Prostate Contours Research Articles

OSAIRIS: Lessons Learned From the Hospital-Based Implementation and Evaluation of an Open-Source Deep-Learning Model for Radiotherapy Image Segmentation

Several studies report the benefits and accuracy of using autosegmentation for organ at risk (OAR) outlining in radiotherapy treatment planning. Typically, evaluations focus on accuracy metrics, and other parameters such as perceived utility and safety are routinely ignored. Here, we report our finding from the implementation and clinical evaluation of OSAIRIS, an open-source AI model for radiotherapy image segmentation that was carried out as part of its development into a medical device. The device contours OARs in the head and neck and male pelvis (referred to as the prostate model), and is designed to be used as a time-saving workflow device, alongside a clinician. Unlike standard evaluation processes, which heavily rely on accuracy metrics alone, our evaluation sought to demonstrate the tangible benefits, quantify utility and assess risk within a specific clinical workflow. We evaluated the time-saving benefit this device affords to clinicians, and how this time-saving might be linked to accuracy metrics, as well as the clinicians' assessment of the usability of the OSAIRIS contours in comparison to their colleagues' contours and those from other commercial AI contouring devices. Our safety evaluation focused on whether clinicians can notice and correct any errors should they be included in the output of the device.We found that OSAIRIS affords a significant time-saving of 36% (5.4 ± 2.1 minutes) when used for prostate contouring and 67% (30.3 ± 8.7 minutes) for head and neck contouring. Combining editing time data with accuracy metrics, we found the Hausdorff distance best correlated with editing-time, outperforming dice, the industry-standard, with a Spearman correlation coefficient of 0.70, and a Kendall coefficient of 0.52. Our safety and risk-mitigation exercise showed that anchoring bias is present when clinicians edit AI-generated contours, with the effect seemingly more pronounced for some structures over others. Most errors, however, were corrected by clinicians, with 72% of the head and neck errors 81% of the prostate errors removed in the editing step. Notably, our blinded clinician contour rating exercise showed that gold standard clinician contours are not rated more highly than the AI-generated contours.We conclude that evaluations of AI in a clinical setting must consider the clinical workflow in which the device will be used, and not rely on accuracy metrics alone, in order to reliably assess the benefits, utility and safety of the device. The effects of human-AI inter-operation must be evaluated to accurately assess the practical usability and potential uptake of the technology, as demonstrated in our blinded clinical utility review. The clinical risks posed by the use of the device must be studied and mitigated as far as possible, and our ‘Mystery Shopping’ experiment provides a template for future such assessments.

Deep-learning-based segmentation using individual patient data on prostate cancer radiation therapy.

Organ-at-risk segmentation is essential in adaptive radiotherapy (ART). Learning-based automatic segmentation can reduce committed labor and accelerate the ART process. In this study, an auto-segmentation model was developed by employing individual patient datasets and a deep-learning-based augmentation method for tailoring radiation therapy according to the changes in the target and organ of interest in patients with prostate cancer. Two computed tomography (CT) datasets with well-defined labels, including contoured prostate, bladder, and rectum, were obtained from 18 patients. The labels of the CT images captured during radiation therapy (CT2nd) were predicted using CT images scanned before radiation therapy (CT1st). From the deformable vector fields (DVFs) created by using the VoxelMorph method, 10 DVFs were extracted when each of the modified CT and CT2nd images were deformed and registered to the fixed CT1st image. Augmented images were acquired by utilizing 110 extracted DVFs and spatially transforming the CT1st images and labels. An nnU-net autosegmentation network was trained by using the augmented images, and the CT2nd label was predicted. A patient-specific model was created for 18 patients, and the performances of the individual models were evaluated. The results were evaluated by employing the Dice similarity coefficient (DSC), average Hausdorff distance, and mean surface distance. The accuracy of the proposed model was compared with those of models trained with large datasets. Patient-specific models were developed successfully. For the proposed method, the DSC values of the actual and predicted labels for the bladder, prostate, and rectum were 0.94 ± 0.03, 0.84 ± 0.07, and 0.83 ± 0.04, respectively. We demonstrated the feasibility of automatic segmentation by employing individual patient datasets and image augmentation techniques. The proposed method has potential for clinical application in automatic prostate segmentation for ART.

