CE 0.525 0.163 all VS. ADC + DWI 0.026 0.120 ADC + T1-CE VS. DWI + T1-CE 0.683 0.382 ADC + T1-CE VS. ADC + DWI 0.391 0.475 ADC + DWI VS. DWI + T1-CE 0.252 0.715 Figure 3 shows the ROC curves of all the radiomics models and their comparison. The results of ROC curves showed that the model of all sequences combined had the highest AUC on both training and testing set, which suggested it had the best prediction performance. To analyze the clinical value of the best model in this study, we conducted the clinical decision curve analysis (Fig. 4 ). We found that using the radiomics model in this study has more benefits than treating all or none of the patients. Discussion With the advance in transplantation techniques and the increasing number of patients treated with immunosuppressive therapy, the incidence of PCNSL has been on the rise in recent years [ 1 ]. The most common histological subtype in PCNSL is DLBCL [ 13 ]. With the advent of DNA microarray technology, DLBCL has been classified into two different subtypes based on different gene expression patterns: GCB and non-GCB [ 14 ]. A combined HD-MTX-based regimen was recommended in the Chinese expert consensus published in 2022 [ 15 ]. Nevertheless, studies have shown that patients with non-GCB subtype respond less well to upfront R-CHOP chemotherapy compared to those with GCB subtype, which leads to a relatively poor prognosis for non-GCB patients [ 16 ]. Therefore, lenalidomide and/or ibrutinib are often added to existing R-CHOP chemotherapy regimens in order to improve efficacy in the treatment of non-GCB patients [ 10 , 17 , 18 ]. Early identification of the patient's pathological type is important to improve the patient's prognosis and achieve optimal neurological recovery and disease control [ 1 ]. Therefore, this study analyzes the imaging characteristics of GCB and non-GCB patients and constructs a radiomics model for tumor differentiation by analyzing medical images, which can help radiologists to differentiate GCB lymphoma from non-GCB lymphoma to a certain extent and can support clinicians in treatment decisions. Past study has shown that T1-CE and DWI can be non-invasive and accurate in assessing the disease status of patients [ 19 ]. Therefore, in this study, TI-CE images, DWI and ADC images of PCNSL patients were analyzed. We found that between the 2 groups there were statistically significant differences (p < 0.05) in terms of the degree of diffusion restriction, degree of peritumoral edema, and gender. Due to the high nucleoplasmic ratio of tumor cells in PCNSL, the diffuse movement of water molecules in the tumor cells is restricted and appears as a distinct high signal on DWI images [ 20 ]. Yet, in the current study, 18.64% of non-GCB cases had unrestricted diffusion, and the difference in the degree of diffusion between GCB and non-GCB lymphomas was statistically significant. This may be caused by the different cellular origins of non-GCB lymphomas. Non-GCB lymphomas can be derived from plasma cells, plasmacytes, immunoblasts, marginal zone cells and memory cells. Non-GCB lymphomas can be derived from plasma cells, plasmablast, immunoblasts, marginal zone B cells and memory cells. Both lymphoplasmacytic lymphoma and nodal marginal zone B-cell lymphoma have abundant cytoplasm and low nucleoplasmic ratio. This may lead to the appearance of isosignal on DWI images for some non-GCB lymphomas. The results showed a statistically significant difference (p < 0.05) in the degree of edema between GCB subtype and non-GCB subtype. This may be since PCNSL is highly invasive and often destroys vascular endothelial cells, which leads to disruption of the blood-brain barrier [ 10 ]. When the blood-brain barrier is disrupted, water, electrolytes and plasma proteins from the microvasculature surrounding the tumor leak out of the blood vessels, causing vasogenic edema. Phosphorylated STAT3 is the "trigger point" of the JAK-STAT pathway, which is associated with cell proliferation, and phosphorylated STAT3 activates the JAK-STAT pathway to cause cell proliferation. Yamada et al. found that phosphorylated STAT3 was expressed at a higher rate in the non-GCB subtype than in the GCB subtype [ 21 ]. This suggests that more rapid cell proliferation accelerates the disruption of the blood-brain barrier, resulting in significantly higher levels of peritumoral edema in non-GCB subtype than that in GCB subtype. This phenomenon has not been reported in the literature yet. In addition, after analyzing the gender of the patients, we found that the majority of male patients presented with non-GCB subtype and the majority of female patients presented with GCB subtype. The difference was statistically significant. However, the reasons for this difference need to be further explored. Intratumoral hemorrhage is usually considered to be rare in PCNSLs, however, Yamada et al [ 21 ] found that all six of the non-GCB patients they reported had significant intratumoral hemorrhage and we also found intratumoral hemorrhage in non-GCB subtype. However, we observed no statistically significant difference in signs of bleeding between GCB subtype and non-GCB subtype. What's more, this study found that GCB subtype was mainly located in the temporal lobe and frontal lobe as well as the lateral ventricle, while the majority of non-GCB subtype occurred in the temporal lobe, frontal lobe and corpus callosum. Overall, the majority of both GCB subtype and non-GCB subtype occurred near the midline, which is consistent with the past study [ 22 ]. With the arise and development of radiomics, the information can be extracted, filtered and analyzed by machine learning algorithms to build radiomics models for differential diagnosis [ 23 ], predicting prognosis [ 24 ] and predicting treatment response [ 25 ]. In this study, based on the results of the analysis of medical images described above, we have developed a radiomics model that can differentiate GCB subtype from non-GCB subtype. To ensure the robustness of the model, four feature screening steps were taken to obtain the features needed to develop the model for this study. Firstly, we standardized the data. Then, radiomics features with high stability (ICC ≥ 0.8) were retained for the next step of feature screening based on the results of intra- and inter-observer ICC analysis. Thirdly, A feature correlation analysis with a threshold of 0.7 was applied to screen redundant features. Finally, an F-test was taken to obtain the features needed for the radiomics models. As shown in Fig. 3 , models constructed by all sequences combined, sequence combination (ADC + T1-CE) and radiomics features in T1-CE images had a good diagnostic performance. In the validation set, the model developed by all sequences combined had the best diagnostic performance with an AUC of 0.854 (95% CI 0.756–0.953). This may result from the fact that the radiomics features extracted from all sequences combined can reflect the biological information of the tumor more comprehensively and accurately. MRI examination and radiomics analysis were two important methods of differentiating GCB subtype from non-GCB subtype. T1-CE was the key sequence for both methods. In past studies, T1-CE images of PCNSLs were also collected when analyzing the lesion sites of the distribution of GCB subtype and non-GCB subtype [ 20 ]. Blasel et al. [ 26 ] suggested that the increased rCBV values in PCNSLs were mainly due to immunoreactive changes of the normal brain vasculature against the infiltrating lymphoma cells. These tumor-vascular interactions would lead to microvascular proliferation. In addition, Hartert et al. [ 19 ] showed that phosphorylated STAT3 was expressed at a higher level in non-GCB subtype than in GCB subtype. It is implied that non-GCB lymphoma cells proliferate more rapidly and have a more pronounced disruption to the blood-brain barrier, which results in more intense immunoreactive changes. In addition, the two subtypes have different cellular origins and different nucleocytoplasmic ratio [ 27 ], and therefore different degrees of diffusion on DWI and ADC images. Hence, the differential diagnostic performance of the model constructed by all sequences combined was higher than that of the models constructed with sequence combination (ADC + T1-CE) and T1-CE. Machine learning algorithms can objectively analyze high-throughput medical image information to develop radiomics models that can differentiate GCB subtype from non-GCB subtype. Different radiologists have different diagnostic experiences, there is subjectivity in the diagnosis of the disease and it is difficult to capture some of the imaging features with the naked-eye. As a result, there may be inter-individual variation in the diagnostic accuracy of radiologists, and junior radiologists may have a slightly lower diagnostic performance than that of the radiomics models [ 28 ]. In addition, compared to stereotactic biopsy, radiomics can be more widely available, more reflective of the overall tumor level and have few side effects. Additionally, ADC, DWI and T1-CE were routine MRI examination sequences for the preoperative diagnosis in patients with PCNSLs. Therefore, the radiomics model developed in this study can help clinicians determine the pathologic classification of PCNSLs with contraindications to biopsy, provide more reliable decision support for the determination of chemotherapy regimens for patients, and is more generalizable in clinical work. These advantages were supported in the decision analysis curve (Fig. 4 ). There were some limitations in our study. Firstly, the number of patients was insufficient and the pathological types of cases were limited to GCB and non-GCB subtypes. Secondly, there was no external validation set for this study, which should be included in future studies. Finally, the edematous, enhanced and necrotic areas of the tumor were not outlined independently. With the further development of automated brain tumor segmentation techniques, biological information from different regions of the tumor will need to be explored in future studies. Conclusion Conventional MRI can distinguish to a certain extent between GCB subtypes and non-GCB subtypes. The GCB subtypes are common in females, most cases present mild peritumoral edema, and diffusion restriction is detected in all of them; however, the non-GCB subtypes are more common in males and severe peritumoral edema can be observed in most cases. Additionally, the radiomics