PathoNet

PathoNet quantifies Ki-67 expression and tumor-infiltrating lymphocytes (TILs) in breast cancer using deep learning to produce quantitative prognostic biomarkers.

Key Features:

• Automated Estimation: Automates evaluation of Ki-67 expression levels and TILs scores at the cell level to address inter-observer variability.
• Deep Learning Approach: Implements a deep neural network backend and advanced deep learning models for simultaneous prediction of Ki-67 and TILs metrics.
• Dataset Contribution: Is supported by the SHIDC-BC-Ki-67 dataset created for Ki-67 cell detection and annotated classification of cells in breast cancer.
• Performance Excellence: Demonstrates higher harmonic mean measures compared to existing state-of-the-art methods as reported.

Scientific Applications:

• Prognostic Assessment: Provides quantitative Ki-67 and TILs metrics for prognostication in heterogeneous tumors such as breast cancer.
• Treatment Stratification: Informs assessment of tumor progression and response to chemotherapy to aid personalized therapeutic decision-making.

Methodology:

The methodology integrates a deep neural network backend and deep learning models in a pipeline to simultaneously estimate Ki-67 expression and intratumoral TILs scores, with the SHIDC-BC-Ki-67 dataset providing annotated Ki-67 cell detection and classification.