New Parametric 2D Curves for Modeling Prostate Shape in Magnetic Resonance Images

Geometric shape models often help to extract specific contours in digital images (the segmentation process) with major precision. Motivated by this idea, we introduce two models for the representation of prostate shape in the axial plane of magnetic resonance images. In more detail, the models are two parametric closed curves of the plane. The analytic study of the models includes the geometric role of the parameters describing the curves, symmetries, invariants, special cases, elliptic Fourier descriptors, conditions for simple curves and area of the enclosed surfaces. The models were validated for prostate shapes by fitting the curves to prostate contours delineated by a radiologist and measuring the errors with the mean distance, the Hausdorff distance and the Dice similarity coefficient. Validation was also conducted by comparing our models with the deformed superellipse model used in literature. Our models are equivalent in fitting metrics to the deformed superellipse model; however, they have the advantage of a more straightforward formulation and they depend on fewer parameters, implying a reduced computational time for the fitting process. Due to the validation, our models may be applied for developing innovative and performing segmentation methods or improving existing ones.

Prostate size, source configuration, and dosimetry dynamics of stranded 125I seed implants

PURPOSETo quantify changes in prostate size and seed movement over time after transperineal implantation of stranded 125I seeds, and to determine their impact on prostate dosimetry. METHODSCT and MR (T2, balanced steady-state free precession) image triplets were acquired on days 0, 3, 10, and 30 for a cohort of 20 patients and registered automatically. Prostate contours were drawn on MR-T2 images; seeds were found and matched in successive CT images. Prostate volume and dimensions, seed movements, and prostate dose metrics V200, V150, V100 and D90 were calculated, and their dynamic behaviors quantified in an operationally defined prostate coordinate system. RESULTSCohort-averaged reductions in prostate A-P dimension (∼8%) and L-R dimension (∼5%) inferred from seed movements agreed with those obtained from contour measurements, whereas prostate volume and S-I dimension (implant direction) reductions inferred from seed movements were overestimated by about 30%. Average overall seed movement was 4.8 ± 3.0 mm, of which the only identifiable systematic component was resolution of prostate edema. Cohort-averaged ratios of prostate V200, V150, V100, and D90 on day 30 relative to day 0 were 1.67, 1.33, 1.02, and 1.08, respectively. CONCLUSIONSPostimplant prostate size reduction in the SI (implant) direction cannot reliably be inferred from stranded seed movements. Apart from large-scale migration, residual seed movements relative to the prostate after accounting for edema resolution appear to be random. Prostate V100 and D90 changes 30 days post implant are modest, whereas those for V150 and V200 are substantial.

Performance deterioration of deep learning models after clinical deployment: a case study with auto-segmentation for definitive prostate cancer radiotherapy

Our study aims to explore the long-term performance patterns for deep learning (DL) models deployed in clinic and to investigate their efficacy in relation to evolving clinical practices. We conducted a retrospective study simulating the clinical implementation of our DL model involving 1328 prostate cancer patients treated between January 2006 and August 2022. We trained and validated a U-Net-based auto-segmentation model on data obtained from 2006 to 2011 and tested on data from 2012 to 2022, simulating the model’s clinical deployment starting in 2012. We visualized the trends of the model performance using exponentially weighted moving average (EMA) curves. Additionally, we performed Wilcoxon Rank Sum Test and multiple linear regression to investigate Dice similarity coefficient (DSC) variations across distinct periods and the impact of clinical factors, respectively. Initially, from 2012 to 2014, the model showed high performance in segmenting the prostate, rectum, and bladder. Post-2015, a notable decline in EMA DSC was observed for the prostate and rectum, while bladder contours remained stable. Key factors impacting the prostate contour quality included physician contouring styles, using various hydrogel spacers, CT scan slice thickness, MRI-guided contouring, and intravenous (IV) contrast (p < 0.0001, p < 0.0001, p = 0.0085, p = 0.0012, p < 0.0001, respectively). Rectum contour quality was notably influenced by factors such as slice thickness, physician contouring styles, and the use of various hydrogel spacers. The quality of the bladder contour was primarily affected by IV contrast. The deployed DL model exhibited a substantial decline in performance over time, aligning with the evolving clinical settings.

A preoperative scoring system for predicting the extraprostatic extension of prostate cancer following radical prostatectomy using magnetic resonance imaging and clinical factors.

