Manual prostate cancer segmentation in MRI: interreader agreement and volumetric correlation with transperineal template core needle biopsy

Deep Learning Prostate MRI Segmentation Accuracy and Robustness: A Systematic Review

Purpose To investigate the accuracy and robustness of prostate segmentation using deep learning across various training data sizes, MRI vendors, prostate zones, and testing methods relative to fellowship-trained diagnostic radiologists. Materials and Methods In this systematic review, Embase, PubMed, Scopus, and Web of Science databases were queried for English-language articles using keywords and related terms for prostate MRI segmentation and deep learning algorithms dated to July 31, 2022. A total of 691 articles from the search query were collected and subsequently filtered to 48 on the basis of predefined inclusion and exclusion criteria. Multiple characteristics were extracted from selected studies, such as deep learning algorithm performance, MRI vendor, and training dataset features. The primary outcome was comparison of mean Dice similarity coefficient (DSC) for prostate segmentation for deep learning algorithms versus diagnostic radiologists. Results Forty-eight studies were included. Most published deep learning algorithms for whole prostate gland segmentation (39 of 42 [93%]) had a DSC at or above expert level (DSC ≥ 0.86). The mean DSC was 0.79 ± 0.06 (SD) for peripheral zone, 0.87 ± 0.05 for transition zone, and 0.90 ± 0.04 for whole prostate gland segmentation. For selected studies that used one major MRI vendor, the mean DSCs of each were as follows: General Electric (three of 48 studies), 0.92 ± 0.03; Philips (four of 48 studies), 0.92 ± 0.02; and Siemens (six of 48 studies), 0.91 ± 0.03. Conclusion Deep learning algorithms for prostate MRI segmentation demonstrated accuracy similar to that of expert radiologists despite varying parameters; therefore, future research should shift toward evaluating segmentation robustness and patient outcomes across diverse clinical settings. Keywords: MRI, Genital/Reproductive, Prostate Segmentation, Deep Learning Systematic review registration link: osf.io/nxaev © RSNA, 2024.

Clinical Use of Hematoma Volume Based On Automated Segmentation of Chronic Subdural Hematoma Using 3D U-Net

PURPOSE

To propose a method for calculating hematoma volume based on automatic segmentation of chronic subdural hematoma (CSDH) using 3D U‑net and investigate whether it can be used clinically to predict recurrence.

METHODS

Hematoma volumes manually measured from pre- and postoperative computed tomography (CT) images were used as ground truth data to train 3D U‑net in 200 patients (400 CT scans). A total of 215 patients (430 CT scans) were used as test data to output segmentation results from the trained 3D U‑net model. The similarity with the ground truth data was evaluated using Dice scores for pre and postoperative separately. The recurrence prediction accuracy was evaluated by obtaining receiver operating characteristic (ROC) curves for the segmentation results. Using a typical mobile PC, the computation time per case was measured and the average time was calculated.

RESULTS

The median Dice score of the test data were preoperative hematoma volume (Pre-HV): 0.764 and postoperative subdural cavity volume (Post-SCV): 0.741. In ROC analyses assessing recurrence prediction, the area under the curve (AUC) of the manual was 0.755 in Pre-HV, whereas the 3D U‑net was 0.735. In Post-SCV, the manual AUC was 0.779; the 3D U‑net was 0.736. No significant differences were found between manual and 3D U‑net for all results. Using a mobile PC, the average time taken to output the test data results was 30 s per case.

CONCLUSION

The proposed method is a simple, accurate, and clinically applicable; it can contribute to the widespread use of recurrence prediction scoring systems for CSDH.

