Predicting Gleason Score of Prostate Cancer Patients Using Radiomic Analysis

Abstract

Purpose: Use of quantitative imaging features and encoding the intra-tumoral heterogeneity from multi-parametric magnetic resonance imaging (mpMRI) for the prediction of Gleason score is gaining attention as a non-invasive biomarker for prostate cancer (PCa). This study tested the hypothesis that radiomic features, extracted from mpMRI, could predict the Gleason score pattern of patients with PCa. Methods: This analysis included T2-weighted (T2-WI) and apparent diffusion coefficient (ADC, computed from diffusion-weighted imaging) scans of 99 PCa patients from The Cancer Imaging Archive (TCIA). A total of 41 radiomic features were calculated from a local tumor sub-volume (i.e., regions of interest) that is determined by a centroid coordinate of PCa volume, grouped based on their Gleason score patterns. Kruskal-Wallis and Spearman's rank correlation tests were used to identify features related to Gleason score groups. Random forest (RF) classifier model was used to predict Gleason score groups and identify the most important signature among the 41 radiomic features. Results: Gleason score groups could be discriminated based on zone size percentage, large zone size emphasis and zone size non-uniformity values (p < 0.05). These features also showed a significant correlation between radiomic features and Gleason score groups with a correlation value of -0.35, 0.32, 0.42 for the large zone size emphasis, zone size non-uniformity and zone size percentage, respectively (corrected p < 0.05). RF classifier model achieved an average of the area under the curves of the receiver operating characteristic (ROC) of 83.40, 72.71, and 77.35% to predict Gleason score groups (G1) = 6; 6 < (G2) < (3 + 4) and (G3) ≥ 4 + 3, respectively. Conclusion: Our results suggest that the radiomic features can be used as a non-invasive biomarker to predict the Gleason score of the PCa patients.

Keywords: biomarkers; classification; gleason score; prostate cancer; radiomics.

Figure 1

Schema of a radiomic model for patients with PCa. Acquisition of pre-treatment PCa patient's MR images; Regions of interest (i.e., subvolume 21 × 21 × 3 voxels); Extraction of 41 radiomic features from ROIs; Feature significance analysis based on Spearman rank correlation and Kruskal-Wallis, and multivariate prediction of Gleason score groups using the random forest model.

Figure 2

(A) Heatmap of Kruskal-Wallis significance test p-values (–log₁₀ scale) using radiomic features to identify patients of different Gleason scores. Significant features to compare between Gleason score groups are indicated with a black-green circle (corrected p < 0.05). (B) Spearman rank correlation between feature value and groups of Gleason score (i.e., 1, 2, 3), color-coded from minimum (dark blue) to maximum (dark red). Features with statistically significant correlation (i.e., corrected p < 0.05) are indicated with a black-green circle.

Figure 3

ROC curves and AUC obtained by the random forest (RF) models for predicting Gleason score of PCa patients using the radiomic features. Gleason score groups: G1 (group 1), G2 (group 2), G3 (group 3). (A) 5-fold cross validation, (B) Validation of the trained (n = 40; balanced classes) RF model by testing new datasets (n = 20).

Figure 4

Importance of features for predicting the Gleason score of PCa patients, G1 vs. G2-G3 (A), G2 vs. G1-G3 (B), G3 vs. G1-G2 (C). Reported values correspond to the average increase in prediction error obtained by permuting the values of individual features across out-of-bag observations (46). Green and red bars represent the positive and negative impact for predicting the Gleason score groups.

Figure 5

ROC curves and AUC obtained by RF models for predicting Gleason Score of PCa patients using the features derived from ADC (A) and T2-WI (B) images. (C) Heatmap of importance value for features predicting the three Gleason score groups of PCa patients. Color-coded from minimum (dark blue) to maximum (dark red). Features with statistically significant correlation (i.e., corrected p < 0.05) are indicated with a black-green circle. Features with predictive value (importance >0) are indicated with a black-green circle.