Clinical Artificial Intelligence in Non-Small Cell Lung Cancer: From Early Detection and Computational Pathology to Precision Oncology
- 1Department of Oncology and Hemato-Oncology, University of Milan, 20122 Milan, Italy
- 2Department of Thoracic Surgery, IEO, European Institute of Oncology IRCCS, 20141 Milan, Italy
- 3Division of Interventional Pulmonology, IEO, European Institute of Oncology IRCCS, 20141 Milan, Italy
Abstract
Artificial intelligence (AI) is being developed to analyze radiological, histopathological, molecular, and clinical data across the non-small cell lung cancer (NSCLC) pathway. This narrative review examines representative applications through August 2026, focusing on intended clinical use, validation, and the evidence required to influence care. Studies were selected purposively for clinical relevance and methodological importance. Imaging models support investigation of nodule detection, malignancy estimation, and future lung cancer risk, although predictive performance does not establish the benefit of changing screening or diagnostic pathways. In computational pathology, models classify tissue, quantify morphological features, and estimate selected molecular alterations. Foundation models extend these capabilities, but their evaluation requires attention to pretraining provenance and independence of downstream testing. Histology-based molecular prediction does not currently justify replacing validated molecular assays; prioritization of testing requires prospective workflow evaluation. Models evaluated in immunotherapy-treated cohorts can predict response or survival, but these associations do not establish differential benefit between treatments. The principal limitations include selected retrospective populations, incomplete assessment of calibration and clinically relevant thresholds, uncertain transportability, and limited evaluation of model-guided care. Evidence should therefore be interpreted at the level of the individual system and intended decision. Prospective evaluation, appropriate comparators, explicit failure procedures, and post-deployment monitoring are required before technical performance can be translated into clinical benefit.
Related
- Editorial: Artificial intelligence and anti-cancer drugs: advancing discovery, personalization, and precision oncology
- Toward Precision Colorectal Cancer Screening: Integrating Clinical Risk Stratification, Molecular Biomarkers, and Artificial Intelligence
- How could artificial intelligence aid clinicians caring for preterm neonates with respiratory pathology?
- Current Advances in Spatial Pathology of Nasal Polyps: Challenges, Opportunities for Artificial Intelligence, and Future Directions
- Artificial Intelligence, Multi-Omics, and Multimodal Integration in Precision Radiotherapy for Head and Neck Cancer: From Biological Insights to Treatment Decisions
BibTeX
@article{bertolaccini2026clinical,
title = {{Clinical Artificial Intelligence in Non-Small Cell Lung Cancer: From Early Detection and Computational Pathology to Precision Oncology}},
author = {Luca Bertolaccini and Matteo Chiari and Sara Degiovanni and Antonio Mazzella and Giorgio Lo Iacono and Juliana Guarize and Monica Casiraghi and Lorenzo Spaggiari},
journal = {Cancers},
year = {2026},
month = oct,
doi = {10.3390/cancers18213415},
url = {https://www.mdpi.com/2072-6694/18/21/3415},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Clinical Artificial Intelligence in Non-Small Cell Lung Cancer: From Early Detection and Computational Pathology to Precision Oncology
AU - Luca Bertolaccini
AU - Matteo Chiari
AU - Sara Degiovanni
AU - Antonio Mazzella
AU - Giorgio Lo Iacono
AU - Juliana Guarize
AU - Monica Casiraghi
AU - Lorenzo Spaggiari
PY - 2026
DA - 2026/10/22
JO - Cancers
DO - 10.3390/cancers18213415
UR - https://www.mdpi.com/2072-6694/18/21/3415
AB - Artificial intelligence (AI) is being developed to analyze radiological, histopathological, molecular, and clinical data across the non-small cell lung cancer (NSCLC) pathway. This narrative review examines representative applications through August 2026, focusing on intended clinical use, validation, and the evidence required to influence care. Studies were selected purposively for clinical relevance and methodological importance. Imaging models support investigation of nodule detection, malignancy estimation, and future lung cancer risk, although predictive performance does not establish the benefit of changing screening or diagnostic pathways. In computational pathology, models classify tissue, quantify morphological features, and estimate selected molecular alterations. Foundation models extend these capabilities, but their evaluation requires attention to pretraining provenance and independence of downstream testing. Histology-based molecular prediction does not currently justify replacing validated molecular assays; prioritization of testing requires prospective workflow evaluation. Models evaluated in immunotherapy-treated cohorts can predict response or survival, but these associations do not establish differential benefit between treatments. The principal limitations include selected retrospective populations, incomplete assessment of calibration and clinically relevant thresholds, uncertain transportability, and limited evaluation of model-guided care. Evidence should therefore be interpreted at the level of the individual system and intended decision. Prospective evaluation, appropriate comparators, explicit failure procedures, and post-deployment monitoring are required before technical performance can be translated into clinical benefit.
SN - 2072-6694
ER -