Multiview Self-Supervised Segmentation for OARs Delineation in Radiotherapy

Abstract

Radiotherapy has become a common treatment option for head and neck (H&N) cancer, and organs at risk (OARs) need to be delineated to implement a high conformal dose distribution. Manual drawing of OARs is time consuming and inaccurate, so automatic drawing based on deep learning models has been proposed to accurately delineate the OARs. However, state-of-the-art performance usually requires a decent amount of delineation, but collecting pixel-level manual delineations is labor intensive and may not be necessary for representation learning. Encouraged by the recent progress in self-supervised learning, this study proposes and evaluates a novel multiview contrastive representation learning to boost the models from unlabelled data. The proposed learning architecture leverages three views of CTs (coronal, sagittal, and transverse plane) to collect positive and negative training samples. Specifically, a CT in 3D is first projected into three 2D views (coronal, sagittal, and transverse planes), then a convolutional neural network takes 3 views as inputs and outputs three individual representations in latent space, and finally, a contrastive loss is used to pull representation of different views of the same image closer (“positive pairs”) and push representations of views from different images (“negative pairs”) apart. To evaluate performance, we collected 220 CT images in H&N cancer patients. The experiment demonstrates that our method significantly improves quantitative performance over the state-of-the-art (from 83% to 86% in absolute Dice scores). Thus, our method provides a powerful and principled means to deal with the label-scarce problem.

Deep Learning-Based Segmentation of Head and Neck Organs-at-Risk with Clinical Partially Labeled Data

“…These results showed that our model was comparable to previously published studies for HN OAR segmentation regarding the DSC [ 19 ] ( Table 4 ), and also to other studies using partially labeled or unlabeled datasets and following similar approaches to ours [ 7 , 11 , 25 ]. Almost all OARs reached a performance close to the maximum DSC reported in the literature, outstanding in some cases such as the temporomandibular joints and inner ears.…”

“…Furthermore, this process is thoroughly subjected to inter- and intra-practitioner variabilities [ 6 , 9 , 10 ] driven by diverse factors such as experience, availability, quality, and interpretation of diagnostic imaging [ 4 ]. The limited soft tissue contrast of computed tomography (CT) images is also a substantial problem in HN delineation [ 10 , 11 ], as many OARs have similar densities to fat, muscle, or other surrounding tissues. Segmenting these OARs, such as the parotid or submandibular glands, is particularly challenging [ 8 ].…”

“…Aiming to exploit the available data in a medical center, Liu et al [ 11 ] proposed a multiview contrastive representative learning framework. They generated 2D CT slices as the input to a self-supervised 2D CNN outputting three 3D segmentations and using a contrastive loss to select the correlated predictions.…”

ContourGAN: Auto‐contouring of organs at risk in abdomen computed tomography images using generative adversarial network

“…The introduction of self-supervised networks mitigates the effect of this issue. 30,31 A fascinating development in machine learning is the invention of a technique called GANs, 10 in which two models compete with each other to make their predictions more accurate. Since their debut in 2014, GANs have excelled in generative image modeling and demonstrated exceptional performance across various medical imaging applications, including classification 32,33 and segmentation.…”

“…The scarcity of labeled data in medical imaging is really affecting the performance of deep learning models. The introduction of self‐supervised networks mitigates the effect of this issue 30,31 …”