Analysis of domain shift in whole prostate gland, zonal and lesions segmentation and detection, using multicentric retrospective data

Despite being one of the most prevalent forms of cancer, prostate cancer (PCa) shows a significantly high survival rate, provided there is timely detection and treatment. Computational methods can help make this detection process considerably faster and more robust. However, some modern machine-learning approaches require accurate segmentation of the prostate gland and the index lesion. Since performing manual segmentations is a very time-consuming task, and highly prone to inter-observer variability, there is a need to develop robust semi-automatic segmentation models. In this work, we leverage the large and highly diverse ProstateNet dataset, which includes 638 whole gland and 461 lesion segmentation masks, from 3 different scanner manufacturers provided by 14 institutions, in addition to other 3 independent public datasets, to train accurate and robust segmentation models for the whole prostate gland, zones and lesions. We show that models trained on large amounts of diverse data are better at generalizing to data from other institutions and obtained with other manufacturers, outperforming models trained on single-institution single-manufacturer datasets in all segmentation tasks. Furthermore, we show that lesion segmentation models trained on ProstateNet can be reliably used as lesion detection models.

Index lesion contouring on prostate MRI for targeted MRI/US fusion biopsy - Evaluation of mismatch between radiologists and urologists

PURPOSE

Mistargeting of focal lesions due to inaccurate segmentations can lead to false-negative findings on MRI-guided targeted biopsies. The purpose of this retrospective study was to examine inter-reader agreement of prostate index lesion segmentations from actual biopsy data between urologists and radiologists.

METHOD

Consecutive patients undergoing transperineal MRI-targeted prostate biopsy for PI-RADS 3-5 lesions between January 2020 and December 2021 were included. Agreement between segmentations on T2w-images between urologists and radiologists was assessed with Dice similarity coefficient (DSC) and 95 % Hausdorff distance (95 % HD). Differences in similarity scores were compared using Wilcoxon test. Differences depending on lesion features (size, zonal location, PI-RADS scores, lesion distinctness) were tested with Mann-Whitney U test. Correlation with prostate signal-intensity homogeneity score (PSHS) and lesion size was tested with Spearman's rank correlation.

RESULTS

Ninety-three patients (mean age 64.9 ± 7.1y, median serum PSA 6.5 [4.33-10.00]) were included. Mean similarity scores were statistically significantly lower between urologists and radiologists compared to radiologists only (DSC 0.41 ± 0.24 vs. 0.59 ± 0.23, p < 0.01; 95 %HD 6.38 ± 5.45 mm vs. 4.47 ± 4.12 mm, p < 0.01). There was a moderate and strong positive correlation between DSC scores and lesion size for segmentations from urologists and radiologists (ρ = 0.331, p = 0.002) and radiologists only (ρ = 0.501, p < 0.001). Similarity scores were worse in lesions ≤ 10 mm while other lesion features did not significantly influence similarity scores.

CONCLUSION

There is significant mismatch of prostate index lesion segmentations between urologists and radiologists. Segmentation agreement positively correlates with lesion size. PI-RADS scores, zonal location, lesion distinctness, and PSHS show no significant impact on segmentation agreement. These findings could underpin benefits of perilesional biopsies.

Multiparametric MRI-based radiomics model to predict pelvic lymph node invasion for patients with prostate cancer

OBJECTIVE

To identify which patient with prostate cancer (PCa) could safely avoid extended pelvic lymph node dissection (ePLND) by predicting lymph node invasion (LNI), via a radiomics-based machine learning approach.

METHODS

An integrative radiomics model (IRM) was proposed to predict LNI, confirmed by the histopathologic examination, integrating radiomics features, extracted from prostatic index lesion regions on MRI images, and clinical features via SVM. The study cohort comprised 244 PCa patients with MRI and followed by radical prostatectomy (RP) and ePLND within 6 months between 2010 and 2019. The proposed IRM was trained in training/validation set and evaluated in an internal independent testing set. The model's performance was measured by area under the curve (AUC), sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV). AUCs were compared via Delong test with 95% confidence interval (CI), and the rest measurements were compared via chi-squared test or Fisher's exact test.

