De Novo Radiomics Approach Using Image Augmentation and Features From T1 Mapping to Predict Gleason Scores in Prostate Cancer

Abstract

Objectives: The aims of this study were to discriminate among prostate cancers (PCa's) with Gleason scores 6, 7, and ≥8 on biparametric magnetic resonance imaging (bpMRI) of the prostate using radiomics and to evaluate the added value of image augmentation and quantitative T1 mapping. Materials and Methods: Eighty-five patients with subsequently histologically proven PCa underwent bpMRI at 3 T (T2-weighted imaging, diffusion-weighted imaging) with 66 patients undergoing additional T1 mapping at 3 T. The PCa lesions as well as the peripheral and transition zones were segmented pixel by pixel in multiple slices of the 3D MRI data sets (T2-weighted images, apparent diffusion coefficient, and T1 maps). To increase the size of the data set, images were augmented for contrast, brightness, noise, and perspective multiple times, effectively increasing the sample size 10-fold, and 322 different radiomics features were extracted before and after augmentation. Four different machine learning algorithms, including a random forest (RF), stochastic gradient boosting (SGB), support vector machine (SVM), and k-nearest neighbor, were trained with and without features from T1 maps to differentiate among 3 different Gleason groups (6, 7, and ≥8). Results: Support vector machine showed the highest accuracy of 0.92 (95% confidence interval [CI], 0.62-1.00) for classifying the different Gleason scores, followed by RF (0.83; 95% CI, 0.52-0.98), SGB (0.75; 95% CI, 0.43-0.95), and k-nearest neighbor (0.50; 95% CI, 0.21-0.79). Image augmentation resulted in an average increase in accuracy between 0.08 (SGB) and 0.48 (SVM). Removing T1 mapping features led to a decline in accuracy for RF (−0.16) and SGB (−0.25) and a higher generalization error. Conclusions: When data are limited, image augmentations and features from quantitative T1 mapping sequences might help to achieve higher accuracy and lower generalization error for classification among different Gleason groups in bpMRI by using radiomics.

Influence of Image Processing on Radiomic Features From Magnetic Resonance Imaging

“…To remain feasible, many other sequences, including 3D isotropic sequences and quantitative sequences such as T1- and T2-mapping and diffusion-weighted MRI, were not benchmarked, important vendor differences/scanner-specific image processing parameters were neither assessed nor compared, and in vivo experiments were not conducted 47,48 . These are major limitations that should be addressed in future multiscanner, multicenter studies.…”

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.…”

Imaging in translational cancer research

“…In prostate cancer, T2 mapping has revealed significantly lower T2 values in cancerous regions 103 , and has shown high diagnostic accuracy in differentiating between chronic prostatitis and cancer 104 . In contrast, T1 mapping has been found to predict Gleason scores 96 . In liver cancer, multiparametric MRI (T1, T2 relaxation times) has been found to be useful in detecting sinusoidal obstruction syndrome after oxaliplatin-based chemotherapy 105 .…”

“…Their use beyond cardiac imaging has increased, owing to better reconstruction and k-space sampling techniques, e.g., for mapping peritumoral infiltration zones in glioblastoma and anaplastic astrocytoma with T2 mapping 94 , or to identify histological types of lung cancer with T1 mapping 95 . T1 mapping has also been used to predict Gleason scores in prostate cancer 96 , whereas T2 mapping in prostate cancer has been used to differentiate between healthy and cancerous gland tissue 97 .…”