model of all sequences combined has the potential in distinguishing between GCB and non-GCB. Abbreviations ADC Apparent diffusion coefficient AUC Area under the curve BCL-6 B-cell lymphoma oncogene-6 DCA Decision curve analysis DWI Diffusion-weighted imaging FOV Field of view GCB Germinal center B cell GEP Gene expression profiling GRE Gradient echo sequence MUM-1 Multiple myeloma oncogene-1 NHL Non-hodgkin's lymphoma OS Overall survival PACS Picture archiving and communication systems PCNSL Primary central nervous system lymphoma PFS Progression-free survival R-CHOP Rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisolone SE Spin echo TE Echo time TR Repetition time T1-CE T1 contrast-enhanced Declarations Acknowledgment The authors would like to thank all the participants. Authors ’ contributions Yelong Shen: Data curation, Formal analysis, Methodology, Project administration, Writing – original draft, Writing – review & editing. Siyu Wu: Formal analysis, Methodology, Visualization, Writing – original draft, Writing – review & editing. Yanan Wu: Data curation. Chao Cui: Data curation. Haiou Li: Data curation. Shuang Yang: Data curation. Xuejun Liu: Data curation. Xingzhi Chen: Project. Chencui Huang: Project. Ximing Wang: Conceptualization, Data curation, Project administration, Supervision, Writing – review & editing. Funding The present study was supported by the National Natural Science Foundation of China (Grant Nos. 8187354, 81571672), and Academic promotion program of Shandong First Medical University (Grant No. 2019QL023). Availability of data and materials The data that support the fundings of this study are available from the corresponding author upon reasonable request. Ethics approval and consent to participate This retrospective study was approved by the Ethics Committee of the Shandong Provincial Hospital and written informed consent for the participants over 16 years of age was discarded because of the retrospective nature of this study. For participants less than 16 years of age, written informed consent has been obtained from their parents/guardians after a brief description of the purpose and objectives of the study has been given to them. We confirmed that the study was carried out in accordance with relevant guidelines and regulations of Helsinki Declaration. Consent for publication Not applicable. Competing interests All the authors declare that they have no conflict of interest. References Schaff LR, Grommes C. Primary central nervous system lymphoma. Blood. 2022;140(9):971–9. .https://www.ncbi.nlm.nih.gov/pubmed/34699590 . Holdhoff M, Mrugala MM, Grommes C, Kaley TJ, Swinnen LJ, Perez-Heydrich C, Nayak L. Challenges in the Treatment of Newly Diagnosed and Recurrent Primary Central Nervous System Lymphoma. J Natl Compr Canc Netw. 2020;18(11):1571–8. https://www.ncbi.nlm.nih.gov/pubmed/33152700 . Marcus C, Maragkos GA, Alterman RL, Uhlmann E, Pihan G, Varma H. GCB-type is a favorable prognostic factor in primary CNS diffuse large B-cell lymphomas. J Clin Neurosci. 2021;83:49–55. https://doi.org/10.1016/j.jocn.2020.11.031 . Goy A, Ramchandren R, Ghosh N, Munoz J, Morgan DS, Dang NH, Knapp M, Delioukina M, Kingsley E, Ping J, et al. Ibrutinib plus lenalidomide and rituximab has promising activity in relapsed/refractory non-germinal center B-cell-like DLBCL. Blood. 2019;134(13):1024–36. Hou K, Yu Z, Jia Y, Fang H, Shao S, Huang L, Feng Y. Efficacy and safety of ibrutinib in diffuse large B-cell lymphoma: A single-arm meta-analysis. Crit Rev Oncol Hematol 2020, 152103010 https://www.ncbi.nlm.nih.gov/pubmed/32540781 . Schmitz R, Wright GW, Huang DW, Johnson CA, Phelan JD, Wang JQ, Roulland S, Kasbekar M, Young RM, Shaffer AL, et al. Genetics and Pathogenesis of Diffuse Large B-Cell Lymphoma. N Engl J Med. 2018;378(15):1396–407. Nowakowski GS, Chiappella A, Gascoyne RD, Scott DW, Zhang Q, Jurczak W, Özcan M, Hong X, Zhu J, Jin J, et al. ROBUST: A Phase III Study of Lenalidomide Plus R-CHOP Versus Placebo Plus R-CHOP in Previously Untreated Patients With ABC-Type Diffuse Large B-Cell Lymphoma. J Clin Oncol. 2021;39(12):1317–28. Lohmann P, Franceschi E, Vollmuth P, Dhermain F, Weller M, Preusser M, Smits M, Galldiks N. Radiomics in neuro-oncological clinical trials. Lancet Digit Health. 2022;4(11):e841–9. https://www.ncbi.nlm.nih.gov/pubmed/36182633 . Mayerhoefer ME, Materka A, Langs G, Häggström I, Szczypiński P, Gibbs P, Cook G. Introduction to Radiomics. J Nucl Med. 2020;61(4):488–95. Barajas RF, Politi LS, Anzalone N, Schoder H, Fox CP, Boxerman JL, Kaufmann TJ, Quarles CC, Ellingson BM, Auer D, et al. Consensus recommendations for MRI and PET imaging of primary central nervous system lymphoma: guideline statement from the International Primary CNS Lymphoma Collaborative Group (IPCG). Neuro Oncol. 2021;23(7):1056–71. https://www.ncbi.nlm.nih.gov/pubmed/33560416 . Kuker W, Nagele T, Korfel A, Heckl S, Thiel E, Bamberg M, Weller M, Herrlinger U. Primary central nervous system lymphomas (PCNSL): MRI features at presentation in 100 patients. J Neurooncol. 2005;72(2):169–77. https://www.ncbi.nlm.nih.gov/pubmed/15925998 . Jin X, Zheng X, Chen D, Jin J, Zhu G, Deng X, Han C, Gong C, Zhou Y, Liu C, et al. Prediction of response after chemoradiation for esophageal cancer using a combination of dosimetry and CT radiomics. Eur Radiol. 2019;29(11):6080–8. https://www.ncbi.nlm.nih.gov/pubmed/31028447 . Yuan Y, Ding T, Wang S, Chen H, Mao Y, Chen T. Current and emerging therapies for primary central nervous system lymphoma. Biomark Res. 2021;9(1):32. https://www.ncbi.nlm.nih.gov/pubmed/33957995 . Yang JM, Jang JY, Jeon YK, Paik JH. Clinicopathologic implication of microRNA-197 in diffuse large B cell lymphoma. J Transl Med. 2018;16(1):162. https://www.ncbi.nlm.nih.gov/pubmed/29890998 . Chen T, Liu Y, Wang Y, Chang Q, Wu J, Wang Z, Geng D, Yu JT, Li Y, Li XQ, et al. Evidence-based expert consensus on the management of primary central nervous system lymphoma in China. J Hematol Oncol. 2022;15(1):136. https://www.ncbi.nlm.nih.gov/pubmed/36176002 . Schultz CJ, Bovi J. Current management of primary central nervous system lymphoma. Int J Radiat Oncol Biol Phys. 2010;76(3):666–78. https://www.ncbi.nlm.nih.gov/pubmed/20159361 . Nowakowski GS, Hong F, Scott DW, Macon WR, King RL, Habermann TM, Wagner-Johnston N, Casulo C, Wade JL, Nagargoje GG, et al. Addition of Lenalidomide to R-CHOP Improves Outcomes in Newly Diagnosed Diffuse Large B-Cell Lymphoma in a Randomized Phase II US Intergroup Study ECOG-ACRIN E1412. J Clin Oncol. 2021;39(12):1329–38. https://www.ncbi.nlm.nih.gov/pubmed/33555941 . Wilson WH, Wright GW, Huang DW, Hodkinson B, Balasubramanian S, Fan Y, Vermeulen J, Shreeve M, Staudt LM. Effect of ibrutinib with R-CHOP chemotherapy in genetic subtypes of DLBCL. Cancer Cell. 2021;39(12):1643–e16531643. https://www.ncbi.nlm.nih.gov/pubmed/34739844 . Hartert KT, Wenzl K, Krull JE, Manske M, Sarangi V, Asmann Y, Larson MC, Maurer MJ, Slager S, Macon WR et al. Targeting of inflammatory pathways with R2CHOP in high-risk DLBCL. Leukemia 2021, 35(2):522–33 https://www.ncbi.nlm.nih.gov/pubmed/32139889 . Kinoshita M, Sasayama T, Narita Y, Yamashita F, Kawaguchi A, Chiba Y, Kagawa N, Tanaka K, Kohmura E, Arita H, et al. Different spatial distribution between germinal center B and non-germinal center B primary central nervous system lymphoma revealed by magnetic resonance group analysis. Neuro Oncol. 2014;16(5):728–34. https://www.ncbi.nlm.nih.gov/pubmed/24497406 . Yamada S, Muto J, Iba S, Shiogama K, Tsuyuki Y, Satou A, Ohba S, Murayama K, Sugita Y, Nakamura S, et al. Primary central nervous system lymphomas with massive intratumoral hemorrhage: Clinical, radiological, pathological, and molecular features of six cases. Neuropathology. 2021;41(5):335–48. https://www.ncbi.nlm.nih.gov/pubmed/34254378 . Liu D, Kong Z, Wang Y, Chen W, Wang Y, Chen W, Liu L, Dang Y, Ma W, Wang Y, et al. Quantitative and Visual Characteristics of Primary Central Nervous System Lymphoma on (18)F-FDG-PET. Interdiscip Sci. 2019;11(2):300–6. https://www.ncbi.nlm.nih.gov/pubmed/31264053 . Bathla G, Priya S, Liu Y, Ward C, Le NH, Soni N, Maheshwarappa RP, Monga V, Zhang H, Sonka M. Radiomics-based differentiation between glioblastoma and primary central nervous system lymphoma: a comparison of diagnostic performance across different MRI sequences and machine learning techniques. Eur Radiol. 2021;31(11):8703–13. https://www.ncbi.nlm.nih.gov/pubmed/33890149 . Zhong S, Ren JX, Yu ZP, Peng YD, Yu CW, Deng D, Xie Y, He ZQ, Duan H, Wu B et al. Predicting glioblastoma molecular subtypes and prognosis with a multimodal model integrating convolutional neural network, radiomics, and semantics. J Neurosurg 2022:1–10 https://www.ncbi.nlm.nih.gov/pubmed/36461822 . Abdel Razek AAK, Alksas A, Shehata M, AbdelKhalek A, Abdel Baky K, El-Baz A, Helmy E. Clinical applications of artificial intelligence and radiomics in neuro-oncology imaging. Insights Imaging. 2021;12(1):152. https://www.ncbi.nlm.nih.gov/pubmed/34676470 . Blasel S, Vorwerk R, Kiyose M, Mittelbronn M, Brunnberg U, Ackermann H, Voss M, Harter PN, Hattingen E. New MR perfusion features in primary central nervous system lymphomas: pattern and prognostic impact. J Neurol. 2018;265(3):647–58. https://www.ncbi.nlm.nih.gov/pubmed/29383512 . Li S, Young KH, Medeiros LJ. Diffuse large B-cell lymphoma. Pathology. 2018;50(1):74–87. https://www.ncbi.nlm.nih.gov/pubmed/29167021 . Xia W, Hu B, Li H, Geng C, Wu Q, Yang L, Yin B, Gao X, Li Y, Geng D. Multiparametric-MRI-Based Radiomics Model for Differentiating Primary Central Nervous System Lymphoma From Glioblastoma: Development and Cross-Vendor Validation. J Magn Reson Imaging. 2021;53(1):242–50. https://www.ncbi.nlm.nih.gov/pubmed/32864825 . Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterial.docx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 24 Jun, 2025 Reviews received at journal 14 May, 2025 Reviewers agreed at journal 07 May, 2025 Reviews received at journal 08 Jul, 2024 Reviewers agreed at journal 07 Jul, 2024 Reviewers invited by journal 02 Jul, 2024 Editor invited by journal 06 Jun, 2024 Editor assigned by journal 06 Jun, 2024 Submission checks completed at journal 06 Jun, 2024 First submitted to journal 30 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-4505854","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":312115712,"identity":"82971a4d-8fdf-4886-8cc6-9889fd2ca351","order_by":0,"name":"Yelong Shen","email":"","orcid":"","institution":"Shandong Provincial Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yelong","middleName":"","lastName":"Shen","suffix":""},{"id":312115714,"identity":"33164065-7b73-4367-82ff-aa9e91a23f0d","order_by":1,"name":"Siyu Wu","email":"","orcid":"","institution":"Shandong Provincial Hospital","correspondingAuthor":false,"prefix":"","firstName":"Siyu","middleName":"","lastName":"Wu","suffix":""},{"id":312115715,"identity":"9447a3a9-2361-4bcd-8409-219ef996e362","order_by":2,"name":"Yanan Wu","email":"","orcid":"","institution":"Shandong Provincial Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yanan","middleName":"","lastName":"Wu","suffix":""},{"id":312115717,"identity":"9f75b192-1b1b-41d8-8fab-5155ac5aa084","order_by":3,"name":"Chao Cui","email":"","orcid":"","institution":"The Affiliated Taian City Central Hospital of Qingdao University","correspondingAuthor":false,"prefix":"","firstName":"Chao","middleName":"","lastName":"Cui","suffix":""},{"id":312115718,"identity":"c0d88148-93e1-4190-acc1-51010b60c98d","order_by":4,"name":"Haiou Li","email":"","orcid":"","institution":"Qilu Hospital of Shandong University","correspondingAuthor":false,"prefix":"","firstName":"Haiou","middleName":"","lastName":"Li","suffix":""},{"id":312115719,"identity":"f55fbe22-e83f-40f6-9437-0b6c3585d834","order_by":5,"name":"Shuang Yang","email":"","orcid":"","institution":"Shandong Provincial QianFoShan Hospital","correspondingAuthor":false,"prefix":"","firstName":"Shuang","middleName":"","lastName":"Yang","suffix":""},{"id":312115720,"identity":"17ec4cbe-1a37-47aa-aa1a-c2b78a2f528b","order_by":6,"name":"Xuejun Liu","email":"","orcid":"","institution":"Affiliated Hospital of Qingdao University","correspondingAuthor":false,"prefix":"","firstName":"Xuejun","middleName":"","lastName":"Liu","suffix":""},{"id":312115721,"identity":"a6b755e6-5ec4-4dbd-86fe-ed8edf9bced9","order_by":7,"name":"Xingzhi Chen","email":"","orcid":"","institution":"Department of Research Collaboration, R\u0026D center, Beijing Deepwise \u0026 League of PHD Technology Co.","correspondingAuthor":false,"prefix":"","firstName":"Xingzhi","middleName":"","lastName":"Chen","suffix":""},{"id":312115722,"identity":"4c9e5ad3-ee79-4c14-b1a6-ee08810bb56f","order_by":8,"name":"Chencui Huang","email":"","orcid":"","institution":"Department of Research Collaboration, R\u0026D center, Beijing Deepwise \u0026 League of PHD Technology Co.","correspondingAuthor":false,"prefix":"","firstName":"Chencui","middleName":"","lastName":"Huang","suffix":""},{"id":312115723,"identity":"8ac11da8-5c54-4c2b-9ae7-9266bbf6eeb1","order_by":9,"name":"Ximing Wang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAyUlEQVRIiWNgGAWjYJCCAwwMNXL87I2NDz+QoOWYsWTP4WZjCRIsYk40uJHeJsBDjFqDG9mJhwt+sSUY3HzYxiDBYCen20BIy5mzGw7P7JPJk7yd2PaggCHZ2OwAAS1mx3s3HObtYSvmu53YbiDBcCBxG0Eth3lBWpgTG24ebJPgIUoLyBaeH8yJE24wEqnFHuQX3gZQICcCA9mACL9Izsjd/JnnDygqjz98+KHCTo6gFjBgbIOxDIhRDgZ/iFY5CkbBKBgFIxEAABXLSrbAVopgAAAAAElFTkSuQmCC","orcid":"","institution":"Shandong Provincial Hospital","correspondingAuthor":true,"prefix":"","firstName":"Ximing","middleName":"","lastName":"Wang","suffix":""}],"badges":[],"createdAt":"2024-05-31 03:08:14","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4505854/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4505854/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":59122451,"identity":"4b973cce-51d6-4d5a-9f96-86f9acdd85f1","added_by":"auto","created_at":"2024-06-26 15:05:03","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":276869,"visible":true,"origin":"","legend":"\u003cp\u003eOverview of the current study workflow\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-4505854/v1/67d865270bcec550e9a5344b.png"},{"id":59122455,"identity":"c6b02658-ac42-4ecf-85f4-004598fd381f","added_by":"auto","created_at":"2024-06-26 15:05:03","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":475407,"visible":true,"origin":"","legend":"\u003cp\u003eThe axial T1-CE, DWI, ADC and Hematoxylin-Eosin stain pathological images of GCB subtype lymphoma (A-D) and non-GCB subtype lymphoma (E-H)\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-4505854/v1/6e38572b01d54c4400a1f0a2.png"},{"id":59122456,"identity":"7ea84e81-1768-44c3-959b-34d5ef50efd5","added_by":"auto","created_at":"2024-06-26 15:05:03","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":168686,"visible":true,"origin":"","legend":"\u003cp\u003eReceiver Operator Characteristic (ROC) curve of all the model in training set (A) and validation set(B)\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-4505854/v1/4f5def8ed008320af925cdb3.png"},{"id":59122453,"identity":"3fb68ca7-9721-4471-a57d-54c2b9da1d8d","added_by":"auto","created_at":"2024-06-26 15:05:03","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":95978,"visible":true,"origin":"","legend":"\u003cp\u003eDecision curve of model built by all sequences combined\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-4505854/v1/90c4e5359ae872f22c8fe19b.png"},{"id":59123288,"identity":"e765fa86-c7d0-41bf-8383-b9e64644a301","added_by":"auto","created_at":"2024-06-26 15:13:03","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1857204,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4505854/v1/70574976-29a7-4dd0-9cbc-bb7c11f7f1a6.pdf"},{"id":59122454,"identity":"cddf5272-e74c-4bc3-ae0d-39a850340f24","added_by":"auto","created_at":"2024-06-26 15:05:03","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":12530,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-4505854/v1/683851bdbe39b05cf295db31.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Differentiating between PCNSL GCB Subtype and Non-GCB Subtype using Radiomics: A Multicenter Study","fulltext":[{"header":"Introduction","content":"\u003cp\u003ePrimary central nervous system lymphoma (PCNSL) is a rare malignant non-hodgkin's lymphoma (NHL), only 1% of lymphoma cases are PCNSL [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. The incidence of PCNSL is relatively low with high aggressiveness and poor prognosis [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Diffuse large B-cell lymphoma (DLBCL) is the most common pathological type [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. DLBCL in PCNSL was divided into germinal center B-cell type (GCB) and non-germinal center B-cell type (non-GCB) according to the expression of immune markers such as CD10, B-cell lymphoma oncogene-6 (BCL-6) and multiple myeloma oncogene-1 (MUM-1) [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Progression-free survival (PFS) and overall survival (OS) were significantly worse in non-GCB patients than in GCB patients [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. In non-GCB patients, lenalidomide and/or ibrutinib will be added to the R-CHOP regimen to improve efficacy [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Hence, it is important to differentiate the GCB and non-GCB subtypes both for prognosis prediction and patient medication instruction [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn clinical practice, gene expression profiling (GEP) and immunohistochemistry are often used to classify the GCB and non-GCB subtypes, in which biopsy was needed [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. However, the therapy of GCB patients often uses radiotherapy and chemotherapy regimens like rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisolone (R-CHOP) [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Surgical operation is not the first choice method of treatment. It is best to avoid a biopsy. A noninvasive method that can predict GCB and non-GCB subtypes is needed.\u003c/p\u003e \u003cp\u003eRadiomics is an emerging field in quantitative imaging that uses advanced imaging features to objectively and quantitatively describe tumor phenotypes [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Radiomics and machine learning can provide a basis for treatment decisions by modeling and analyzing radiomics features in medical images, which can also be used to predict the prognosis of diseases [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIt is recommended to use apparent diffusion coefficient (ADC), diffusion-weighted imaging (DWI), and T1 contrast-enhanced (T1-CE) sequences to diagnose PCNSL [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Then multi-mode fault-guided biopsy or minimally invasive surgery is to achieve a pathological diagnosis. Previous studies have described the MRI signs of PCNSL, but the imaging features of GCB and non-GCB patients have not been summarized and analyzed [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. In this study, the MRI characteristics will be summarized. ADC, DWI, and T1-CE are used to differentiate the GCB and non-GCB subtypes using radiomics.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003ePatients\u003c/h2\u003e \u003cp\u003eThis study included 80 PCNSL patients with complete ADC, DWI, and T1-CE MRI image data, clinic-pathological features, and follow-up information. The imaging data of 145 patients were retrospectively collected from the picture archiving and communication systems (PACS) of Shandong Provincial Hospital Affiliated to Shandong First Medical University, Qilu Hospital of Shandong University, Shandong Cancer Hospital and Institute, the Affiliated Hospital of Qingdao University and Shandong Provincial Qianfoshan Hospital. Excluding lesions cannot accurately measure cases. Inclusion criteria include: 1) GCB or non-GCB was confirmed by immunohistochemical examination; 2) Complete MRI examination data with DWI (b\u0026thinsp;=\u0026thinsp;1000), ADC, and T1-CE images; 3) No history of congenital or acquired immune deficiency; 4) No history of organ transplantation; 5) No other lymphoma; 6) No treatment was performed before MRI examination; 7) Patient information is complete. Finally, eighty patients (44 men; 5\u0026ndash;80 years, mean age 59.5 years) met the inclusion criteria.