RESULTS

Overall, 17 (10.6%) and 14 (16.7%) patients with LNI were included in training/validation set and testing set, respectively. Shape and first-order radiomics features showed usefulness in building the IRM. The proposed IRM achieved an AUC of 0.915 (95% CI: 0.846-0.984) in the testing set, superior to pre-existing nomograms whose AUCs were from 0.698 to 0.724 (p < 0.05).

CONCLUSION

The proposed IRM could be potentially feasible to predict the risk of having LNI for patients with PCa. With the improved predictability, it could be utilized to assess which patients with PCa could safely avoid ePLND, thus reduce the number of unnecessary ePLND.

KEY POINTS

• The combination of MRI-based radiomics features with clinical information improved the prediction of lymph node invasion, compared with the model using only radiomics features or clinical features. • With improved prediction performance on predicting lymph node invasion, the number of extended pelvic lymph node dissection (ePLND) could be reduced by the proposed integrative radiomics model (IRM), compared with the existing nomograms.

Evaluation of Traumatic Subdural Hematoma Volume by Using Image Segmentation Assessment Based on Deep Learning

Rapid and accurate evaluations of hematoma volume can guide the treatment of traumatic subdural hematoma. We aim to explore the consistency between the measurement results of traumatic subdural hematoma (TSDH) using a deep learn-based image segmentation algorithm. A retrospective study was conducted on 90 CT images of patients diagnosed with TSDH in our hospital from January 2019 to January 2022. All image data were measured by manual segmentation, convolutional neural networks (CNN) algorithm segmentation, and ABC/2 volume formula. With manual segmentation as the “golden standard,” a consistency test was carried out with CNN algorithm segmentation and ABC/2 volume formula, respectively. The percentage error of CNN algorithm segmentation is less than ABC/2 volume formula. There is no significant difference between CNN algorithm segmentation and manual segmentation (P > 0.05). The area under curve of the ABC/2 volume formula, manual segmentation, and CNN algorithm segmentation is 0.811 (95% CI: 0.717~0.905), 0.840 (95% CI: 0.753~0.928), and 0.832 (95% CI: 0.742~0.922), respectively. From our results, the algorithm based on CNN has a good efficiency in segmentation and accurate calculation of TSDH hematoma volume.

Fully Automatic Whole-Volume Tumor Segmentation in Cervical Cancer

Simple Summary

Uterine cervical cancer (CC) is a leading cause of cancer-related deaths in women worldwide. Pelvic magnetic resonance imaging (MRI) allows the assessment of local tumor extent and guides the choice of primary treatment. MRI tumor segmentation enables whole-volume radiomic tumor profiling, which is potentially useful for prognostication and individualization of therapy in CC. Manual tumor segmentation is, however, labor intensive and thus not part of routine clinical workflow. In the current work, we trained a deep learning (DL) algorithm to automatically segment the primary tumor in CC patients. Although the achieved segmentation performance of the trained DL algorithm is slightly lower than that for human experts, it is still relatively good. This study suggests that automated MRI primary tumor segmentations by DL algorithms without any human interaction is possible in patients with CC.

Abstract

Uterine cervical cancer (CC) is the most common gynecologic malignancy worldwide. Whole-volume radiomic profiling from pelvic MRI may yield prognostic markers for tailoring treatment in CC. However, radiomic profiling relies on manual tumor segmentation which is unfeasible in the clinic. We present a fully automatic method for the 3D segmentation of primary CC lesions using state-of-the-art deep learning (DL) techniques. In 131 CC patients, the primary tumor was manually segmented on T2-weighted MRI by two radiologists (R1, R2). Patients were separated into a train/validation (n = 105) and a test- (n = 26) cohort. The segmentation performance of the DL algorithm compared with R1/R2 was assessed with Dice coefficients (DSCs) and Hausdorff distances (HDs) in the test cohort. The trained DL network retrieved whole-volume tumor segmentations yielding median DSCs of 0.60 and 0.58 for DL compared with R1 (DL-R1) and R2 (DL-R2), respectively, whereas DSC for R1-R2 was 0.78. Agreement for primary tumor volumes was excellent between raters (R1-R2: intraclass correlation coefficient (ICC) = 0.93), but lower for the DL algorithm and the raters (DL-R1: ICC = 0.43; DL-R2: ICC = 0.44). The developed DL algorithm enables the automated estimation of tumor size and primary CC tumor segmentation. However, segmentation agreement between raters is better than that between DL algorithm and raters.