We aimed to develop a preoperative prediction model for extraprostatic extension (EPE) in prostate cancer (PCa) patients following radical prostatectomy (RP) using MRI and clinical factors. This retrospective study enrolled 266 consecutive patients who underwent RP for PCa in 2022. These patients were divided into a training set (n = 187) and a test set (n = 79) through random assignment. The evaluated variables included age, prostate-specific antigen (PSA) level, prostate volume, PSA density (PSAD), index tumor length on MRI, Prostate Imaging-Reporting and Data System (PI-RADS) category, and EPE-related MRI features as defined by PI-RADS v2.1. A predictive model was constructed through multivariable logistic regression and subsequently translated into a scoring system. The performance of this scoring system in terms of prediction and calibration was assessed using C statistics and the Hosmer‒Lemeshow test. Among patients in the training and test cohorts, 74 (39.6%) and 25 (31.6%), respectively, exhibited EPE after RP. The formulated scoring system incorporated the following factors: PSAD, index tumor length, bulging prostatic contour, and tumor-capsule interface > 10mm as identified on MRI. This scoring system demonstrated strong prediction performance for EPE in both the training (C statistic, 0.87 [95% confidence interval, 0.86-0.87]) and test cohorts (C statistic, 0.85 [0.83-0.89]). Furthermore, the scoring system exhibited good calibration in both cohorts (P = 0.988 and 0.402, respectively). Our scoring system, built upon MRI features defined by the PI-RADS, offers valuable assistance in assessing the likelihood of EPE after RP.

Deep-learning-based joint rigid and deformable contour propagation for magnetic resonance imaging-guided prostate radiotherapy.

Deep learning-based unsupervised image registration has recently been proposed, promising fast registration. However, it has yet to be adopted in the online adaptive magnetic resonance imaging-guided radiotherapy (MRgRT)workflow. In this paper, we design an unsupervised, joint rigid, and deformable registration framework for contour propagation in MRgRT of prostatecancer. Three-dimensional pelvic T2-weighted MRIs of 143 prostate cancer patients undergoing radiotherapy were collected and divided into 110, 13, and 20 patients for training, validation, and testing. We designed a framework using convolutional neural networks (CNNs) for rigid and deformable registration. We selected the deformable registration network architecture among U-Net, MS-D Net, and LapIRN and optimized the training strategy (end-to-end vs. sequential). The framework was compared against an iterative baseline registration. We evaluated registration accuracy (the Dice and Hausdorff distance of the prostate and bladder contours), structural similarity index, and folding percentage to compare the methods. We also evaluated the framework's robustness to rigid and elastic deformations and bias fieldperturbations. The end-to-end trained framework comprising LapIRN for the deformable component achieved the best median (interquartile range) prostate and bladder Dice of0.89 (0.85-0.91) and 0.86 (0.80-0.91), respectively. This accuracy was comparable to the iterative baseline registration: prostate and bladder Dice of 0.91 (0.88-0.93) and 0.86 (0.80-0.92). The best models complete rigid and deformable registration in 0.002 (0.0005) and 0.74 (0.43) s (Nvidia Tesla V100-PCIe 32 GB GPU), respectively. We found that the models are robust to translations up to 52 mm, rotations up to 15 , elastic deformations up to 40 mm, and biasfields. Our proposed unsupervised, deep learning-based registration framework can perform rigid and deformable registration in less than a second with contour propagation accuracy comparable with iterative registration.

Deep learning performance on MRI prostate gland segmentation: evaluation of two commercially available algorithms compared with an expert radiologist.

Accurate whole-gland prostate segmentation is crucial for successful ultrasound-MRI fusion biopsy, focal cancer treatment, and radiation therapy techniques. Commercially available artificial intelligence (AI) models, using deep learning algorithms (DLAs) for prostate gland segmentation, are rapidly increasing in numbers. Typically, their performance in a true clinical context is scarcely examined or published. We used a heterogenous clinical MRI dataset in this study aiming to contribute to validation of AI-models. We included 123 patients in this retrospective multicenter (7 hospitals), multiscanner (8 scanners, 2 vendors, 1.5T and 3T) study comparing prostate contour assessment by 2 commercially available Food and Drug Association (FDA)-cleared and CE-marked algorithms (DLA1 and DLA2) using an expert radiologist's manual contours as a reference standard (RSexp) in this clinical heterogeneous MRI dataset. No in-house training of the DLAs was performed before testing. Several methods for comparing segmentation overlap were used, the Dice similarity coefficient (DSC) being the most important. The DSC mean and standard deviation for DLA1 versus the radiologist reference standard (RSexp) was and for DLA2 versus RSexp it was . A paired -test to compare the DSC for DLA1 and DLA2 showed no statistically significant difference (). Two commercially available DL algorithms (FDA-cleared and CE-marked) can perform accurate whole-gland prostate segmentation on a par with expert radiologist manual planimetry on a real-world clinical dataset. Implementing AI models in the clinical routine may free up time that can be better invested in complex work tasks, adding more patient value.

