SAA-SDM: Neural Networks Faster Learned to Segment Organ Images

• DOI: 10.1007/s10278-023-00947-1
• Publication Date: 2024-01-10T12:02:27Z
• Authors: Chao Gao; Yongtao Shi; Shuai Yang; Bangjun Lei
• Keywords: Feature (linguistics)

In the field of medicine, rapidly and accurately segmenting organs in medical images is a crucial application computer technology. This paper introduces feature map module, Strength Attention Area Signed Distance Map (SAA-SDM), based on principal component analysis (PCA) principle. The module designed to accelerate neural networks' convergence speed achieving high precision. SAA-SDM provides network with confidence information regarding target background, similar signed distance (SDM), thereby enhancing network's understanding semantic related target. Furthermore, this presents training scheme tailored for aiming achieve finer segmentation improved generalization performance. Validation our approach carried out using TRUS chest X-ray datasets. Experimental results demonstrate that method significantly enhances For instance, UNet UNET + by more than 30%. Moreover, Segformer achieves an increase over 6% 3% mIoU (mean Intersection Union) two test datasets without requiring pre-trained parameters. Our reduces time resource costs associated networks organ tasks while effectively guiding meaningful learning even