The role of radiogenomics in the diagnosis of breast cancer: a systematic review

Abstract

Background One of the most common cancers diagnosed worldwide is breast cancer (BC), which is the leading cause of cancer death among women. The radiogenomics method is more accurate for managing and inhibiting this disease, which takes individual diagnosis on genes, environments, and lifestyles of each person. The present study aims to highlight the current state-of-the-art, the current role and limitations, and future directions of radiogenomics in breast cancer. Method This systematic review article was searched from databases such as Embase, PubMed, Web of Science, Google Scholar, Scopus, and Cochrane Library without any date or language limitations of databases. Searches were performed using Boolean OR and AND operators between the main terms and keywords of particular topic of the subject under investigation. All retrospective, prospective, cohort, and pilot studies were included, which were provided with more details about the topic. Articles such as letter to the editor, review, and short communications were excluded because of lack of information, discussions, or use of radiogenomics method on other cancers. For quality assessment of articles, STROBE checklist was used. Result For the systematic review, 18 articles were approved after assessing the full text of selected articles. In this review, 3614 patients with BC of selected articles were evaluated, and all radiogenomics were associated with more power in classification, differential diagnosis, and prognosis of BC. Among the various modalities to predict genomic indicators and molecular subtypes, DCE-MRI has the higher performance and finally the highest amount of AUC value (0.956) belonged to PI3K gene. Conclusion This review shows that radiogenomics can help with the diagnosis and treatment of breast cancer in patients. It has shown that recognizing and specifying radiogenomic phenotypes in the genomic signatures can be helpful in treatment and diagnosis of disease. The molecular methods used in these articles are limited to miRNAs expression, gene expression, Ki67 proliferation index, next-generation RNA sequencing, whole RNA sequencing, and molecular histopathology that can be completed in future studies by other methods such as exosomal miRNAs, specific proteins expression, DNA repair capacity, and other biomarkers that have prognostic and predictive value for cancer treatment response. Studies with control group and large sample size for evaluation of radiogenomics in diagnosis and treatment recommended.

Application of Artificial Intelligence Methods for Imaging of Spinal Metastasis

“…To date, there are no other published studies on radiomics models for identifying other clinically important genotypes in spinal oncology imaging, such as KRAS, BRAF and ALK [ 133 ], which have been shown to have diagnostic and prognostic significance. However, there are many studies in the literature for other organs, including breast oncology [ 134 ] imaging (e.g., miRNAs expression, gene expression and Ki67 proliferation index), and for brain imaging ( Supplementary Figure S1 ), including biomarkers such as isocitrate dehydrogenase (IDH) [ 135 ], chromosome arms 1p/19q-codeletion [ 136 , 137 , 138 ] and methylguanine-DNA methyltransferase status (MGMT) [ 139 , 140 , 141 , 142 ] as prognostic markers for glioma [ 143 , 144 , 145 ]. Future work will include the study of these other clinically important genotypes to aid in deciding treatment for patients with spinal metastases or other primary tumours, supporting the new era of precision medicine [ 146 , 147 ].…”

A Focus on the Synergy of Radiomics and RNA Sequencing in Breast Cancer

“…The size and morphologic radiomics features such as texture, diameter, and perimeter, derived from DCE-MRI have been demonstrated to provide information about the metastatic capacity of BC and the composition of the microenvironment, in particular neutrophils, fibroblasts, and endothelial cells in a small cohort of patients [ 69 ]. The DCE-MRI is a high-performance modality to individuate genomic biomarkers, including cell cycle check points, genes such as Myc, PI3K, RTK/RAS, P53, and ER + /ER−, PR + /PR−, HER2 + /HER2, and triple-negative indicators [ 70 ]. Bismeijer and colleagues established that the tumor size changes in concomitant with the proliferative rate of the mass [ 71 ].…”

“…Moreover, radiogenomics, in the radiotherapy field, aims to find indicators which can predict the sensitivity of tumor and healthy tissues to radiation in order to develop a more personalized therapy [ 70 ]. PET imaging of 18F-fluorodeoxygucose (FDG) is frequently used to mark tumors in order to determine its diffusion in the body and assess the response to a treatment, although the FDG uptake is still not fully understood [ 75 ].…”

“…Systemic reviews and multiple studies on small cohort of patients are suggesting that the integration of molecular, anatomical, and functional approaches in radiogenomics, might allow us to classify significant features of tumor tissue, that could become potential biomarkers for the development of personalized medicine. Specifically, radiogenomics and/or radiomics could assist in the early detection and classification of cancers, differential diagnosis, mapping BC before surgery, correlation between imaging features and tumor molecular biomarkers, prediction of metastasis, and chemotherapy benefit prediction [ 33 , 70 , 77 , 78 , 79 , 80 ].…”

Application of Machine Learning for Differentiating Bone Malignancy on Imaging: A Systematic Review

“…Emerging Artificial Intelligence (AI) tools continue to demonstrate remarkable progress in medical imaging applications, especially in the field of oncology [ 36 ]. These applications include cancer screening and diagnosis [ 37 , 38 , 39 , 40 ], diagnosis and classification [ 41 , 42 , 43 , 44 ], predicting prognosis and treatment response [ 45 , 46 , 47 , 48 , 49 ], automated segmentation [ 50 , 51 , 52 , 53 , 54 ], and radiology-pathology correlation (radiogenomics) [ 55 , 56 , 57 , 58 ]. In particular, within the field of diagnosis and classification, the ability of AI models to classify benign vs. malignant tumours has been shown to achieve high accuracy, sensitivity, and specificity in various organs, such as in the case of breast [ 59 , 60 , 61 ], prostate [ 62 , 63 ], lung [ 38 , 64 , 65 , 66 ], and brain lesions [ 67 , 68 ].…”