Optimal planning target margin for prostate radiotherapy based on interfractional and intrafractional variability assessment during 1.5T MRI-guided radiotherapy.

We analyzed daily pre-treatment- (PRE) and real-time motion monitoring- (MM) MRI scans of patients receiving definitive prostate radiotherapy (RT) with 1.5 T MRI guidance to assess interfractional and intrafractional variability of the prostate and suggest optimal planning target volume (PTV) margin. Rigid registration between PRE-MRI and planning CT images based on the pelvic bone and prostate anatomy were performed. Interfractional setup margin (SM) and interobserver variability (IO) were assessed by comparing the centroid values of prostate contours delineated on PRE-MRIs. MM-MRIs were used for internal margin (IM) assessment, and PTV margin was calculated using the van Herk formula. We delineated 400 prostate contours on PRE-MRI images. SM was 0.57 ± 0.42, 2.45 ± 1.98, and 2.28 ± 2.08 mm in the left-right (LR), anterior-posterior (AP), and superior-inferior (SI) directions, respectively, after bone localization and 0.76 ± 0.57, 1.89 ± 1.60, and 2.02 ± 1.79 mm in the LR, AP, and SI directions, respectively, after prostate localization. IO was 1.06 ± 0.58, 2.32 ± 1.08, and 3.30 ± 1.85 mm in the LR, AP, and SI directions, respectively, after bone localization and 1.11 ± 0.55, 2.13 ± 1.07, and 3.53 ± 1.65 mm in the LR, AP, and SI directions, respectively, after prostate localization. Average IM was 2.12 ± 0.86, 2.24 ± 1.07, and 2.84 ± 0.88 mm in the LR, AP, and SI directions, respectively. Calculated PTV margin was 2.21, 5.16, and 5.40 mm in the LR, AP, and SI directions, respectively. Movements in the SI direction were the largest source of variability in definitive prostate RT, and interobserver variability was a non-negligible source of margin. The optimal PTV margin should also consider the internal margin.

Optimal Planning Target Margin for Prostate Radiotherapy Based on Interfractional and Intrafractional Variability Assessment during 1.5T MR-Guided Radiotherapy.

MR-guided radiotherapy (MRgRT) provides superior soft-tissue contrast over CT-based image guidance. We collected and analyzed daily pre-treatment (PRE) and real-time motion-monitoring (MM) MR images of patients receiving prostate radiotherapy to assess interfractional and intrafractional variability of prostate using two localization methods: pelvic bony anatomy (bone) and prostate during online adaptive radiotherapy (ART). PRE and MM MRIs for the first five fractions of twenty prostate cancer patients who received definitive MRgRT with 1.5T MRI were collected. Using MIM software, rigid registration between PRE MRI and planning CT images based on pelvic bony anatomy and prostate reproduced bone localization and online ART, respectively. To determine interfractional setup margin (SM), prostate was delineated on all PRE MRIs registered after bone and prostate localizations by a radiation oncologist, and centroid values of prostate contours between planning CT and PRE MRIs were compared. To determine interobserver variability, another radiation oncologist, a medical physicist, and a radiotherapist contoured prostate for both localization methods. For internal margin (IM) assessment, we used MM MRIs of the five patients who had all three sets of coronal, sagittal, and axial cine images and determined the maximum contour displacement using in-house MATLAB-based software converting binary image files to 2D cine images with a superimposed grid of 1 mm spacing. A total of 100 PRE and 25 MM MRIs were analyzed. Four hundred prostate contours were delineated on MR images registered with planning CT based on both bony anatomy and prostate. After bone localization, SM was 0.57±0.42 mm in left-right (LR), 2.45±1.98 mm in anterior-posterior (AP), and 2.28±2.08 mm in superior-inferior (SI) directions, and IO was 1.06±0.58 mm in LR, 2.32±1.08 mm in AP, and 3.30±1.85 mm in SI directions. After prostate localization, SM was 0.76±0.57 mm in LR, 1.89±1.60 mm in AP, and 2.2±1.79 mm in SI directions, and IO was 1.11±0.55 mm in LR, 2.13±1.07 mm in AP, and 3.53±1.65 mm in SI directions. Average IM was 2.12±0.86 mm, 2.24±1.07 mm, and 2.84±0.88 mm in LR, AP, and SI directions, respectively. Using daily MRIs from MRgRT, we showed that movements in the SI direction were the largest source of variability in prostate definitive RT. In addition, interobserver variability was a non-negligible source of margin. Optimal PTV margin should also consider internal margin, especially in the SI direction.

