Deep Learning and Pathomics Analyses predict prognosis of high-grade gliomas

Yuchen Zhu¹Yuxi Gong¹Weilin Xu¹Xingjian Sun¹Gefei Jiang¹Lei Qiu¹Kexin Shi¹Wu Mengxing¹Yinjiao Fei¹Jinling Yuan¹Jinyan Luo¹Yurong Li²Yuandong Cao¹Minhong Pan¹SHU ZHOU^1*

• ¹First Affiliated Hospital, Nanjing Medical University, Nanjing, China
• ²Second Affiliated Hospital, School of Medicine, Zhejiang University, Hangzhou, Zhejiang Province, China

Utilizing pathomics to analyze high-grade gliomas and provide prognostic insights. Initially, regions of interest (ROIs) within the entire slice images (WSIs) were identified and underwent cropping, removal of blank areas, and standardization to select tumor patches. Subsequently, a deep learning model trained on these patches aggregated predictions for WSIs. Finally, pathological features were extracted using Pearson correlation, univariate Cox regression, and LASSO-Cox regression, and three models were developed: a pathology-based model, a clinical model, and a combined model integrating both. In the study, the combined model demonstrated the best performance, with a C-index of 0.847 in the training set and 0.739 in the test set. Based on the combined model, the patient population was divided into high-risk and low-risk groups.The median progression-free survival (PFS) for high-risk patients was 10 months (p<0.001), while the median PFS for low-risk patients was not reached. Stratification by IDH status revealed significant differences in PFS. In conclusion, the combined model can better predict the progression risk of high-grade gliomas and provides valuable guidance for personalized treatment of high-grade gliomas.