IEEE Trans Med Imaging. 2024 Aug;43(8):3044-3056. doi: 10.1109/TMI.2024.3383466. Epub 2024 Aug 1.
Ophthalmic diseases such as central serous chorioretinopathy (CSC) significantly impair the vision of millions of people globally. Precise segmentation of choroid and macular edema is critical for diagnosing and treating these conditions. However, existing 3D medical image segmentation methods often fall short due to the heterogeneous nature and blurry features of these conditions, compounded by medical image clarity issues and noise interference arising from equipment and environmental limitations. To address these challenges, we propose the Spectrum Analysis Synergy Axial-Spatial Network (SASAN), an approach that innovatively integrates spectrum features using the Fast Fourier Transform (FFT). SASAN incorporates two key modules: the Frequency Integrated Neural Enhancer (FINE), which mitigates noise interference, and the Axial-Spatial Elementum Multiplier (ASEM), which enhances feature extraction. Additionally, we introduce the Self-Adaptive Multi-Aspect Loss ( L ), which balances image regions, distribution, and boundaries, adaptively updating weights during training. We compiled and meticulously annotated the Choroid and Macular Edema OCT Mega Dataset (CMED-18k), currently the world's largest dataset of its kind. Comparative analysis against 13 baselines shows our method surpasses these benchmarks, achieving the highest Dice scores and lowest HD95 in the CMED and OIMHS datasets. Our code is publicly available at https://github.com/IMOP-lab/SASAN-Pytorch.
眼科疾病，如中心性浆液性脉络膜视网膜病变（CSC），严重影响了全球数百万人的视力。准确地分割脉络膜和黄斑水肿对于这些疾病的诊断和治疗至关重要。然而，现有的 3D 医学图像分割方法由于这些疾病的异质性和模糊特征，以及设备和环境限制引起的医学图像清晰度问题和噪声干扰，往往效果不佳。为了解决这些挑战，我们提出了 Spectrum Analysis Synergy Axial-Spatial Network（SASAN），这是一种创新地利用快速傅里叶变换（FFT）集成谱特征的方法。SASAN 包含两个关键模块：Frequency Integrated Neural Enhancer（FINE），它可以减轻噪声干扰，以及 Axial-Spatial Elementum Multiplier（ASEM），它可以增强特征提取。此外，我们引入了 Self-Adaptive Multi-Aspect Loss（L），它可以平衡图像区域、分布和边界，在训练过程中自适应地更新权重。我们编译并精心标注了 Choroid 和 Macular Edema OCT Mega Dataset（CMED-18k），这是目前世界上最大的同类数据集。与 13 个基线的对比分析表明，我们的方法优于这些基准，在 CMED 和 OIMHS 数据集上实现了最高的 Dice 分数和最低的 HD95。我们的代码在 https://github.com/IMOP-lab/SASAN-Pytorch 上公开可用。