Clinical Impact of Contouring Variability for Prostate Cancer Tumor Boost.

In the FLAME randomized phase III trial, adding a focal radiotherapy (RT) boost to tumors visible on MRI improved prostate cancer disease-free survival, local control, and regional/distant metastasis-free survival without increasing toxicity. In a prospective study (ReIGNITE RT Boost), we found substantial variability in radiation oncologists' attempts to contour prostate cancer tumors on MRI. Participants' accuracy and reliability improved when they used a quantitative MRI biomarker for cancer called the restriction spectrum imaging restriction score (RSIrs). Here, we measure the impact of radiation oncologists' tumor contour attempts on RT plans and predicted probability of biochemical failure. A total of 44 radiation oncologists (participants) from multiple institutions contoured prostate tumors on 30 patient cases, some with only conventional MRI and some with conventional MRI plus RSIrs maps. We developed a knowledge-based planning automated algorithm to generate RT plans with focal tumor boost per the FLAME trial protocol: 77 Gy in 35 fractions to prostate and integrated boost up to 95 Gy to the focal target, provided no normal tissue constraints were violated. We applied this algorithm to each participant's tumor contour and compared dosimetric parameters to those achieved when using the expert-defined tumor (consensus of two radiologists and a radiation oncologist). The primary metric was dose covering 98% of the expert-defined tumor (D98%), which was associated with probability of biochemical failure in a model published with the FLAME trial. In this preliminary analysis, 42 target volumes were analyzed from 20 participants and two patient cases: case 1 was contoured with conventional MRI alone and case 2 with RSIrs. All plans had adequate coverage of the prostate and met all key normal tissue constraints. For case 1 (without RSIrs), the expert's D98% was 87.1 Gy. By comparison, median D98% for participants was 82.2 Gy (IQR 77.8 - 84.6 Gy). Per the FLAME trial model, the predicted probability of biochemical failure at 7 years is 6% for the expert, but participants' plans yielded a median failure probability of 11% (IQR 18 - 9%). For case 2 (with RSIrs), the expert's D98% was 82.8 Gy, while median D98% for participants was 80.6 Gy (IQR 80.0 - 81.0 Gy). Predicted probability of biochemical failure is 12% for the expert-defined target and median 13% (IQR 14 - 13%) for participants. Variability in radiation oncologists' prostate tumor contours can lead to clinically meaningful changes to focal RT boost plans. The probability of biochemical failure for one patient case increased from 6% to a median of 11% when using conventional MRI alone. Use of RSIrs may mitigate this problem by increasing the accuracy and reliability of radiation oncologists' tumor contours.

Observer preference of artificial intelligence-generated versus clinical prostate contours for ultrasound-based high dose rate brachytherapy.

For trans-rectal ultrasound (TRUS)-based high dose rate (HDR) prostate brachytherapy, prostate contouring can be challenging due to artifacts from implanted needles, bleeding, and calcifications. To evaluate the geometric accuracy and observer preference of an artificial intelligence (AI) algorithm for generating prostate contours on TRUS images with implanted needles. We conducted a retrospective study of 150 patients, who underwent HDR brachytherapy. These patients were randomly divided into training (104), validation (26) and testing (20) sets. An AI algorithm was trained/validated utilizing the TRUS image and reference (clinical) contours. The algorithm then provided contours for the test set. For evaluation, we calculated the Dice coefficient between AI and reference prostate contours. We then presented AI and reference contours to eight clinician observers, and asked observers to select their preference. Observers were blinded to the source of contours. We calculated the percentage of cases in which observers preferred AI contours. Lastly, we evaluate whether the presence of AI contours improved the geometric accuracy of prostate contours provided by five resident observers for a 10-patient subset. The median Dice coefficient between AI and reference contours was 0.92 (IQR: 0.90-0.94). Observers preferred AI contours for a median of 57.5% (IQR: 47.5, 65.0) of the test cases. For resident observers, the presence of AI contours was associated with a 0.107 (95% CI: 0.086, 0.128; p<0.001) improvement in Dice coefficient for the 10-patient subset. The AI algorithm provided high-quality prostate contours on TRUS with implanted needles. Further prospective study is needed to better understand how to incorporate AI prostate contours into the TRUS-based HDR brachytherapy workflow.

