用椭圆变换与深度学习结合，自动检测青光眼，准确率高达99.3%。
Ellipse Meets Bit-Planes: A Novel Approach to RNFL based Glaucoma Detection Using Advanced Image Processing and Deep Learning
- 基于椭圆极坐标变换增强视网膜神经纤维层分析
- 两种框架分别达99.3%和92.31%准确率，适配不同计算需求
- 适合资源有限地区快速筛查，具低成本可扩展性
本文提出一种基于自适应椭圆极坐标变换的综合流程，用于从常规彩色眼底图像中自动检测青光眼。该方法以视网膜神经纤维层（RNFL）为主要生物标志物，不受视盘和黄斑位置影响。通过该变换，设计了两种针对不同需求的框架：首个基于深度学习的特征融合框架在高精度场景下实现99.3%的检测率；第二个框架采用新型位平面切片图像处理算法，在低资源环境下实现92.31%准确率，支持快速推理。两种方案均具备可扩展性和成本效益，为青光眼早期筛查提供了有效工具，尤其适用于医疗资源匮乏地区。
原文摘要 · Abstract (English)
This work proposes an integrated pipeline for automatic glaucoma detection method from easily available colour fundas images based on an adaptive algorithm for ellipse-based polar transformation, to enhance the analysis of the Retinal Nerve Fiber Layer (RNFL) as the primary biomarker for observing glaucomatous changes, regardless of optic disc and macula position. Utilizing this transformation, we introduce two distinct frameworks tailored to different operational needs. The first framework, a deep learning-inspired feature fusion approach, achieves a 99.3% detection rate, ideal for settings where high precision is essential, despite higher computational demands. The second framework employs a novel image-processing algorithm based on bit-plane slicing, offering 92.31% accuracy and optimized for environments requiring rapid inference with minimal resource consumption. Both frameworks provide scalable and cost-effective solutions for early glaucoma detection. This study highlights the potential of RNFL-based diagnostic tools in addressing the global challenge of glaucoma, particularly in underserved regions.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。