Analysis of Routine Computed Tomographic Scans With Radiomics and Machine Learning: One Step Closer to Clinical Practice
Menée à l'aide de l'intelligence artificielle et d'images tomographiques réalisées sur 575 patients atteints d'un mélanome de stade avancé traité par immunothérapie, cette étude évalue la performance d'une signature radiomique pour estimer la survie globale des patients et éclairer la décision thérapeutique
Response Evaluation Criteria in Solid Tumors (RECIST) has been the standard approach to assess response and predict survival in oncology clinical trials for many years. In RECIST 1.1, the percentage change in the sum of the longest single dimension of up to 5 target lesions is used to classify response. Although many studies have supported the utility of this approach, other studies have shown that there is room for improvement. For example, volumetric changes in tumor size and radiomics signatures of image textural qualities have shown promise as tools to predict survival that are complementary to RECIST.
JAMA Oncology , commentaire, 2021