Practice-based training strategy for therapist-driven prostate MR-Linac adaptive radiotherapy.

To develop a practice-based training strategy to transition from radiation oncologist to therapist-driven prostate MR-Linac adaptive radiotherapy. In phase 1, 7 therapists independently contoured the prostate and organs-at-risk on T2-weighted MR images from 11 previously treated MR-Linac prostate patients. Contours were evaluated quantitatively (i.e. Dice similarity coefficient [DSC] calculated against oncologist generated online contours) and qualitatively (i.e. oncologist using a 5-point Likert scale; a score≥4 was deemed a pass, a 90% pass rate was required to proceed to the next phase). Phase 2 consisted of supervised online workflow with therapists required no intervention from the oncologist on 10 total cases to advance. Phase 3 involved unsupervised therapist-driven workflow, with offline support from oncologists prior to the next fraction. In phase 1, the mean DSC was 0.92 (range 0.85-0.97), and mean Likert score was 3.7 for the prostate. Five therapists did not attain a pass rate (3-5 cases with prostate contour score<4), underwent follow-up one-on-one review, and performed contours on a further training set (n=5). Each participant completed a median of 12 (range 10-13) cases in phase 2; of 82 cases, minor direction were required from the oncologist on 5 regarding target contouring. Radiation oncologists reviewed 179 treatment fractions in phase 3, and deemed 5 cases acceptable but with suggestions for next fraction; all other cases were accepted without suggestions. A training stepwise program was developed and successfully implemented to enable a therapist-driven workflow for online prostate MR-Linac adaptive radiotherapy.

PSOR04 Presentation Time: 12:15 PM: Observer Preference of AI-Generated versus Clinically-Utilized Prostate Contours for Ultrasound-Based High Dose Rate Brachytherapy

PSOR04 Presentation Time: 12:15 PM: Observer Preference of AI-Generated versus Clinically-Utilized Prostate Contours for Ultrasound-Based High Dose Rate Brachytherapy

Reference standard for the evaluation of automatic segmentation algorithms: Quantification of inter observer variability of manual delineation of prostate contour on MRI

PurposeThe purpose of this study was to investigate the relationship between inter-reader variability in manual prostate contour segmentation on magnetic resonance imaging (MRI) examinations and determine the optimal number of readers required to establish a reliable reference standard. Materials and methodsSeven radiologists with various experiences independently performed manual segmentation of the prostate contour (whole-gland [WG] and transition zone [TZ]) on 40 prostate MRI examinations obtained in 40 patients. Inter-reader variability in prostate contour delineations was estimated using standard metrics (Dice similarity coefficient [DSC], Hausdorff distance and volume-based metrics). The impact of the number of readers (from two to seven) on segmentation variability was assessed using pairwise metrics (consistency) and metrics with respect to a reference segmentation (conformity), obtained either with majority voting or simultaneous truth and performance level estimation (STAPLE) algorithm. ResultsThe average segmentation DSC for two readers in pairwise comparison was 0.919 for WG and 0.876 for TZ. Variability decreased with the number of readers: the interquartile ranges of the DSC were 0.076 (WG) / 0.021 (TZ) for configurations with two readers, 0.005 (WG) / 0.012 (TZ) for configurations with three readers, and 0.002 (WG) / 0.0037 (TZ) for configurations with six readers. The interquartile range decreased slightly faster between two and three readers than between three and six readers. When using consensus methods, variability often reached its minimum with three readers (with STAPLE, DSC = 0.96 [range: 0.945–0.971] for WG and DSC = 0.94 [range: 0.912–0.957] for TZ, and interquartile range was minimal for configurations with three readers. ConclusionThe number of readers affects the inter-reader variability, in terms of inter-reader consistency and conformity to a reference. Variability is minimal for three readers, or three readers represent a tipping point in the variability evolution, with both pairwise-based metrics or metrics with respect to a reference. Accordingly, three readers may represent an optimal number to determine references for artificial intelligence applications.