Explainable AI for CNN-based prostate tumor segmentation in multi-parametric MRI correlated to whole mount histopathology

Automatic prostate tumor segmentation is often unable to identify the lesion even if multi-parametric MRI data is used as input, and the segmentation output is difficult to verify due to the lack of clinically established ground truth images. In this work we use an explainable deep learning model to interpret the predictions of a convolutional neural network (CNN) for prostate tumor segmentation. The CNN uses a U-Net architecture which was trained on multi-parametric MRI data from 122 patients to automatically segment the prostate gland and prostate tumor lesions. In addition, co-registered ground truth data from whole mount histopathology images were available in 15 patients that were used as a test set during CNN testing. To be able to interpret the segmentation results of the CNN, heat maps were generated using the Gradient Weighted Class Activation Map (Grad-CAM) method. The CNN achieved a mean Dice Sorensen Coefficient 0.62 and 0.31 for the prostate gland and the tumor lesions -with the radiologist drawn ground truth and 0.32 with whole-mount histology ground truth for tumor lesions. Dice Sorensen Coefficient between CNN predictions and manual segmentations from MRI and histology data were not significantly different. In the prostate the Grad-CAM heat maps could differentiate between tumor and healthy prostate tissue, which indicates that the image information in the tumor was essential for the CNN segmentation.

Quantifying Tumor and Vasculature Deformations during Laryngoscopy

Retractors and scopes used in head and neck surgery to provide adequate surgical exposure also deform critical structures in the region. Surgeons typically use preoperative imaging to plan and guide their tumor resections, however the large tissue deformation resulting from placement of retractors and scopes reduces the utility of preoperative imaging as a reliable roadmap. We quantify the extent of tumor and vasculature deformation in patients with tumors of the larynx and pharynx undergoing diagnostic laryngoscopy. A mean tumor displacement of 1.02 cm was observed between the patients' pre- and intra-operative states. Mean vasculature displacement at key bifurcation points was 0.99 cm. Registration to the hyoid bone can reduce tumor displacement to 0.67 cm and improve carotid stem angle deviations but increase overall vasculature displacement. The large deformation results suggest limitations in reliance on preoperative imaging and that using specific landmarks intraoperatively or having more intraoperative information could help to compensate for these deviations and ultimately improve surgical success.

On the usage of average Hausdorff distance for segmentation performance assessment: hidden error when used for ranking

Average Hausdorff distance is a widely used performance measure to calculate the distance between two point sets. In medical image segmentation, it is used to compare ground truth images with segmentations allowing their ranking. We identified, however, ranking errors of average Hausdorff distance making it less suitable for applications in segmentation performance assessment. To mitigate this error, we present a modified calculation of this performance measure that we have coined “balanced average Hausdorff distance”. To simulate segmentations for ranking, we manually created non-overlapping segmentation errors common in magnetic resonance angiography cerebral vessel segmentation as our use-case. Adding the created errors consecutively and randomly to the ground truth, we created sets of simulated segmentations with increasing number of errors. Each set of simulated segmentations was ranked using both performance measures. We calculated the Kendall rank correlation coefficient between the segmentation ranking and the number of errors in each simulated segmentation. The rankings produced by balanced average Hausdorff distance had a significantly higher median correlation (1.00) than those by average Hausdorff distance (0.89). In 200 total rankings, the former misranked 52 whilst the latter misranked 179 segmentations. Balanced average Hausdorff distance is more suitable for rankings and quality assessment of segmentations than average Hausdorff distance.

