Application of an Artificial Intelligence Algorithm to Prognostically Stratify Grade II Gliomas.

Area of Science:

• Neuro-oncology
• Computational Biology
• Medical Statistics

Background:

• Extent of resection (EOR) and molecular classification impact low-grade glioma (LGG) prognosis.
• A comprehensive prognostic stratification integrating multiple factors for LGG patients is currently lacking.

Purpose of the Study:

• To develop a multiparametric prognostic stratification for grade II glioma patients.
• To integrate classic statistics and artificial intelligence for enhanced prognostic accuracy.

Main Methods:

• Analysis of clinical, neuroradiological, surgical, histopathological, and molecular data from 241 adult LGG patients.
• Assessment of prognostic predictors for overall survival (OS), progression-free survival (PFS), and malignant progression-free survival (MPFS).
• Application of a decision-tree algorithm for patient stratification.

Main Results:

• Extent of resection (EOR), tumor volumes, Ki67, and molecular classification independently predict OS, PFS, and MPFS.
• The decision tree identified prognostic factor hierarchy and cut-off levels for distinct prognostic classes.
• EOR demonstrated a superior prognostic role compared to molecular class; second surgery and molecular class-specific factors were also significant.

Conclusions:

• A novel stratification for LGG patients integrating clinical, molecular, and imaging data was proposed using a supervised non-parametric learning method.
• This approach may offer improved clinical utility for prognostic assessment in LGG.
• Validation in independent cohorts is necessary to confirm the clinical utility of this innovative stratification.