Role of textural and radiomic analysis parameters in predicting histopathological parameters of the tumor in breast cancer patients.

Abstract

Introduction: Texture and radiomic analysis characterizes the tumor's phenotype and evaluates its microenvironment in quantitative terms. This study aims to investigate the role of textural and radiomic analysis parameters in predicting histopathological factors in breast cancer patients.

Materials and methods: Two hundred and twelve primary breast cancer patients underwent 18F-FDG PET/computed tomography for staging. The images were processed in a commercially available textural analysis software. ROI was drawn over the primary tumor with a 40% threshold and was processed further to derive textural and radiomic parameters. These parameters were then compared with histopathological factors of tumor. Receiver-operating characteristic analysis was performed with a P-value <0.05 for statistical significance. The significant parameters were subsequently utilized in various machine learning models to assess their predictive accuracy.

Results: A retrospective study of 212 primary breast cancer patients was done. Among all the significant parameters, SUVmin, SUVmean, SUVstd, SUVmax, discretized HISTO_Entropy, and gray level co-occurrence matrix_Contrast were found to be significantly associated with ductal carcinoma type. Four parameters (SUVmin, SUVmean, SUVstd, and SUVmax) were significant in differentiating the luminal subtypes of the tumor. Five parameters (SUVmin, SUVmean, SUVstd, SUVmax, and SUV kurtosis) were significant in predicting the grade of the tumor. These parameters showcased robust capabilities in predicting multiple histopathological parameters when tested using machine learning algorithms.

Conclusion: Though textural analysis could not predict hormonal receptor status, lymphovascular invasion status, perineural invasion status, microcalcification status of tumor, and all the molecular subtypes of the tumor, it could predict the tumor's histologic type, triple-negative subtype, and score of the tumor noninvasively.