MR Radiomics in the Risk Stratification of Prostate Cancer
Sources of Funding: None
Introduction
The current paradigm in prostate cancer risk stratification, including DRE, PSA values, and prostate biopsy, has resulted in overdiagnosis and overtreatment. A noninvasive marker is needed to more accurately differentiate between aggressive and indolent disease. This study evaluated multiparametric magnetic resonance imaging (mpMRI)-derived texture metrics as a biomarker for prostate cancer risk stratification.
Methods
In this IRB approved, retrospective study, we identified 66 prostate cancer lesions in patients who underwent 3T mpMRI prior to prostate biopsy. Biopsy proven Prostate cancer lesions were divided into high, intermediate, and low risk categories per National Comprehensive Cancer Network guidelines. Lesion regions of interest were manually segmented from apparent diffusion coefficient (ADC) and T2 weighted images (T2WI). Texture analysis was performed using gray-level co-occurrence matrices (GLCM), fast Fourier transfer-based spectral metrics, and ADC and T2 signal intensity. Kruskall Wallis test and analysis of variance were used to determine if there is an association between texture metrics and prostate cancer risk categories. Stepwise logistic regression was used to select the best predictors in discriminating high risk lesions from other lesions.
Results
Of the spectral metrics, Complexity Index on ADC and T2WI was significantly different (p<0.01) between the risk categories. ADC-derived GLCM metrics variance, contrast, homogeneity, dissimilarity, and difference of average were significantly different (p<0.01) between the risk categories. Of the texture metrics, GLCM Variance on ADC (ADC_Var) and Information Measures of Correlation 1 on T2WI (T2_ICM1) were the best metrics in discriminating high risk lesions from intermediate and low risk lesions and were selected in the final prediction model. Used alone, the areas under the receiver operator curve (AUC) for ADC_Var and T2_IMC1 were 0.77 (95%CI: 0.64-0.9) and 0.71 (95%CI: 0.59-0.82) respectively. The AUC when using both metrics together was 0.83 (95%CI: 0.72-0.94).
Conclusions
mpMRI-based texture analysis can differentiate high risk prostate cancer lesions from intermediate and low risk lesions, demonstrating promise as a biomarker for prostate cancer risk stratification._x000D_ _x000D_
Funding
None
Bino Varghese
Darryl Hwang
Steve Cen
Mihir Desai
Suzanne Palmer
Monish Aron
Manju Aron
Inderbir Gill
Gangning Liang
Andre Abreu
Sameer Chopra
Osamu Ukimura
Vinay Duddalwar