Deciphering the biological underpinnings behind prognostic MRI-based imaging signatures in breast cancer: a systematic review.
Authors
Affiliations (8)
Affiliations (8)
- Department of Radiology, The First Affiliated Hospital of Zhejiang Chinese Medical University (Zhejiang Provincial Hospital of Chinese Medicine), 54 Youdian Road, Hangzhou, 310006, P.R. China.
- The First School of Clinical Medicine, Zhejiang Chinese Medical University, Hangzhou, China.
- Department of Radiology, The First Affiliated Hospital of Zhejiang Chinese Medical University (Zhejiang Provincial Hospital of Chinese Medicine), 54 Youdian Road, Hangzhou, 310006, P.R. China. [email protected].
- The First School of Clinical Medicine, Zhejiang Chinese Medical University, Hangzhou, China. [email protected].
- Department of Radiology, The First Affiliated Hospital of Zhejiang Chinese Medical University (Zhejiang Provincial Hospital of Chinese Medicine), 54 Youdian Road, Hangzhou, 310006, P.R. China. [email protected].
- The First School of Clinical Medicine, Zhejiang Chinese Medical University, Hangzhou, China. [email protected].
- Department of Radiology, The First Affiliated Hospital of Zhejiang Chinese Medical University (Zhejiang Provincial Hospital of Chinese Medicine), 54 Youdian Road, Hangzhou, 310006, P.R. China. [email protected].
- The First School of Clinical Medicine, Zhejiang Chinese Medical University, Hangzhou, China. [email protected].
Abstract
To explore the biological foundations of MRI-based prognostic imaging signatures (including radiomics and deep learning signatures) in breast cancer, and to assess the methodological quality of existing studies. This review identified studies through a comprehensive search of PubMed, Embase, Web of Science Core Collection, and the Cochrane Library through February 25, 2025. Studies on MRI-based prognostic radiomics or deep learning models with elaborated biological relevance were included. The Radiomics Quality Score (RQS), Newcastle-Ottawa Scale (NOS), and Quality Assessment of Prognostic Accuracy Studies (QUAPAS) were employed to appraise the quality of studies. Data extraction included details on study characteristics, specifics of radiomics or deep learning models, and methods leveraged for biological analysis. Sixteen studies published from 2015 to 2025, comprising 61-2279 breast cancer patients, were included. Most studies employed supervised machine learning methods, with a few utilizing unsupervised machine learning methods. The underlying biological correlations mainly focused on genomic, tumor microenvironment-related, and multiomics data. The median RQS was 12.5 (range 5-17), and the mean NOS score was 7.3, reflecting limited methodological rigor. The overall risk of bias (ROB) among the studies was high, according to QUAPAS. The underlying biological associations of prognostic imaging signatures are mainly elucidated through genomic and transcriptomic factors. Further in-depth exploration is essential to facilitate personalized and precise treatment.