Journal: NPJ digital medicine
This publication provides a clinically focused review of computer-aided diagnosis in lung cancer, covering traditional imaging, machine learning, and deep learning approaches.
It highlights significant bedside-validated advancements, including:
- Multimodal CT/PET and clinical data fusion
- Strategies tailored for small datasets
- Interpretable AI models
- Privacy-preserving multi-center learning frameworks
The reviewed systems achieve high diagnostic performance (AUC ≥ 0.95 with less than 0.1 false positives per CT) and improve early detection rates by approximately 20–30%, with prognostic accuracy reflected by C-index values around 0.85–0.90.
The authors also discuss critical implementation steps and priorities to translate these technical accuracies into meaningful patient outcomes.