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eRadiographic Data and Image Segmentation\u003c/h2\u003e \u003cp\u003eAll images were acquired during routine clinical examination using a 3.0T MRI scanner with a standard Siemens 20-channel head coil. DWI was performed using a single-shot spin echo (SE) echo plane sequence with the following parameters: Echo time (TE)/repetition time (TR)\u0026thinsp;=\u0026thinsp;74.6/5300 ms, 90\u0026deg; flip angle, 38 transverse sections, slice thickness\u0026thinsp;=\u0026thinsp;5 mm, field of view (FOV)\u0026thinsp;=\u0026thinsp;240 mm. A diffusion sensitization gradient is applied sequentially in the x, y and z directions with a b-factor of 1000 s/mm\u003csup\u003e2\u003c/sup\u003e.The ADC is automatically calculated and displayed as a corresponding ADC plot on the operator console of the MR scanner. Enhanced T1-weighted 3D gradient echo sequence (GRE) imaging was obtained using the following parameters: TE/TR\u0026thinsp;=\u0026thinsp;15.87/1884 ms, flip angle 90\u0026deg;, 1-mm section thickness and 240 mm FOV. A standard dose (0.1 mmol/kg body weight) of gadoteric acid (Gd-DOTA, Dotarem; Laboratoire Guerbet, Aulnay-sous-Bois, France) was administered intravenously. Without knowing the pathology results, these images were analysed by two experienced radiologists.\u003c/p\u003e \u003cp\u003eSigns analyzed include lesion sites, type of lesions, type of enhancement, degree of diffusion, peritumor edema, signs of bleeding and butterfly pattern. Types of lesions include solitary demarcated lesion, solitary infiltrative type, diffuse infiltrative type and multiple infiltrative lesion. Solitary or multiple infiltrative lesions were defined as follows (at least one criterion): no clear demarcation; non-enhancing parts of the tumor other than the enhancing parts; invasion of the meninges, ependyma or cranial nerves. Inclusion criteria for the diffuse infiltrative type were as follows: both white and grey matter of the brain were involved; tumor spread along the white matter tract and affected more than 2/3 of the unilateral cerebral hemisphere and/or both supratentorial and infratentorial regions were affected. Based on comparison with conventional T1-weighted fat-suppressed TSE sequence, the degree of enhancement was assessed by two diagnostic radiologists, the results include homogeneous enhancement, nonhomogeneous enhancement, non-enhancing areas indicating necrotic tissue and diffuse infiltrative. Grade the degree of peri-tumoral edema: grade 0: no peri-tumoral edema; grade 1: width of edema zone\u0026thinsp;\u0026le;\u0026thinsp;2cm; grade 2: width of edema zone\u0026thinsp;\u0026gt;\u0026thinsp;2cm and not exceeding 50% of the maximum transverse diameter of the cerebral hemisphere; grade 3: width of the edema zone exceeding 50% of the maximum transverse diameter of the cerebral hemisphere. In case of disagreement between the two physicians, they will discuss it with the senior physician to reach an agreement.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eFeature Extraction\u003c/h2\u003e \u003cp\u003eWe resampled and normalized the ADC, DWI and T1-CE images before the radiomics features were extracted, and nine methods of imaging transformation were performed to protect the data from heterogeneity bias. Features were extracted from the ROIs of the original images. Radiomics features were extracted from the ROI of each of the three sequences separately via the Deepwise Multimodal Research Platform (version 2.0, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://keyan.deepwise.com\u003c/span\u003e\u003cspan address=\"https://keyan.deepwise.com\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eFeature Selection\u003c/h2\u003e \u003cp\u003eInter-observer and intra-observer repeatability were assessed to evaluate the stability of the extracted radiomics features. Fifteen cases were randomly selected to create two identical DWI image labeling tasks. ROIs were segmented independently by observer 1 and observer 2 (both with 10 years of neuro-MRI diagnostic experience). The radiomic features obtained were used for inter-observer ICC analysis. Observer 1 re-segmented the above 15 cases at monthly intervals, the first and second extracted radiomic features were used for intra-observer ICC analysis; finally, Observer 1 segmented the ROI of all the remaining cases.\u003c/p\u003e \u003cp\u003eFeature selection is a key step in the process of model building. First, unstable radiomics features were removed based on the results of intra- and inter-observer ICC analysis. A feature correlation analysis with a threshold of 0.7 was used to screen out redundant features. F-test was conducted for further feature screening\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eModel construction and evaluating\u003c/h2\u003e \u003cp\u003eSupport vector machine is a linear classifier with a maximum interval defined on the feature space, and its algorithm is an optimization algorithm for solving convex quadratic programming. Previous studies suggest that SVM is effective in high-dimensional feature space, and its sparse feature can suppress data noise [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. As a result, SVM was used in this study.\u003c/p\u003e \u003cp\u003eThe dataset of 80 samples was randomly divided in a ratio of 7:3. 56 samples were used as a training set to train the model and 24 samples were used as testing set to validate the performance of the model. Based on the features of single-sequence images (DWI, T1-CE, ADC) and multiple-sequence images (ADC\u0026thinsp;+\u0026thinsp;DWI, ADC\u0026thinsp;+\u0026thinsp;T1-CE, DWI\u0026thinsp;+\u0026thinsp;T1-CE, ADC\u0026thinsp;+\u0026thinsp;DWI\u0026thinsp;+\u0026thinsp;T1-CE), seven radiomics models were established using the SVM algorithm, respectively. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e describes the workflow of the current study.\u003c/p\u003e \u003cp\u003eThe performance of these models was evaluated by calculating AUC, accuracy, sensitivity and specificity. In order to analyze the clinical efficacy of the best machine learning model, decision curve was plotted in this study. Furthermore, Delong test was applied to calculate whether the models were statistically different.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eThe chi-square test was performed to compare type of lesions, type of enhancement, degree of diffusion, peritumor edema, butterfly pattern, signs of bleeding and gender between patients with GCB subtype and those with non-GCB subtype. Independent samples t-test was utilized to compare the ages of patients across the two pathology types. Delong test was applied to calculate the statistical differences among the different models by medcalc (version 20.014, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.medcalc.org\u003c/span\u003e\u003cspan address=\"https://www.medcalc.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Statistical analysis and plots were conducted by the Deepwise Multimodal Research Platform, R statistical software application (version 4.2.2, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.Rproject.org\u003c/span\u003e\u003cspan address=\"http://www.Rproject.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) and SPSS 25.0. P values\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was regarded as statistically significant.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003ePatient Characteristics\u003c/h2\u003e \u003cp\u003eAmong the 80 cases, 24 (9 men; age 56.4\u0026thinsp;\u0026plusmn;\u0026thinsp;15.1 years) and 56 (35 men; age 61.1\u0026thinsp;\u0026plusmn;\u0026thinsp;9.3 years) were in the GCB group and the non-GCB group, respectively. As shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, there were statistical differences between GCB patients and non-GCB patients in terms of both gender and age (both p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Additionally, we found that GCB subtype mostly located in the frontal lobe, temporal lobe and lateral ventricle, non-GCB subtype mostly located in the frontal lobe, temporal lobe and corpus callosum, as shown in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. Moreover, diffusion restriction was observed in all GCB subtypes, however there were 19.64% of non-GCB subtypes that performed with no diffusion restriction (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). There were also statistical differences in the tumor edema between the two histological subtypes (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). In order to facilitate clinical application, this study defined grade 0 and 1 edema as mild edema and grade 2 and 3 edema as severe edema. We found that most GCB subtype was mild edema (62.5%) and most non-GCB subtype was severe edema (66.07%), with a statistically significant difference between them (p\u0026thinsp;=\u0026thinsp;0.018). According to Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e and Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, the main factors in decision making process between GCB and non-GCB are gender and age of the patients, presence or absence of diffusion restriction and the degree of edema. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e illustrates MRI images and pathology of GCB subtype lymphoma and non-GCB subtype lymphoma.\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\u003eCharacteristics of GCB and non-GCB patients.\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=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" 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\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003ePatient characteristics\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGCB subtype\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003enon-GCB subtype\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\u003eGender\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e37.50%\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e62.50%\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.039\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e62.50%\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e37.50%\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.039\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\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\u003cb\u003e0.029\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \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\u003eLesion sites of GCB and non-GCB patients.