A self-tuned graph-based framework for localization and grading prostate cancer lesions: An initial evaluation based on multiparametric magnetic resonance imaging

摘要: Multiparametric magnetic resonance imaging (mpMRI) has been established as the state-of-the-art examination for detection and localization of prostate cancer lesions. Prostate Imaging-Reporting Data System (PI-RADS) a scheme to standardize reporting mpMRI findings. Although lesion delineation PI-RADS ratings could be performed manually, human are subjective time-consuming. In this article, we developed validated self-tuned graph-based model rating prediction. 34 features were obtained at pixel level from T2-weighted (T2W), apparent diffusion coefficient (ADC) dynamic contrast enhanced (DCE) images, which scores predicted. Two major innovations involved in model. First, approaches sensitive choice edge weight. The proposed tuned weights automatically based on structure data, thereby obviating empirical weight selection. Second, feature give heavier important estimation. framework was evaluated its performance datasets 12 patients. evaluation, score distribution map generated by algorithm observers' binarized thresholds 3 4. sensitivity, specificity accuracy these two threshold settings ranged 65 77%, 86 93% 85 88% respectively, comparable results previous studies non-clinical T2 maps available. took 10s estimate an axial image. efficiency achievable suggests that technique can into MR analysis system suitable clinical use after thorough validation involving more