A Trustworthy Second Opinion: Self-Explanatory AI for Lung Nodule Diagnosis

PhD defence by Jiahao Lu

Assessment Committee

Associate Professor Jens Petersen, Computer Science, University of Copenhagen (Chairperson)
Associate Professor Dimitrios Papadopoulos, DTU Computer
Assistant Professor Axel Carlier, ISAE-SUPAERO

Supervisors

Professor Sune Darkner

Professor Michael Bachman Nielsen

Department

Department of Computer Science

Place

Building: Vermundsgade, Room: SCI-DIKU-VG5-01-0-078, SCI-DIKU-VG5-01-0-074, Vermundsgade 5, 2100 Copenhagen Ø

Email address to gain access to the thesis: lu@di.ku.dk.
You will either receive a copy of the thesis or be informed where you can read a physical copy.
Recipients of copies of the thesis are not allowed to share or distribute it due to copyright compliance.

Short description of the thesis

Lung cancer is a leading cause of cancer-related deaths worldwide, making early diagnosis crucial. Small nodules found in lung CT scans may be harmless or cancerous. While artificial intelligence (AI) models offer promising diagnostic support, their opacity—the “black-box” problem—limits clinical acceptance.

The research investigates self-explanatory AI models that integrate explainability directly into their diagnostic processes while significantly reducing the annotation burden typically required for such explainability. Through clinically meaningful nodule features, highlighted image regions, and comparisons with similar cases, these models provide understandable assessments with competitive prediction accuracy. The open-source code supports reproducibility and further validation. Together, these advances represent a step towards more practical, transparent, and trustworthy AI as a clinical second opinion.