From Black Box to Glass Box Interpretable AI for Lung Cancer Radiomics

Abstract

Lung cancer is the leading cause of cancer mortality, with deaths due to the disease anticipated to exceed 4.6M new cases and 3.5 million lives each year by 2050. AI-based radiomics tools can potentially analyze the levels of pixel distribution and identify hidden biomarkers related to tumor aggressiveness. However, clinicians are reluctant to adopt these methodologies due to the lack of transparency in the DL models, known as the “black-box” phenomenon. The chapter “From Black Box to Glass Box: Interpretable AI for Lung Cancer Radiomics” aims to facilitate the transition of AI-based applications from laboratories to clinics. The protocol for AI-enhanced radiomics is systematically described. The core focus is on enhancing transparency through explainable AI (XAI) techniques. AI designs should be transparent (glass boxes) and have the purpose of assisting clinicians rather than replacing them. The roadmap to success and excellence in the diagnosis and treatment of lung cancer is clear, with data-driven insights.

Ähnliche Arbeiten