Purpose To evaluate the value of preoperative MRI-based radiomic models for assessing tumor-infiltrating CD8+T-cell expression in glioblastoma patients,and to identify the most stable and efficient radiomic feature region for predicting prognosis following immunotherapy.Materials and Methods This retrospective study included 150 patients with histopathologically confirmed glioblastoma from Lanzhou University Second Hospital(January 2018 to April 2022).Tumor-infiltrating CD8+T-cell expression was quantitatively assessed using immunohistochemical staining,with patients stratified into CD8-high and CD8-low expression groups based on overall survival.A total of 1 185 radiomic features were extracted from each patient's contrast-enhanced T1C and T2WI images,covering the original tumor region and sequentially expanded peritumoral regions(2.5 mm,5.0 mm,7.5 mm,10.0 mm,12.5 mm,15.0 mm morphological dilation of tumor core+peritumoral area).Feature selection was performed using variance threshold,minimum redundancy maximum relevance,and least absolute shrinkage and selection operator methods.XGBoost classifier was employed to construct clinical,radiomic,and clinical-radiomic multimodal combined prediction models.Diagnostic performance was evaluated using receiver operating characteristic curve analysis.Results The radiomic model based on tumor expansion of 7.5 mm(tumor+peritumoral region)demonstrated optimal predictive performance.The clinical-radiomic multimodal combined model showed superior predictive capability compared to clinical and radiomic models alone,achieving an area under the curve of 0.991 and accuracy of 99.0%in the training set,and area under the curve of 0.840 with accuracy of 80.0%in the validation set.Conclusion MRI radiomics provides a feasible approach for evaluating tumor-infiltrating CD8+T-cell expression in glioblastoma patients,offering potential for preoperative prognosis prediction.

Purpose To evaluate the value of preoperative MRI-based radiomic models for assessing tumor-infiltrating CD8+T-cell expression in glioblastoma patients,and to identify the most stable and efficient radiomic feature region for predicting prognosis following immunotherapy.Materials and Methods This retrospective study included 150 patients with histopathologically confirmed glioblastoma from Lanzhou University Second Hospital(January 2018 to April 2022).Tumor-infiltrating CD8+T-cell expression was quantitatively assessed using immunohistochemical staining,with patients stratified into CD8-high and CD8-low expression groups based on overall survival.A total of 1 185 radiomic features were extracted from each patient's contrast-enhanced T1C and T2WI images,covering the original tumor region and sequentially expanded peritumoral regions(2.5 mm,5.0 mm,7.5 mm,10.0 mm,12.5 mm,15.0 mm morphological dilation of tumor core+peritumoral area).Feature selection was performed using variance threshold,minimum redundancy maximum relevance,and least absolute shrinkage and selection operator methods.XGBoost classifier was employed to construct clinical,radiomic,and clinical-radiomic multimodal combined prediction models.Diagnostic performance was evaluated using receiver operating characteristic curve analysis.Results The radiomic model based on tumor expansion of 7.5 mm(tumor+peritumoral region)demonstrated optimal predictive performance.The clinical-radiomic multimodal combined model showed superior predictive capability compared to clinical and radiomic models alone,achieving an area under the curve of 0.991 and accuracy of 99.0%in the training set,and area under the curve of 0.840 with accuracy of 80.0%in the validation set.Conclusion MRI radiomics provides a feasible approach for evaluating tumor-infiltrating CD8+T-cell expression in glioblastoma patients,offering potential for preoperative prognosis prediction.

Purpose To explore the value of three-dimensions densely connected convolutional networks(3D-DenseNet)in the differential diagnosis of high-grade gliomas(HGGs)and single brain metastases(BMs)via MRI,and to compare the diagnostic performance of models built with different sequences.Materials and Methods T2WI and T1WI contra-enhanced(T1C)imaging data of 230 cases of HGGs and 111 cases of BMs confirmed by surgical pathology in Lanzhou University Second Hospital from June 2016 to June 2021 were retrospectively collected,and the volume of interest under the 3D model was delineated in advance as the input data.All data were randomly divided into a training set(n=254)and a validation set(n=87)in a ratio of 7∶3.Based on the 3D-DenseNet,T2WI,T1C and two sequence fusion prediction models(T2-net,T1C-net and TS-net)were constructed respectively.The predictive efficiency of each model was evaluated and compared by the receiver operating characteristic curve,and the predictive performance of models built with different sequences were compared.Results The area under curve(AUC)of T1C-net,T2-net and TS-net in the training and validation sets were 0.852,0.853,0.802,0.721,0.856 and 0.745,respectively.The AUC and accuracy of the validation set of T1C-net were significantly higher than those of T2-net and TS-net,respectively,and the AUC and accuracy of the validation set of TS-net were significantly higher than those of T2-net.There was a significant difference between T1C-net and T2-net models(P<0.05),while there were no statistical differences between the models of TS-net and T2-net,T1C-net and TS-net(P>0.05).The T1C-net model based on 3D-DenseNet had the best performance,the accuracy of the validation set was 80.5%,the sensitivity was 90.9%,the specificity was 62.5%.Conclusion The 3D-DenseNet model based on MRI conventional sequence has better diagnostic performance,and the model built by T1C-net sequence has better performance in differentiating HGGs and BMs.Deep learning models can be a potential tool to identify HGGs and BMs and to guide the clinical formulation of precise treatment plans.