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLesion sites\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGCB subtype\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003enon-GCB subtype\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFrontal lobe\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTemporal lobe\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e20\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eParietal lobe\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOccipital lobe\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCerebellum\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCorpus callosum\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBasal ganglia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLateral ventricle\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eThird ventricle\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\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\u003eMRI findings of GCB and non-GCB.\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=\"char\" char=\".\" 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\u003eMRI finding\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGCB\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003enon-GCB\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\u003eType of lesions\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003esolitary demarcated lesion\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e33.33%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e33.93%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.959\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003esolitary infiltrative type\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20.83%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e7.14%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.165\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ediffuse infiltrative type\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.79%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003emultiple infiltrative lesion\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e45.83%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e57.14%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.353\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eType of enhancement\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ehomogeneous\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e33.33%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e30.36%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.068\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003enonhomogeneous\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e45.83%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e39.29%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.586\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003epresence of necrosis\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20.83%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e28.57%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.471\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ediffuse infiltrative\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.79%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.000\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDWI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003efree diffusion\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e100.00%\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e80.36%\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.047\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003erestricted diffusion in some part of the solid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0%\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e19.64%\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.047\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTumor edema\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003egrade 0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12.50%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5.36%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.517\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003egrade 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e50.00%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e28.57%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.066\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003egrade 2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e29.17%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e16.07%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.300\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003egrade 3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e8.33%\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e50.00%\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003emild\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e62.50%\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e33.93%\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.018\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003esevere\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e37.50%\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e66.07%\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.018\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eButterfly pattern\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12.50%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e21.43%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.532\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSigns of bleeding\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20.83%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e28.57%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.471\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eFeature Selection and Establishment of Radiomics Models\u003c/h2\u003e \u003cp\u003eA total of 1906,1906 and 1743 radiomics features were extracted from each patient 's ADC, DWI and T1-CE images, respectively. After inter-observer and intra-observer ICC analysis, ADC, DWI and T1-CE retained 1497,1447 and 1530 features, respectively. Finally, after feature reduction, 9 features of ADC images, 10 features of DWI images and 9 features of T1-CE features were retained.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eModels performance\u003c/h2\u003e \u003cp\u003eThe models that established by features from individual MRI sequence achieved an AUC from 0.753 to 0.976 on training set and from 0.641 to 0.746 on testing set. Based on features from a combination of sequences, the models achieved an AUC from 0.830 to 0.879 on training set and from 0.725 to 0.831 on testing set. And based on features combined all different sequences, the model had the best performance with an AUC of 0.902, accuracy of 0.813, specificity of 0.871, sensitivity of 0.797 on training set and an AUC of 0.854, accuracy of 0.790, specificity of 0.539, sensitivity of 0.857 on testing set. The results were shown in supplementary material. Furthermore, there were statistical significance of differences between ADC\u0026thinsp;+\u0026thinsp;DWI\u0026thinsp;+\u0026thinsp;T1-CE model and the ADC model (p\u0026thinsp;=\u0026thinsp;0.005), DWI model (p\u0026thinsp;=\u0026thinsp;0.018) and T1-CE model (p\u0026thinsp;=\u0026thinsp;0.003) on the training set, respectively. And there were statistical differences between ADC\u0026thinsp;+\u0026thinsp;DWI\u0026thinsp;+\u0026thinsp;T1-CE model and the ADC model (p\u0026thinsp;=\u0026thinsp;0.048) and DWI model (p\u0026thinsp;=\u0026thinsp;0.043) on testing set, respectively, as shown in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDelong test results of radiomics models based on different MRI sequences. Results with p value \u0026lt; 0.05 were marked coarse.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e value of the training set\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e value of the testing set\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eADC VS. DWI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.839\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.744\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eADC VS. T1-CE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.502\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDWI VS. T1-CE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.406\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eADC VS. ADC\u0026thinsp;+\u0026thinsp;DWI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.410\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eADC VS. ADC\u0026thinsp;+\u0026thinsp;T1-CE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.128\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eADC VS. DWI\u0026thinsp;+\u0026thinsp;T1-CE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.665\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDWI VS. ADC\u0026thinsp;+\u0026thinsp;DWI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.359\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDWI VS. ADC\u0026thinsp;+\u0026thinsp;T1-CE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.082\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDWI VS. DWI\u0026thinsp;+\u0026thinsp;T1-CE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.002\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.500\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1-CE VS. ADC\u0026thinsp;+\u0026thinsp;T1-CE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.029\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.301\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1-CE VS. DWI\u0026thinsp;+\u0026thinsp;T1-CE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.022\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.901\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1-CE VS. ADC\u0026thinsp;+\u0026thinsp;DWI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.148\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.867\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eall VS. T1-CE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.003\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.104\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eall VS. ADC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.005\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.048\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eall VS. DWI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.018\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.043\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eall VS. ADC\u0026thinsp;+\u0026thinsp;T1-CE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.152\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.738\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eall VS. DWI\u0026thinsp;+\u0026thinsp;T1-CE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.525\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.163\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eall VS. ADC\u0026thinsp;+\u0026thinsp;DWI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.026\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.120\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eADC\u0026thinsp;+\u0026thinsp;T1-CE VS. DWI\u0026thinsp;+\u0026thinsp;T1-CE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.683\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.382\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eADC\u0026thinsp;+\u0026thinsp;T1-CE VS. ADC\u0026thinsp;+\u0026thinsp;DWI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.391\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.475\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eADC\u0026thinsp;+\u0026thinsp;DWI VS. DWI\u0026thinsp;+\u0026thinsp;T1-CE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.252\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.715\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e shows the ROC curves of all the radiomics models and their comparison. The results of ROC curves showed that the model of all sequences combined had the highest AUC on both training and testing set, which suggested it had the best prediction performance.