SwinVFTR: A Novel Volumetric Feature-learning Transformer for 3D OCT Fluid Segmentation

Sharif Amit KamranKhondker Fariha HossainAlireza TavakkoliStewart Lee ZuckerbrodKenton M. SandersSalah A. Baker

Mar 2023

摘要

Accurately segmenting fluid in 3D volumetric optical coherence tomography(OCT) images is a crucial yet challenging task for detecting eye diseases.Traditional autoencoding-based segmentation approaches have limitations inextracting fluid regions due to successive resolution loss in the encodingphase and the inability to recover lost information in the decoding phase.Although current transformer-based models for medical image segmentationaddresses this limitation, they are not designed to be applied out-of-the-boxfor 3D OCT volumes, which have a wide-ranging channel-axis size based ondifferent vendor device and extraction technique. To address these issues, wepropose SwinVFTR, a new transformer-based architecture designed for precisefluid segmentation in 3D volumetric OCT images. We first utilize a channel-wisevolumetric sampling for training on OCT volumes with varying depths (B-scans).Next, the model uses a novel shifted window transformer block in the encoder toachieve better localization and segmentation of fluid regions. Additionally, wepropose a new volumetric attention block for spatial and depth-wise attention,which improves upon traditional residual skip connections. Consequently,utilizing multi-class dice loss, the proposed architecture outperforms otherexisting architectures on the three publicly available vendor-specific OCTdatasets, namely Spectralis, Cirrus, and Topcon, with mean dice scores of 0.72,0.59, and 0.68, respectively. Additionally, SwinVFTR outperforms otherarchitectures in two additional relevant metrics, mean intersection-over-union(Mean-IOU) and structural similarity measure (SSIM).

展开全部

图表提取

论文十问

参考文献

被引用

提问与回答

暂无人提供速读十问回答

被引用

笔记

问答

领域