NSCLC histological subtype classification from CT scans using generalist 3D medical foundation models

Fatih Aksu, Fabrizia Gelardi, Arturo Chiti, Paolo Soda

Abstract: Lung cancer is one of the leading causes of cancer-related deaths worldwide, with Non-Small Cell Lung Cancer (NSCLC) accounting for approximately 85% of all cases. Accurate histological subtype classification of NSCLC is critical for personalized treatment planning; however, current methods rely heavily on invasive biopsies, which can pose risks to patients. Deep learning (DL) models have shown promising results in disease diagnosis and prognosis, but they typically require large labeled datasets for training, which can be challenging to obtain in healthcare settings. Recently, foundation models have gained significant attention in various domains, including medicine, due to their ability to generalize across tasks with limited task-specific data. In this study, we investigate the potential of foundation models for NSCLC histological subtype classification. We evaluate the performance of three generalist medical foundation models, trained on 3D CT scans, and compare them with three task-specific models designed for this classification task. Using a dataset of 714 NSCLC patients, we found that all three foundation models outperformed the task-specific models, highlighting their potential to improve classification accuracy in NSCLC histological subtype identification.

External IDs:dblp:conf/ichi/AksuGCS25