\u003c/p\u003e \u003cp\u003eTo analyze the clinical value of the best model in this study, we conducted the clinical decision curve analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). We found that using the radiomics model in this study has more benefits than treating all or none of the patients.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eWith the advance in transplantation techniques and the increasing number of patients treated with immunosuppressive therapy, the incidence of PCNSL has been on the rise in recent years [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. The most common histological subtype in PCNSL is DLBCL [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. With the advent of DNA microarray technology, DLBCL has been classified into two different subtypes based on different gene expression patterns: GCB and non-GCB [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. A combined HD-MTX-based regimen was recommended in the Chinese expert consensus published in 2022 [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Nevertheless, studies have shown that patients with non-GCB subtype respond less well to upfront R-CHOP chemotherapy compared to those with GCB subtype, which leads to a relatively poor prognosis for non-GCB patients [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. Therefore, lenalidomide and/or ibrutinib are often added to existing R-CHOP chemotherapy regimens in order to improve efficacy in the treatment of non-GCB patients [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Early identification of the patient's pathological type is important to improve the patient's prognosis and achieve optimal neurological recovery and disease control [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Therefore, this study analyzes the imaging characteristics of GCB and non-GCB patients and constructs a radiomics model for tumor differentiation by analyzing medical images, which can help radiologists to differentiate GCB lymphoma from non-GCB lymphoma to a certain extent and can support clinicians in treatment decisions.\u003c/p\u003e \u003cp\u003ePast study has shown that T1-CE and DWI can be non-invasive and accurate in assessing the disease status of patients [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Therefore, in this study, TI-CE images, DWI and ADC images of PCNSL patients were analyzed. We found that between the 2 groups there were statistically significant differences (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) in terms of the degree of diffusion restriction, degree of peritumoral edema, and gender. Due to the high nucleoplasmic ratio of tumor cells in PCNSL, the diffuse movement of water molecules in the tumor cells is restricted and appears as a distinct high signal on DWI images [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. Yet, in the current study, 18.64% of non-GCB cases had unrestricted diffusion, and the difference in the degree of diffusion between GCB and non-GCB lymphomas was statistically significant. This may be caused by the different cellular origins of non-GCB lymphomas. Non-GCB lymphomas can be derived from plasma cells, plasmacytes, immunoblasts, marginal zone cells and memory cells. Non-GCB lymphomas can be derived from plasma cells, plasmablast, immunoblasts, marginal zone B cells and memory cells. Both lymphoplasmacytic lymphoma and nodal marginal zone B-cell lymphoma have abundant cytoplasm and low nucleoplasmic ratio. This may lead to the appearance of isosignal on DWI images for some non-GCB lymphomas.\u003c/p\u003e \u003cp\u003eThe results showed a statistically significant difference (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) in the degree of edema between GCB subtype and non-GCB subtype. This may be since PCNSL is highly invasive and often destroys vascular endothelial cells, which leads to disruption of the blood-brain barrier [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. When the blood-brain barrier is disrupted, water, electrolytes and plasma proteins from the microvasculature surrounding the tumor leak out of the blood vessels, causing vasogenic edema. Phosphorylated STAT3 is the \"trigger point\" of the JAK-STAT pathway, which is associated with cell proliferation, and phosphorylated STAT3 activates the JAK-STAT pathway to cause cell proliferation. Yamada et al. found that phosphorylated STAT3 was expressed at a higher rate in the non-GCB subtype than in the GCB subtype [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. This suggests that more rapid cell proliferation accelerates the disruption of the blood-brain barrier, resulting in significantly higher levels of peritumoral edema in non-GCB subtype than that in GCB subtype. This phenomenon has not been reported in the literature yet. In addition, after analyzing the gender of the patients, we found that the majority of male patients presented with non-GCB subtype and the majority of female patients presented with GCB subtype. The difference was statistically significant. However, the reasons for this difference need to be further explored. Intratumoral hemorrhage is usually considered to be rare in PCNSLs, however, Yamada et al [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] found that all six of the non-GCB patients they reported had significant intratumoral hemorrhage and we also found intratumoral hemorrhage in non-GCB subtype. However, we observed no statistically significant difference in signs of bleeding between GCB subtype and non-GCB subtype. What's more, this study found that GCB subtype was mainly located in the temporal lobe and frontal lobe as well as the lateral ventricle, while the majority of non-GCB subtype occurred in the temporal lobe, frontal lobe and corpus callosum. Overall, the majority of both GCB subtype and non-GCB subtype occurred near the midline, which is consistent with the past study [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eWith the arise and development of radiomics, the information can be extracted, filtered and analyzed by machine learning algorithms to build radiomics models for differential diagnosis [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e], predicting prognosis [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e] and predicting treatment response [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. In this study, based on the results of the analysis of medical images described above, we have developed a radiomics model that can differentiate GCB subtype from non-GCB subtype. To ensure the robustness of the model, four feature screening steps were taken to obtain the features needed to develop the model for this study. Firstly, we standardized the data. Then, radiomics features with high stability (ICC\u0026thinsp;\u0026ge;\u0026thinsp;0.8) were retained for the next step of feature screening based on the results of intra- and inter-observer ICC analysis. Thirdly, A feature correlation analysis with a threshold of 0.7 was applied to screen redundant features. Finally, an F-test was taken to obtain the features needed for the radiomics models.\u003c/p\u003e \u003cp\u003eAs shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, models constructed by all sequences combined, sequence combination (ADC\u0026thinsp;+\u0026thinsp;T1-CE) and radiomics features in T1-CE images had a good diagnostic performance. In the validation set, the model developed by all sequences combined had the best diagnostic performance with an AUC of 0.854 (95% CI 0.756\u0026ndash;0.953). This may result from the fact that the radiomics features extracted from all sequences combined can reflect the biological information of the tumor more comprehensively and accurately. MRI examination and radiomics analysis were two important methods of differentiating GCB subtype from non-GCB subtype. T1-CE was the key sequence for both methods. In past studies, T1-CE images of PCNSLs were also collected when analyzing the lesion sites of the distribution of GCB subtype and non-GCB subtype [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. Blasel et al. [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e] suggested that the increased rCBV values in PCNSLs were mainly due to immunoreactive changes of the normal brain vasculature against the infiltrating lymphoma cells. These tumor-vascular interactions would lead to microvascular proliferation. In addition, Hartert et al. [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] showed that phosphorylated STAT3 was expressed at a higher level in non-GCB subtype than in GCB subtype. It is implied that non-GCB lymphoma cells proliferate more rapidly and have a more pronounced disruption to the blood-brain barrier, which results in more intense immunoreactive changes. In addition, the two subtypes have different cellular origins and different nucleocytoplasmic ratio [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e], and therefore different degrees of diffusion on DWI and ADC images. Hence, the differential diagnostic performance of the model constructed by all sequences combined was higher than that of the models constructed with sequence combination (ADC\u0026thinsp;+\u0026thinsp;T1-CE) and T1-CE.