AI-assisted Diagnosis of Malignant Brain Tumors

Overview

This study aims to establish a large-scale, multi-center MRI database for malignant brain tumors. It will develop an artificial intelligence system for the segmentation and classification of multiple subtypes of brain tumors (including glioma, metastatic tumor and lymphoma et al.) using deep learning technology. This will address the issues of small sample sizes and limited classification performance in existing methods, thereby improving the accuracy of non-invasive preoperative diagnosis, reducing the need for biopsies, and having significant clinical translational value.

This study is mainly based on two centers, the Second Affiliated Hospital of Zhejiang University School of Medicine and the Zhejiang Cancer Hospital. It retrospectively collects cases of malignant brain tumors (including gliomas, brain metastases, and brain lymphomas) that have been confirmed by histopathology and have preoperative multimodal MRI images (mainly including CE-T1WI and T2-FLAIR). It is expected to include 3,000 cases. Axial CE-T1WI and T2-FLAIR images of all patients were obtained on 3.0T or 1.5T magnetic resonance imaging systems. A large-scale, multi-center MRI image database for common malignant brain tumors (gliomas, brain metastases, and brain lymphomas) was planned to be constructed. To address the automatic segmentation of complex lesion tissues in brain tumors and the auxiliary diagnosis of common malignant brain tumors, a deep learning technical approach was adopted. A deep learning-based multi-subtype brain tumor segmentation and classification diagnostic method was proposed, aiming to build an image artificial intelligence-assisted diagnostic system for common malignant brain tumors and improve the accuracy of auxiliary diagnosis of common brain malignancies.

Conditions