Objective:To explore the current status of the artificial intelligence (AI) developments in medical imaging in China, and to provide data for the development of AI.Methods:In May 2022, the Radiology Branch of the Chinese Medical Association and the China Medical Imaging AI Industry-University-Research Innovation Alliance jointly launched a nationwide survey on the application status and development needs of medical imaging AI in China in the form of a questionnaire. This survey was carried out for different groups of people, focusing on the clinical applications of medical imaging AI, enterprise development, and educational needs in colleges and universities, with the descriptive statistical analysis performed.Results:China′s medical imaging AI has made great progress in clinical applications, in enterprise developments, as well as in the education and teaching areas. In terms of clinical application, 90.8% (5 765/6 347) of the survey respondents had a preliminary understanding of AI. There were 62.1% (3 798/6 119) doctors confirmed the applications medical imaging AI products in their departments. AI products were applied in the whole process of medical imaging examination, especially in assistance of the diagnosis. The application of pulmonary nodules screening accounted for 89.5% (3 401/3 798) of all medical imaging AIs. The main factors restricting the rapid development of medical imaging AI included lack of experts [47.3% (3 002/6 347)], poor data quality [45.7% (2 898/6 347)] and imperfect function of the products [40.4% (2 566/6 347)]; in terms of enterprises, there were 65.4% enterprises with a scale of less than 100 employees (17/26), and 34.6% with a scale of more than 100 employees (9/26). The main group of the customers were the hospitals above the second level, accounting for about 92.3% (24/26); in terms of education, the number and quality of AI courses, practical operations and lectures currently carried out by schools vary between different levels. The AI courses for graduated students accounted for about 22.5% (86/381), which were the largest in number; while the proportion of AI courses for junior college students, undergraduates and regular trainees were less than 15%. More than 60% of the students thought it necessary for schools to establish AI courses. Among all the students, the master′s and doctoral candidates had the greatest demand for additional AI courses [84.8% (323/381)].Conclusions:The development and popularization of medical imaging AI in China continues to prosper, with opportunities and challenges coexisting. It is necessary to adhere to the orientation of clinical needs, and to realize the coordinated development of clinical application, enterprise development, as well as education and teaching.

Objective:To explore the value of a deep learning model based on MRI in predicting the methylation status of MGMT in WHO Ⅱ, Ⅲ gliomas.Methods:The clinical and imaging data of 121 patients with WHO grade Ⅱ, Ⅲ glioma confirmed by surgical pathology and molecular pathology in the Second Hospital of Lanzhou University from June 2016 to June 2020 were retrospectively analyzed. Among them, the MGMT promoter was methylated. A total of 78 cases were metabolized and 43 cases were unmethylated. T 2WI and T 1WI enhanced sequence images of 121 cases of WHO Ⅱ, Ⅲ gliomas were collected, and all the images of each patient including the lesion level were selected manually, and were randomly divided into training set and validation set according to 7∶3. The EfficientNet-B3 convolutional neural network was used to build independent prediction models (T 2-net, T 1C-net, TS-net) based on T 2WI, T 1WI enhancement, T 2WI combined with T 1WI enhancement, and the prediction performance of each model was evaluated separately through the ROC curve. Results:The T 2-net model in the validation set presented an accuracy of 72.3%, a sensitivity of 64.7%, a specificity of 73.3%, and an area under the curve (AUC) of 0.72 for predicting the methylation status of the MGMT promoter in WHO Ⅱ, Ⅲ gliomas. The T 1C-net model showed an accuracy of 66.8%, a sensitivity of 68.3%, a specificity of 66.9%, and an AUC of 0.72. The TS-net model showed an accuracy of 81.8%, a sensitivity of 63.1%, a specificity of 85.0%, and AUC of 0.78. Conclusions:The EfficientNet-B3 convolutional neural network based on MRI can predict the methylation status of the MGMT promoter of WHO Ⅱ, Ⅲ gliomas; the TS-net model has the best prediction performance.

In recent years, with the continuous development of artificial intelligence related technologies, deep learning (DL) technology kept improving and had become a research hotspot in medical field. Research of DL in medical imaging has brought a new development direction for precise diagnosis, individualized treatment and prognosis evaluation of brain neoplasms. The current status and future development of DL in brain neoplasms medical images were reviewed in this paper.