Predicting early response to ablative radiotherapy in oligometastatic disease: a scoping review of radiomics-based machine learning and deep learning models.
Authors
Affiliations (6)
Affiliations (6)
- Department of Radiation Oncology, Hospital Universitari de Sant Joan, Institut d'Investigació Sanitària Pere Virgili, Universitat Rovira i Virgili, Reus, Spain.
- Unitat de Recerca Biomèdica, Hospital Universitari Sant Joan de Reus, Institut d'Investigació Sanitària Pere Virgili, Universitat Rovira i Virgili, Reus, Spain.
- Medical Image Analysis and Biometry Laboratory, Universidad Rey Juan Carlos, Móstoles, Madrid, Spain.
- Organisation for Economic Co-operation and Development (OECD) Health Division, Paris, France.
- Cochrane Ecuador, Centro de Investigación en Salud Pública y Epidemiología Clínica (CISPEC), Facultad de Ciencias de la Salud Eugenio Espejo, Universidad UTE, Quito, Ecuador.
- Unitat de Recerca Biomèdica, Hospital Universitari Sant Joan de Reus, Institut d'Investigació Sanitària Pere Virgili, Universitat Rovira i Virgili, Reus, Spain. [email protected].
Abstract
Oligometastatic disease represents an intermediate stage of cancer, often treated with surgery or ablative radiotherapy (ART). This scoping review aimed to systematically summarize current evidence on the use of radiomics, including machine learning and deep learning approaches, to predict response to ART. We also aimed to assess the methodological quality and reporting transparency of published studies, identifying gaps and opportunities for future research. A systematic search in PubMed, Web of Science, Scopus, Embase, Cochrane, and Google Scholar identified studies that used radiomics for predicting ART response. Two reviewers independently selected and assessed the methodological quality using the Radiomics Quality Score (RQS) and the METhodological RadiomICs Score (METRICS). In addition, reporting transparency was evaluated using the CheckList for EvaluAtion of Radiomics research (CLEAR). This scoping review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) extension for Scoping Reviews guidelines. The systematic search identified 9463 records, of which 29 studies (3946 patients) were included. Most studies used MRI-derived features, with 24 focusing on brain metastases. Radiomics-based models demonstrated variable predictive performance (area under the curve, AUC: 0.69-0.95), with deep learning models achieving the highest accuracies (AUC: 0.85-1.00). Methodological quality of the studies was moderate (mean RQS: 13; METRICS: 64.2-78%). Radiomics-based models show potential for identifying patients unlikely to benefit from ART, but their clinical implementation remains limited, especially for extracranial metastases. Future research should focus on multicenter, prospective studies with standardized protocols, incorporating clinical and dosimetric data for broader clinical application. Question Can radiomics-based predictive models reliably assess treatment response to ablative radiotherapy in oligometastatic disease, and how robust is the current methodological evidence supporting their use? Findings Radiomics models show encouraging predictive performance, mainly for brain metastases, yet substantial methodological heterogeneity and limited validation hinder their clinical translation. Clinical relevance Radiomics-based prediction models hold potential for identifying patients unlikely to benefit from ablative radiotherapy, enabling more personalized treatment. Further prospective, multicenter, and methodologically standardized studies are essential before clinical implementation.