Development of an AI-based predictive model for surgical outcomes in lung cancer treatment
Seance of wednesday 16 april 2025 (Intelligence Artificielle et Chirurgie)
DOI number : 10.26299/z8fd-ca87/2025.15.05
Abstract
The BRC-Lung project, submitted under the European Innovation Health Initiative call, aims to develop an artificial intelligence (AI) model to predict surgical outcomes in thoracic surgery, particularly in the management of lung tumors. It is based on a public-private partnership, with the project being co-developed by hospital, academic, and industry stakeholders.
The initial analysis highlights a high variability in surgical practices (lobectomy vs. segmentectomy) and clinical outcomes (complications, length of stay, readmission rates) across hospitals. One of the main hypotheses is that the quality of surgical planning may play a key role in this variability.
Recent studies show that the use of 3D pulmonary reconstruction significantly reduces intraoperative conversions, postoperative complications, and length of hospital stay. The project aims to go a step further by integrating this data into an AI-based predictive model.
This model would be used early in the care pathway, during multidisciplinary team meetings (MDTs), to help choose the most appropriate surgical option and predict outcomes such as: risk of conversion, hospital length of stay, complications, and functional recovery (via PROMs/PREMs).
BRC-Lung aims to reduce false positives by adapting surgical strategies based on 3D imaging and AI scores, such as switching from lobectomy to segmentectomy, thereby achieving significant benefits in lung preservation and complication reduction.
The development will rely on data sources such as EPITHOR, platforms like Synapse 3D, natural language processing (NLP) tools to extract information from medical records, and a structured data repository. The project is planned in several phases over five years: retrospective study, model development, prospective validation, and industrialization, including CE marking, business strategy, and scientific publications.
The ambition is to create a European reference standard of excellence and to standardize practices through an AI-powered tool embedded in clinical decision-making for thoracic surgery.
The initial analysis highlights a high variability in surgical practices (lobectomy vs. segmentectomy) and clinical outcomes (complications, length of stay, readmission rates) across hospitals. One of the main hypotheses is that the quality of surgical planning may play a key role in this variability.
Recent studies show that the use of 3D pulmonary reconstruction significantly reduces intraoperative conversions, postoperative complications, and length of hospital stay. The project aims to go a step further by integrating this data into an AI-based predictive model.
This model would be used early in the care pathway, during multidisciplinary team meetings (MDTs), to help choose the most appropriate surgical option and predict outcomes such as: risk of conversion, hospital length of stay, complications, and functional recovery (via PROMs/PREMs).
BRC-Lung aims to reduce false positives by adapting surgical strategies based on 3D imaging and AI scores, such as switching from lobectomy to segmentectomy, thereby achieving significant benefits in lung preservation and complication reduction.
The development will rely on data sources such as EPITHOR, platforms like Synapse 3D, natural language processing (NLP) tools to extract information from medical records, and a structured data repository. The project is planned in several phases over five years: retrospective study, model development, prospective validation, and industrialization, including CE marking, business strategy, and scientific publications.
The ambition is to create a European reference standard of excellence and to standardize practices through an AI-powered tool embedded in clinical decision-making for thoracic surgery.