An uncertainty‐aware deep learning architecture with outlier mitigation for prostate gland segmentation in radiotherapy treatment planning

Abstract

Purpose Task automation is essential for efficient and consistent image segmentation in radiation oncology. We report on a deep learning architecture, comprising a U‐Net and a variational autoencoder (VAE) for automatic contouring of the prostate gland incorporating interobserver variation for radiotherapy treatment planning. The U‐Net/VAE generates an ensemble set of segmentations for each image CT slice. A novel outlier mitigation (OM) technique was implemented to enhance the model segmentation accuracy. Methods The primary source dataset (source_prim) consisted of 19 200 CT slices (from 300 patient planning CT image datasets) with manually contoured prostate glands. A smaller secondary source dataset (source_sec) comprised 640 CT slices (from 10 patient CT datasets), where prostate glands were segmented by 5 independent physicians on each dataset to account for interobserver variability. Data augmentation via random rotation (<5 degrees), cropping, and horizontal flipping was applied to each dataset to increase sample size by a factor of 100. A probabilistic hierarchical U‐Net with VAE was implemented and pretrained using the augmented source_prim dataset for 30 epochs. Model parameters of the U‐Net/VAE were fine‐tuned using the augmented source_sec dataset for 100 epochs. After the first round of training, outlier contours in the training dataset were automatically detected and replaced by the most accurate contours (based on Dice similarity coefficient, DSC) generated by the model. The U‐Net/OM‐VAE was retrained using the revised training dataset. Metrics for comparison included DSC, Hausdorff distance (HD, mm), normalized cross‐correlation (NCC) coefficient, and center‐of‐mass (COM) distance (mm). Results Results for U‐Net/OM‐VAE with outliers replaced in the training dataset versus U‐Net/VAE without OM were as follows: DSC = 0.82 ± 0.01 versus 0.80 ± 0.02 (p = 0.019), HD = 9.18 ± 1.22 versus 10.18 ± 1.35 mm (p = 0.043), NCC = 0.59 ± 0.07 versus 0.62 ± 0.06, and COM = 3.36 ± 0.81 versus 4.77 ± 0.96 mm over the average of 15 contours. For the average of 15 highest accuracy contours, values were as follows: DSC = 0.90 ± 0.02 versus 0.85 ± 0.02, HD = 5.47 ± 0.02 versus 7.54 ± 1.36 mm, and COM = 1.03 ± 0.58 versus 1.46 ± 0.68 mm (p < 0.03 for all metrics). Results for the U‐Net/OM‐VAE with outliers removed were as follows: DSC = 0.78 ± 0.01, HD = 10.65 ± 1.95 mm, NCC = 0.46 ± 0.10, COM = 4.17 ± 0.79 mm for the average of 15 contours, and DSC = 0.88 ± 0.02, HD = 7.00 ± 1.17 mm, COM = 1.58 ± 0.63 mm for the average of 15 highest accuracy contours. All metrics for U‐Net/VAE trained on the source_prim and source_sec datasets via pretraining, followed by fine‐tuning, show statistically significant improvement over that trained on the source_sec dataset only. Finally, all metrics for U‐Net/VAE with or without OM showed statistically significant improvement over those for the standard U‐Net. Conclusions A VAE combined with a hierarchical U‐Net and an OM strategy (U‐Net/OM‐VAE) demonstrates promise toward capt...

Artificial Intelligence Uncertainty Quantification in Radiotherapy Applications - A Scoping Review

“…A minority of studies also used uncertainty in active learning frameworks, where the model selects the most informative data points for labeling based on their uncertainty, to improve model training. Ambiguity modeling and out-of-distribution detection were vastly underrepresented, with only one study each (Li et al [71], and Yang et al [80], respectively) investigating these areas.…”

Real-Time Reconstruction of HIFU Focal Temperature Field Based on Deep Learning

“…With the rapid development of artificial intelligence, deep learning–based medical image processing provides promising solutions in many data-driven clinical application challenges [ 16 – 18 ]. For example, big data technology enables breaking through the limits in various aspects such as tumor identification [ 19 ], super-resolution reconstruction [ 20 ], slice staining techniques [ 21 ], and treatment planning [ 22 ], to resolve the challenge mentioned above to monitor the temperature distribution in clinical applications. In this paper, a temperature distribution reconstruction strategy based on a deep multimodal teacher–student (MMTS) model is proposed to establish a nonlinear mapping relationship from ultrasonic echo signal to temperature.…”

A new architecture combining convolutional and transformer‐based networks for automatic 3D multi‐organ segmentation on CT images

“…[4][5][6] The manual contouring process in addition to often being inefficient can also suffer from inconsistencies in contouring preferences or related intra-and inter-observer uncertainties. 4,7 Inaccuracies in contouring impact on planning margin design-erroneous planning margins may lead to possible underdosage of the target and excess radiation delivered to surrounding healthy tissues. 8 To address these issues, a method for accurate automatic segmentation is needed to improve efficiency and consistency in radiation treatment planning.…”

“…Currently, the clinical practice for contour delineation involves a labor‐intensive and operator‐dependent manual process 4–6 . The manual contouring process in addition to often being inefficient can also suffer from inconsistencies in contouring preferences or related intra‐and inter‐observer uncertainties 4,7 . Inaccuracies in contouring impact on planning margin design—erroneous planning margins may lead to possible underdosage of the target and excess radiation delivered to surrounding healthy tissues 8 .…”