Prediction of Prostate Cancer Disease Aggressiveness Using Bi-Parametric Mri Radiomics

Abstract

Prostate cancer is one of the most prevalent cancers in the male population. Its diagnosis and classification rely on unspecific measures such as PSA levels and DRE, followed by biopsy, where an aggressiveness level is assigned in the form of Gleason Score. Efforts have been made in the past to use radiomics coupled with machine learning to predict prostate cancer aggressiveness from clinical images, showing promising results. Thus, the main goal of this work was to develop supervised machine learning models exploiting radiomic features extracted from bpMRI examinations, to predict biological aggressiveness; 288 classifiers were developed, corresponding to different combinations of pipeline aspects, namely, type of input data, sampling strategy, feature selection method and machine learning algorithm. On a cohort of 281 lesions from 183 patients, it was found that (1) radiomic features extracted from the lesion volume of interest were less stable to segmentation than the equivalent extraction from the whole gland volume of interest; and (2) radiomic features extracted from the whole gland volume of interest produced higher performance and less overfitted classifiers than radiomic features extracted from the lesions volumes of interest. This result suggests that the areas surrounding the tumour lesions offer relevant information regarding the Gleason Score that is ultimately attributed to that lesion.

Beyond Multiparametric MRI and towards Radiomics to Detect Prostate Cancer: A Machine Learning Model to Predict Clinically Significant Lesions

“…Recently, radiomics has emerged as the application of well-established machine learning (ML) and artificial intelligence techniques to medical image analysis [ 9 , 10 , 11 ]. The large popularity gained by radiomics arises from the possibility of ML algorithms to considerably enrich the information retrievable from imaging, represented by latent radiomic features, which ultimately allow improvements in the diagnostic potential of mpMRI [ 12 , 13 ].…”

The Use of MRI-Derived Radiomic Models in Prostate Cancer Risk Stratification: A Critical Review of Contemporary Literature

“…Rather, we focused on stratification of clinically significant PCa by GS risk groups [ 15 , 16 , 17 ]. Table 1 summarizes the 15 studies utilizing MRI-derived radiomic models to predict pathologic GS [ 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 ], with one study predicting GS upgrading between biopsy and surgery [ 23 ].…”

“…In delineating regions of interest, eight studies [ 15 , 17 , 18 , 19 , 22 , 24 , 25 , 27 ] utilized only PI-RADS lesions, while the remaining seven utilized the full prostate [ 16 , 20 , 21 , 23 , 26 , 28 , 29 ]. Feature extraction and selection was most commonly done via minimum redundancy maximum relevance (mRMR) [ 21 , 28 , 29 ] or random forest (RF) [ 15 , 16 , 24 ], and the receiver operator characteristic–area under the curve (ROC-AUC) of the final models ranged from 0.50 to 0.92. Castillo and colleagues [ 27 ] were the only group to perform a multi-institutional external validation of their model for the prediction of GS ≤ 6 versus GS ≥ 7, yielding an ROC-AUC of 0.75.…”

The Potential and Emerging Role of Quantitative Imaging Biomarkers for Cancer Characterization

“…Besides traditional analyses, it was demonstrated in patients with colorectal cancer that readily available biomarkers, such as CT attenuation/Hounsfield intensity values, could also be predictors of survival [ 20 ]. In line with these results, it was possible to use radiomics for aggressiveness prediction in papillary thyroid cancer using multiparametric MRI [ 90 ], in prostate cancer using bi-parametric MRI [ 91 ], and in diffuse lower-grade gliomas using diffusion- and perfusion-weighted sequences [ 92 ]. For prostate cancer, it was shown that differentiation and aggressiveness assessment using a radiomics-based model outperforms the traditional PI-RADS visual assessment.…”