A Quality Control System for Automated Prostate Segmentation on T2-Weighted MRI

Abstract

Computer-aided detection and diagnosis (CAD) systems have the potential to improve robustness and efficiency compared to traditional radiological reading of magnetic resonance imaging (MRI). Fully automated segmentation of the prostate is a crucial step of CAD for prostate cancer, but visual inspection is still required to detect poorly segmented cases. The aim of this work was therefore to establish a fully automated quality control (QC) system for prostate segmentation based on T2-weighted MRI. Four different deep learning-based segmentation methods were used to segment the prostate for 585 patients. First order, shape and textural radiomics features were extracted from the segmented prostate masks. A reference quality score (QS) was calculated for each automated segmentation in comparison to a manual segmentation. A least absolute shrinkage and selection operator (LASSO) was trained and optimized on a randomly assigned training dataset (N = 1756, 439 cases from each segmentation method) to build a generalizable linear regression model based on the radiomics features that best estimated the reference QS. Subsequently, the model was used to estimate the QSs for an independent testing dataset (N = 584, 146 cases from each segmentation method). The mean ± standard deviation absolute error between the estimated and reference QSs was 5.47 ± 6.33 on a scale from 0 to 100. In addition, we found a strong correlation between the estimated and reference QSs (rho = 0.70). In conclusion, we developed an automated QC system that may be helpful for evaluating the quality of automated prostate segmentations.

Prostate Cancer Radiogenomics—From Imaging to Molecular Characterization

“…These models were already studied in a variety of cancers [ 35 , 36 , 37 , 38 , 39 ]. The recent rise in artificial intelligence (AI) and machine learning (ML) algorithms has introduced new classifications for PCa, regarding the differentiation of favorable from unfavorable disease [ 40 , 41 ]; the quantitative assessment of information predicting the tumor Gleason score [ 31 , 32 , 42 , 43 , 44 ] and biochemical recurrence (BCR)-free survival [ 45 ]; the identification of tumors through mpMRI [ 43 , 46 ]; the development of new detection features, such as advanced zoomed diffusion-weighted imaging (DWI) and conventional full-field-of-view DWI [ 47 ]; texture analysis of prostate MRI in the prostate imaging reporting and data system (PIRADS) for PI-RADS 3 score lesions [ 48 ]; the creation of frameworks for automated PCa localization and detection [ 49 ]; and, finally, the management of radiotherapy treatment and toxicity [ 50 , 51 , 52 , 53 , 54 , 55 , 56 ], and the prediction of BCR [ 57 , 58 , 59 , 60 , 61 , 62 , 63 , 64 , 65 ]. Additionally, radiomics and AI algorithms will help to limit the discrepancies between different readers [ 66 ].…”

“…Articles that studied radiomics in PCa are just briefly reminded and not analyzed in detail, and we had them incorporated in a table, along with their clinical outcomes and results. In PCa, the use of radiomics aids prostate volume selection and segmentation [ 30 , 40 , 46 , 73 , 74 , 75 , 76 ], PCa screening [ 28 , 77 , 78 ], detection and classification [ 29 , 77 , 79 , 80 , 81 ], in addition to its role in risk stratification [ 61 , 76 , 82 , 83 ], treatment [ 59 , 75 , 78 , 84 , 85 , 86 ], and prognosis. One of the first studies that analyzed the imaging features for PCA was performed by Khalvati et al [ 87 ], with the goal of creating a radiomics-based auto detection method utilizing an mpMRI feature model that combined computed high b-value DWI (diffusion-weighted imaging) and correlated diffusion imaging, which was then evaluated through a support vector machine (SVM) classifier.…”

The Reproducibility of Deep Learning-Based Segmentation of the Prostate Gland and Zones on T2-Weighted MR Images

“…In addition, the reproducibility results were compared to the corresponding results where (1) the post-processing step was excluded, and (2) patients with a poor segmentation quality score were excluded. To enable the last comparison, our previously proposed automated segmentation quality control system (SQCS) [ 33 ] was implemented, and the patients with a quality score of less than 85 for scan 1 or/and scan 2 were excluded from further analysis. As per [ 33 ], the SQCS was implemented using pre-processed T2W images and WP segmentations.…”

“…To enable the last comparison, our previously proposed automated segmentation quality control system (SQCS) [ 33 ] was implemented, and the patients with a quality score of less than 85 for scan 1 or/and scan 2 were excluded from further analysis. As per [ 33 ], the SQCS was implemented using pre-processed T2W images and WP segmentations.…”

Magnetic Resonance Imaging Based Radiomic Models of Prostate Cancer: A Narrative Review

“…To date, there has been a focus on conducting proof of concept studies. Radiomic models have been used to discriminate low from higher-grade PCa [ 38 , 39 ], directly predict the GS [ 23 , 24 , 40 , 41 ], lesion identification [ 42 , 43 ], and plan radiotherapy [ 44 , 45 , 46 ]. More recently, radiomic models have been utilized to predict genetic characteristics, a field known as radiogenomics.…”