A mathematical and neural network-based hybrid technique for detecting the prostate contour from medical image data

Extraction of prostate contour on transrectal ultrasound (TRUS) is significant in prostate cancer diagnosis. With the low-contrast images of TRUS and the existence of imaging artifacts, detecting the prostate accurately from TRUS images becomes more and more difficult. In this paper, we present four strategies to achieve improved segmentation precision on difficult images. First, to achieve the data sequence, we adopt a variant of the principal curve-based algorithm, where a small part of prior point information is utilized as the approximate initialization. Secondly, we devise an evolution neural network to assist in finding the optimal neural network. Thirdly, we train the fractional-order-based neural network using data sequence as input, thereby decreasing the model error and increasing the precision of results during training. Finally, a smooth and explainable mathematical function of organ boundary is represented by the parameters of a fractional-order-based neural network. The performance of our method was evaluated on two clinical datasets with 1,338 TRUS prostate images using Accuracy (ACC), Dice similarity coefficient (DSC), and Jaccard similarity coefficient (OMG). Results show that our model 1) achieves better performance than the state-of-the-art segmentation models; the corresponding ACC, DSC, and OMG of our method are 95.3 ± 2.2 %, 95.9 ± 2.3 %, and 94.9 ± 2.4 %, respectively, and 2) detects unseen or vague regions satisfactorily.

A Technique to Enable Efficient Adaptive Radiation Therapy: Automated Contouring of Prostate and Adjacent Organs

The purpose of this work was to investigate the use of a segmentation approach that could potentially improve the speed and reproducibility of contouring during magnetic resonance-guided adaptive radiation therapy. The segmentation algorithm was based on a hybrid deep neural network and graph optimization approach that also allows rapid user intervention (Deep layered optimal graph image segmentation of multiple objects and surfaces [LOGISMOS]+just enough interaction [JEI]). A total of 115 magnetic resonance-data sets were used for training and quantitative assessment. Expert segmentations were used as the independent standard for the prostate, seminal vesicles, bladder, rectum, and femoral heads for all 115 data sets. In addition, 3 independent radiation oncologists contoured the prostate, seminal vesicles, and rectum for a subset of patients such that the interobserver variability could be quantified. Consensus contours were then generated from these independent contours using a simultaneous truth and performance level estimation approach, and the deviation of Deep LOGISMOS+JEI contours to the consensus contours was evaluated and compared with the interobserver variability. The absolute accuracy of Deep LOGISMOS+JEI generated contours was evaluated using median absolute surface-to-surface distance which ranged from a minimum of 0.20 mm for the bladder to a maximum of 0.93 mm for the prostate compared with the independent standard across all data sets. The median relative surface-to-surface distance was less than 0.17 mm for all organs, indicating that the Deep LOGISMOS+JEI algorithm did not exhibit a systematic under- or oversegmentation. Interobserver variability testing yielded a mean absolute surface-to-surface distance of 0.93, 1.04, and 0.81 mm for the prostate, seminal vesicles, and rectum, respectively. In comparison, the deviation of Deep LOGISMOS+JEI from consensus simultaneous truth and performance level estimation contours was 0.57, 0.64, and 0.55 mm for the same organs. On average, the Deep LOGISMOS algorithm took less than 26 seconds for contour segmentation. Deep LOGISMOS+JEI segmentation efficiently generated clinically acceptable prostate and normal tissue contours, potentially limiting the need for time intensive manual contouring with each fraction.

Prostate segmentation accuracy using synthetic MRI for high-dose-rate prostate brachytherapy treatment planning