Prostate cancer measurements on serial MRI during active surveillance: it's time to be PRECISE

OBJECTIVE

The PRECISE criteria for reporting multiparametric MRI in patients on active surveillance (AS) for prostate cancer (PCa) score the likelihood of clinically significant change over time using a 1-5 scale, where 4 or 5 indicates radiological progression. According to the PRECISE recommendations, the index lesion size can be reported using different definitions of volume (planimetry or ellipsoid formula) or by measuring one or two diameters. We compared different measurements using planimetry as the reference standard and stratified changes according to the PRECISE scores.

METHODS

We retrospectively analysed 196 patients on AS with PCa confirmed by targeted biopsy who had two MR scans (baseline and follow-up). Lesions were measured on T₂ weighted imaging (T₂WI) according to all definitions. A PRECISE score was assessed for each patient.

RESULTS

The ellipsoid formula exhibited the highest correlation with planimetry at baseline (ρ = 0.97) and follow-up (ρ = 0.98) imaging, compared to the biaxial measurement and single maximum diameter. There was a significant difference (p < 0.001) in the yearly percentage volume change between radiological regression/stability (PRECISE 2-3) and progression (PRECISE 4-5) for planimetry (39.64%) and for the ellipsoid formula (46.78%).

CONCLUSION

The ellipsoid formula could be used to monitor tumour growth during AS. Evidence of a significant yearly percentage volume change between radiological regression/stability (PRECISE 2-3) and progression (PRECISE 4-5) has been also observed.

ADVANCES IN KNOWLEDGE

The ellipsoid formula is a reasonable surrogate for planimetry in capturing tumour volume changes on T₂WI in patients on imaging-led AS. This is also associated with radiological changes using the PRECISE recommendations.

Test-retest repeatability of a deep learning architecture in detecting and segmenting clinically significant prostate cancer on apparent diffusion coefficient (ADC) maps

OBJECTIVES

To evaluate short-term test-retest repeatability of a deep learning architecture (U-Net) in slice- and lesion-level detection and segmentation of clinically significant prostate cancer (csPCa: Gleason grade group > 1) using diffusion-weighted imaging fitted with monoexponential function, ADC_m.

METHODS

One hundred twelve patients with prostate cancer (PCa) underwent 2 prostate MRI examinations on the same day. PCa areas were annotated using whole mount prostatectomy sections. Two U-Net-based convolutional neural networks were trained on three different ADC_m b value settings for (a) slice- and (b) lesion-level detection and (c) segmentation of csPCa. Short-term test-retest repeatability was estimated using intra-class correlation coefficient (ICC(3,1)), proportionate agreement, and dice similarity coefficient (DSC). A 3-fold cross-validation was performed on training set (N = 78 patients) and evaluated for performance and repeatability on testing data (N = 34 patients).

RESULTS

For the three ADC_m b value settings, repeatability of mean ADC_m of csPCa lesions was ICC(3,1) = 0.86-0.98. Two CNNs with U-Net-based architecture demonstrated ICC(3,1) in the range of 0.80-0.83, agreement of 66-72%, and DSC of 0.68-0.72 for slice- and lesion-level detection and segmentation of csPCa. Bland-Altman plots suggest that there is no systematic bias in agreement between inter-scan ground truth segmentation repeatability and segmentation repeatability of the networks.

CONCLUSIONS

For the three ADC_m b value settings, two CNNs with U-Net-based architecture were repeatable for the problem of detection of csPCa at the slice-level. The network repeatability in segmenting csPCa lesions is affected by inter-scan variability and ground truth segmentation repeatability and may thus improve with better inter-scan reproducibility.

KEY POINTS

• For the three ADC_m b value settings, two CNNs with U-Net-based architecture were repeatable for the problem of detection of csPCa at the slice-level. • The network repeatability in segmenting csPCa lesions is affected by inter-scan variability and ground truth segmentation repeatability and may thus improve with better inter-scan reproducibility.