\u003c/p\u003e \u003cp\u003eMachine learning algorithms can objectively analyze high-throughput medical image information to develop radiomics models that can differentiate GCB subtype from non-GCB subtype. Different radiologists have different diagnostic experiences, there is subjectivity in the diagnosis of the disease and it is difficult to capture some of the imaging features with the naked-eye. As a result, there may be inter-individual variation in the diagnostic accuracy of radiologists, and junior radiologists may have a slightly lower diagnostic performance than that of the radiomics models [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. In addition, compared to stereotactic biopsy, radiomics can be more widely available, more reflective of the overall tumor level and have few side effects. Additionally, ADC, DWI and T1-CE were routine MRI examination sequences for the preoperative diagnosis in patients with PCNSLs. Therefore, the radiomics model developed in this study can help clinicians determine the pathologic classification of PCNSLs with contraindications to biopsy, provide more reliable decision support for the determination of chemotherapy regimens for patients, and is more generalizable in clinical work. These advantages were supported in the decision analysis curve (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThere were some limitations in our study. Firstly, the number of patients was insufficient and the pathological types of cases were limited to GCB and non-GCB subtypes. Secondly, there was no external validation set for this study, which should be included in future studies. Finally, the edematous, enhanced and necrotic areas of the tumor were not outlined independently. With the further development of automated brain tumor segmentation techniques, biological information from different regions of the tumor will need to be explored in future studies.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eConventional MRI can distinguish to a certain extent between GCB subtypes and non-GCB subtypes. The GCB subtypes are common in females, most cases present mild peritumoral edema, and diffusion restriction is detected in all of them; however, the non-GCB subtypes are more common in males and severe peritumoral edema can be observed in most cases. Additionally, the radiomics model of all sequences combined has the potential in distinguishing between GCB and non-GCB.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003e \u003cp\u003eADC Apparent diffusion coefficient\u003c/p\u003e \u003cp\u003eAUC Area under the curve\u003c/p\u003e \u003cp\u003eBCL-6 B-cell lymphoma oncogene-6\u003c/p\u003e \u003cp\u003eDCA Decision curve analysis\u003c/p\u003e \u003cp\u003eDWI Diffusion-weighted imaging\u003c/p\u003e \u003cp\u003eFOV Field of view\u003c/p\u003e \u003cp\u003eGCB Germinal center B cell\u003c/p\u003e \u003cp\u003eGEP Gene expression profiling\u003c/p\u003e \u003cp\u003eGRE Gradient echo sequence\u003c/p\u003e \u003cp\u003eMUM-1 Multiple myeloma oncogene-1\u003c/p\u003e \u003cp\u003eNHL Non-hodgkin's lymphoma\u003c/p\u003e \u003cp\u003eOS Overall survival\u003c/p\u003e \u003cp\u003ePACS Picture archiving and communication systems\u003c/p\u003e \u003cp\u003ePCNSL Primary central nervous system lymphoma\u003c/p\u003e \u003cp\u003ePFS Progression-free survival\u003c/p\u003e \u003cp\u003eR-CHOP Rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisolone\u003c/p\u003e \u003cp\u003eSE Spin echo\u003c/p\u003e \u003cp\u003eTE Echo time\u003c/p\u003e \u003cp\u003eTR Repetition time\u003c/p\u003e \u003cp\u003eT1-CE T1 contrast-enhanced\u003c/p\u003e \u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgment\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors would like to thank all the participants.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u003c/strong\u003e\u003cstrong\u003e\u0026rsquo;\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;contributions\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eYelong Shen: Data curation, Formal analysis, Methodology, Project administration, Writing \u0026ndash; original draft, Writing \u0026ndash; review \u0026amp; editing. Siyu Wu: Formal analysis, Methodology, Visualization, Writing \u0026ndash; original draft, Writing \u0026ndash; review \u0026amp; editing. Yanan Wu: Data curation. Chao Cui: Data curation. Haiou Li: Data curation. Shuang Yang: Data curation. Xuejun Liu: Data curation. Xingzhi Chen: Project. Chencui Huang: Project. Ximing Wang: Conceptualization, Data curation, Project administration, Supervision, Writing \u0026ndash; review \u0026amp; editing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe present study was supported by the National Natural Science Foundation of China (Grant Nos. 8187354, 81571672), and Academic promotion program of Shandong First Medical University (Grant No. 2019QL023).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data that support the fundings of this study are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis retrospective study was approved by the Ethics Committee of the Shandong Provincial Hospital and written informed consent for the participants over 16 years of age was discarded because of the retrospective nature of this study. For participants less than 16 years of age, written informed consent has been obtained from their parents/guardians after a brief description of the purpose and objectives of the study has been given to them. We confirmed that the study was carried out in accordance with relevant guidelines and regulations of Helsinki Declaration.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll the authors declare that they have no conflict of interest.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSchaff LR, Grommes C. Primary central nervous system lymphoma. Blood. 2022;140(9):971\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e.https://www.ncbi.nlm.nih.gov/pubmed/34699590\u003c/span\u003e\u003cspan address=\"http://.https://www.ncbi.nlm.nih.gov/pubmed/34699590\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoldhoff M, Mrugala MM, Grommes C, Kaley TJ, Swinnen LJ, Perez-Heydrich C, Nayak L. Challenges in the Treatment of Newly Diagnosed and Recurrent Primary Central Nervous System Lymphoma. J Natl Compr Canc Netw. 2020;18(11):1571\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/33152700\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/33152700\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMarcus C, Maragkos GA, Alterman RL, Uhlmann E, Pihan G, Varma H. GCB-type is a favorable prognostic factor in primary CNS diffuse large B-cell lymphomas. J Clin Neurosci. 2021;83:49\u0026ndash;55. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jocn.2020.11.031\u003c/span\u003e\u003cspan address=\"10.1016/j.jocn.2020.11.031\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoy A, Ramchandren R, Ghosh N, Munoz J, Morgan DS, Dang NH, Knapp M, Delioukina M, Kingsley E, Ping J, et al. Ibrutinib plus lenalidomide and rituximab has promising activity in relapsed/refractory non-germinal center B-cell-like DLBCL. Blood. 2019;134(13):1024\u0026ndash;36.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHou K, Yu Z, Jia Y, Fang H, Shao S, Huang L, Feng Y. Efficacy and safety of ibrutinib in diffuse large B-cell lymphoma: A single-arm meta-analysis. Crit Rev Oncol Hematol 2020, 152103010\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/32540781\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/32540781\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchmitz R, Wright GW, Huang DW, Johnson CA, Phelan JD, Wang JQ, Roulland S, Kasbekar M, Young RM, Shaffer AL, et al. Genetics and Pathogenesis of Diffuse Large B-Cell Lymphoma. N Engl J Med. 2018;378(15):1396\u0026ndash;407.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNowakowski GS, Chiappella A, Gascoyne RD, Scott DW, Zhang Q, Jurczak W, \u0026Ouml;zcan M, Hong X, Zhu J, Jin J, et al. ROBUST: A Phase III Study of Lenalidomide Plus R-CHOP Versus Placebo Plus R-CHOP in Previously Untreated Patients With ABC-Type Diffuse Large B-Cell Lymphoma. J Clin Oncol. 2021;39(12):1317\u0026ndash;28.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLohmann P, Franceschi E, Vollmuth P, Dhermain F, Weller M, Preusser M, Smits M, Galldiks N. Radiomics in neuro-oncological clinical trials. Lancet Digit Health. 2022;4(11):e841\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/36182633\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/36182633\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMayerhoefer ME, Materka A, Langs G, H\u0026auml;ggstr\u0026ouml;m I, Szczypiński P, Gibbs P, Cook G. Introduction to Radiomics. J Nucl Med. 2020;61(4):488\u0026ndash;95.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarajas RF, Politi LS, Anzalone N, Schoder H, Fox CP, Boxerman JL, Kaufmann TJ, Quarles CC, Ellingson BM, Auer D, et al. Consensus recommendations for MRI and PET imaging of primary central nervous system lymphoma: guideline statement from the International Primary CNS Lymphoma Collaborative Group (IPCG). Neuro Oncol. 2021;23(7):1056\u0026ndash;71. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/33560416\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/33560416\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKuker W, Nagele T, Korfel A, Heckl S, Thiel E, Bamberg M, Weller M, Herrlinger U. Primary central nervous system lymphomas (PCNSL): MRI features at presentation in 100 patients. J Neurooncol. 2005;72(2):169\u0026ndash;77. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/15925998\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/15925998\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJin X, Zheng X, Chen D, Jin J, Zhu G, Deng X, Han C, Gong C, Zhou Y, Liu C, et al. Prediction of response after chemoradiation for esophageal cancer using a combination of dosimetry and CT radiomics. Eur Radiol. 2019;29(11):6080\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/31028447\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/31028447\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYuan Y, Ding T, Wang S, Chen H, Mao Y, Chen T. Current and emerging therapies for primary central nervous system lymphoma. Biomark Res. 2021;9(1):32. