Enhancing Single-Slice Segmentation with 3D-to-2D Unpaired Scan Distillation
This work addresses the challenge of accurate organ segmentation in low-radiation 2D CT scans for medical imaging, offering a method to reduce annotation burdens, though it is incremental as it builds on existing knowledge distillation techniques.
The paper tackles the problem of 2D single-slice CT segmentation by proposing a 3D-to-2D distillation framework that uses unpaired 3D scans to enhance 2D network performance, achieving improved segmentation results and efficacy in low-data regimes, such as outperforming models with all training data using only 200 subjects.
2D single-slice abdominal computed tomography (CT) enables the assessment of body habitus and organ health with low radiation exposure. However, single-slice data necessitates the use of 2D networks for segmentation, but these networks often struggle to capture contextual information effectively. Consequently, even when trained on identical datasets, 3D networks typically achieve superior segmentation results. In this work, we propose a novel 3D-to-2D distillation framework, leveraging pre-trained 3D models to enhance 2D single-slice segmentation. Specifically, we extract the prediction distribution centroid from the 3D representations, to guide the 2D student by learning intra- and inter-class correlation. Unlike traditional knowledge distillation methods that require the same data input, our approach employs unpaired 3D CT scans with any contrast to guide the 2D student model. Experiments conducted on 707 subjects from the single-slice Baltimore Longitudinal Study of Aging (BLSA) dataset demonstrate that state-of-the-art 2D multi-organ segmentation methods can benefit from the 3D teacher model, achieving enhanced performance in single-slice multi-organ segmentation. Notably, our approach demonstrates considerable efficacy in low-data regimes, outperforming the model trained with all available training subjects even when utilizing only 200 training subjects. Thus, this work underscores the potential to alleviate manual annotation burdens.