Objective. Both computed tomography (CT) and magnetic resonance imaging (MRI) images are acquired for high-dose-rate (HDR) prostate brachytherapy patients at our institution. CT is used to identify catheters and MRI is used to segment the prostate. To address scenarios of limited MRI access, we developed a novel generative adversarial network (GAN) to generate synthetic MRI (sMRI) from CT with sufficient soft-tissue contrast to provide accurate prostate segmentation without MRI (rMRI). Approach. Our hybrid GAN, PxCGAN, was trained utilizing 58 paired CT-MRI datasets from our HDR prostate patients. Using 20 independent CT-MRI datasets, the image quality of sMRI was tested using mean absolute error (MAE), mean squared error (MSE), peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM). These metrics were compared with the metrics of sMRI generated using Pix2Pix and CycleGAN. The accuracy of prostate segmentation on sMRI was evaluated using the Dice similarity coefficient (DSC), Hausdorff distance (HD) and mean surface distance (MSD) on the prostate delineated by three radiation oncologists (ROs) on sMRI versus rMRI. To estimate inter-observer variability (IOV), these metrics between prostate contours delineated by each RO on rMRI and the prostate delineated by treating RO on rMRI (gold standard) were calculated. Main results. Qualitatively, sMRI images show enhanced soft-tissue contrast at the prostate boundary compared with CT scans. For MAE and MSE, PxCGAN and CycleGAN have similar results, while the MAE of PxCGAN is smaller than that of Pix2Pix. PSNR and SSIM of PxCGAN are significantly higher than Pix2Pix and CycleGAN (p < 0.01). The DSC for sMRI versus rMRI is within the range of the IOV, while the HD for sMRI versus rMRI is smaller than the HD for the IOV for all ROs (p ≤ 0.03). Significance. PxCGAN generates sMRI images from treatment-planning CT scans that depict enhanced soft-tissue contrast at the prostate boundary. The accuracy of prostate segmentation on sMRI compared to rMRI is within the segmentation variation on rMRI between different ROs.

Prostate-specific membrane antigen positron emission tomography detected local failure after post-prostatectomy radiation therapy: Low rates of out-of-field recurrence validates current Australian prostate bed contouring guidelines.

The Australian Faculty of Radiation Oncology Genitourinary Group (FROGG) developed prostate bed clinical target volume (CTV) contouring guidelines which were subsequently used to develop the National EviQ guidelines for adjuvant and salvage post-prostatectomy radiotherapy (PPRT). These guidelines were based mainly upon consensus agreement. With the advent of prostate-specific membrane antigen (PSMA) positron emission tomography (PET), sites of recurrence can now be detected with low prostate-specific antigen (PSA) levels following radical prostatectomy. We evaluated sites of recurrence in patients treated with FROGG/EviQ CTVs to inform upcoming modifications of these guidelines. At our institution, we use the FROGG/EviQ guidelines for PPRT. From 2015, patients with PSA failure following PPRT have been re-staged using PSMA PET imaging. We identified patients with PET-avid local, nodal, and distant recurrences, fusing them with original treatment plans to determine whether recurrences were within or outside the prostate bed CTV. Regional nodal failures were reviewed to determine if they were within current elective node contouring guidelines. Ninety-four patients had positive PSMA PET following PPRT. Nine (9.6%) recurrences were local, seven being local-only. One local recurrence (1.1%) was just superior to the contoured prostate bed CTV, located within the vas deferens. Seventy-three (77.7%) patients had a component of node failure, with 56 (59.6%) having node-only failure. Sites of nodal relapses were covered by standard contouring guidelines 60.3% of the time. The low recurrence rate outside of current prostate bed CTV contouring guidelines is consistent with other studies using contemporary contouring, and validates the efficacy of the current FROGG/EviQ prostate bed CTV definition.

Dosimetric and workflow impact of synthetic-MRI use in prostate high-dose-rate brachytherapy

Target and organ delineation during prostate high-dose-rate (HDR) brachytherapy treatment planning can be improved by acquiring both a postimplant CT and MRI. However, this leads to a longer treatment delivery workflow and may introduce uncertainties due to anatomical motion between scans. We investigated the dosimetric and workflow impact of MRI synthesized from CT for prostate HDR brachytherapy. Seventy-eight CT and T2-weighted MRI datasets from patients treated with prostate HDR brachytherapy at our institution were retrospectively collected to train and validate our deep-learning-based image-synthesis method. Synthetic MRI was assessed against real MRI using the dice similarity coefficient (DSC) between prostate contours drawn using both image sets. The DSC between the same observer's synthetic and real MRI prostate contours was compared with the DSC between two different observers' real MRI prostate contours. New treatment plans were generated targeting the synthetic MRI-defined prostate and compared with the clinically delivered plans using target coverage and dose to critical organs. Variability between the same observer's prostate contours from synthetic and real MRI was not significantly different from the variability between different observer's prostate contours on real MRI. Synthetic MRI-planned target coverage was not significantly different from that of the clinically delivered plans. There were no increases above organ institutional dose constraints in the synthetic MRI plans. We developed and validated a method for synthesizing MRI from CT for prostate HDR brachytherapy treatment planning. Synthetic MRI use may lead to a workflow advantage and removal of CT-to-MRI registration uncertainty without loss of information needed for target delineation and treatment planning.