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/33957995\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/33957995\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang JM, Jang JY, Jeon YK, Paik JH. Clinicopathologic implication of microRNA-197 in diffuse large B cell lymphoma. J Transl Med. 2018;16(1):162. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/29890998\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/29890998\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen T, Liu Y, Wang Y, Chang Q, Wu J, Wang Z, Geng D, Yu JT, Li Y, Li XQ, et al. Evidence-based expert consensus on the management of primary central nervous system lymphoma in China. J Hematol Oncol. 2022;15(1):136. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/36176002\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/36176002\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchultz CJ, Bovi J. Current management of primary central nervous system lymphoma. Int J Radiat Oncol Biol Phys. 2010;76(3):666\u0026ndash;78. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/20159361\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/20159361\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNowakowski GS, Hong F, Scott DW, Macon WR, King RL, Habermann TM, Wagner-Johnston N, Casulo C, Wade JL, Nagargoje GG, et al. Addition of Lenalidomide to R-CHOP Improves Outcomes in Newly Diagnosed Diffuse Large B-Cell Lymphoma in a Randomized Phase II US Intergroup Study ECOG-ACRIN E1412. J Clin Oncol. 2021;39(12):1329\u0026ndash;38. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/33555941\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/33555941\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilson WH, Wright GW, Huang DW, Hodkinson B, Balasubramanian S, Fan Y, Vermeulen J, Shreeve M, Staudt LM. Effect of ibrutinib with R-CHOP chemotherapy in genetic subtypes of DLBCL. Cancer Cell. 2021;39(12):1643\u0026ndash;e16531643. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/34739844\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/34739844\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHartert KT, Wenzl K, Krull JE, Manske M, Sarangi V, Asmann Y, Larson MC, Maurer MJ, Slager S, Macon WR et al. Targeting of inflammatory pathways with R2CHOP in high-risk DLBCL. Leukemia 2021, 35(2):522\u0026ndash;33\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/32139889\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/32139889\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKinoshita M, Sasayama T, Narita Y, Yamashita F, Kawaguchi A, Chiba Y, Kagawa N, Tanaka K, Kohmura E, Arita H, et al. Different spatial distribution between germinal center B and non-germinal center B primary central nervous system lymphoma revealed by magnetic resonance group analysis. Neuro Oncol. 2014;16(5):728\u0026ndash;34. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/24497406\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/24497406\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYamada S, Muto J, Iba S, Shiogama K, Tsuyuki Y, Satou A, Ohba S, Murayama K, Sugita Y, Nakamura S, et al. Primary central nervous system lymphomas with massive intratumoral hemorrhage: Clinical, radiological, pathological, and molecular features of six cases. Neuropathology. 2021;41(5):335\u0026ndash;48. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/34254378\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/34254378\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu D, Kong Z, Wang Y, Chen W, Wang Y, Chen W, Liu L, Dang Y, Ma W, Wang Y, et al. Quantitative and Visual Characteristics of Primary Central Nervous System Lymphoma on (18)F-FDG-PET. Interdiscip Sci. 2019;11(2):300\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/31264053\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/31264053\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBathla G, Priya S, Liu Y, Ward C, Le NH, Soni N, Maheshwarappa RP, Monga V, Zhang H, Sonka M. Radiomics-based differentiation between glioblastoma and primary central nervous system lymphoma: a comparison of diagnostic performance across different MRI sequences and machine learning techniques. Eur Radiol. 2021;31(11):8703\u0026ndash;13. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/33890149\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/33890149\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhong S, Ren JX, Yu ZP, Peng YD, Yu CW, Deng D, Xie Y, He ZQ, Duan H, Wu B et al. Predicting glioblastoma molecular subtypes and prognosis with a multimodal model integrating convolutional neural network, radiomics, and semantics. J Neurosurg 2022:1\u0026ndash;10\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/36461822\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/36461822\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAbdel Razek AAK, Alksas A, Shehata M, AbdelKhalek A, Abdel Baky K, El-Baz A, Helmy E. Clinical applications of artificial intelligence and radiomics in neuro-oncology imaging. Insights Imaging. 2021;12(1):152. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/34676470\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/34676470\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlasel S, Vorwerk R, Kiyose M, Mittelbronn M, Brunnberg U, Ackermann H, Voss M, Harter PN, Hattingen E. New MR perfusion features in primary central nervous system lymphomas: pattern and prognostic impact. J Neurol. 2018;265(3):647\u0026ndash;58. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/29383512\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/29383512\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi S, Young KH, Medeiros LJ. Diffuse large B-cell lymphoma. Pathology. 2018;50(1):74\u0026ndash;87. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/29167021\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/29167021\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXia W, Hu B, Li H, Geng C, Wu Q, Yang L, Yin B, Gao X, Li Y, Geng D. Multiparametric-MRI-Based Radiomics Model for Differentiating Primary Central Nervous System Lymphoma From Glioblastoma: Development and Cross-Vendor Validation. J Magn Reson Imaging. 2021;53(1):242\u0026ndash;50. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/pubmed/32864825\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/pubmed/32864825\" targettype=\"URL\" 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":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Machine learning. Primary central nervous system lymphoma. Magnetic resonance imaging. Pathological classification. Radiomics","lastPublishedDoi":"10.21203/rs.3.rs-4505854/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4505854/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjectives\u003c/h2\u003e \u003cp\u003eTo investigate the characteristics and pathological basis of MRI in germinal center B cell (GCB) and non-germinal center B cell (non-GCB) in PCNSL (primary central nervous system lymphoma). And to explore the predictive ability of MRI radiomics-based in differentiating the GCB and non-GCB of PCNSL.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThis study retrospectively analyzed standard diagnostic MRI examinations in 24 immunocompetent patients (9 men; age 56.4\u0026thinsp;\u0026plusmn;\u0026thinsp;15.1 years) with GCB and 56 immunocompetent patients (35 men; age 61.1\u0026thinsp;\u0026plusmn;\u0026thinsp;9.3 years) with non-GCB. The radiomics features were extracted from ADC, DWI, and T1-CE images respectively, and the features were screened by machine learning algorithm and statistical method. Finally, radiomics models of seven different sequence permutations were constructed. The area under the receiver operating characteristic (ROC AUC) curve was used to evaluate the predictive performance of all models. Delong test was utilized to compare the differences among models.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe GCB cases all showed diffusion restriction, which was observed in 80.36% of the non-GBM cases; p\u0026thinsp;\u0026lt;\u0026thinsp;0.05. Grade 3 edema was rare in GCB cases (8.33%) and common in non-GCB cases (50.00%); p\u0026thinsp;\u0026lt;\u0026thinsp;0.001. 62.50% of male patients were non-GCB and 37.50% of female patients were non-GCB; p\u0026thinsp;\u0026lt;\u0026thinsp;0.05. Additionally, patients with the GCB subtype are younger than those with the non-GCB subtype; p\u0026thinsp;\u0026lt;\u0026thinsp;0.05. The best prediction model in our study used a combination of ADC, DWI, and T1-CE achieving the highest AUC of 0.854. And there was a significant difference between the best-combined model and some of the other models.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThe GCB subtype is commonly seen in women, with mild peritumoral edema in most cases and diffusion restriction in all cases; however, the non-GCB subtype is commonly seen in men, with severe peritumoral edema in most cases. Additionally, the radiomics model developed by all sequences combined had good performance in discriminating between GCB and non-GCB.\u003c/p\u003e","manuscriptTitle":"Differentiating between PCNSL GCB Subtype and Non-GCB Subtype using Radiomics: A Multicenter Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-06-26 15:04:58","doi":"10.21203/rs.3.rs-4505854/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-06-24T16:23:31+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-05-14T19:45:17+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"180587170332839588641944006782516036976","date":"2025-05-07T14:02:11+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-07-08T23:40:56+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"176658809500981632175501781865379998648","date":"2024-07-08T03:12:46+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-07-02T04:36:13+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2024-06-06T11:57:45+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-06-06T08:53:21+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-06-06T08:52:52+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Imaging","date":"2024-05-31T03:04:55+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"11de2ca0-78d3-43c6-9887-d9b517b8f6da","owner":[],"postedDate":"June 26th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-10T07:08:32+00:00","versionOfRecord":[],"versionCreatedAt":"2024-06-26 15:04:58","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4505854","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4505854","identity":"rs-4505854","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}