基于感受野与多尺度路径增强的腰椎图像分割网络DPS-UNet

打开文本图片集

中图分类号：TP391.4文献标识码：A

文章编号：2096-4706（2026）04-0067-06

Lumbar Image Segmentation Network DPS-UNet Based on Receptive Field Amplification and Multi-scale Path Augmentation

HEZhiyuan，WANGCanhua （JiangxiUniversityof ChineseMedicine，Nanchang 33oo04， China）

Abstract： To addressegmentation challenges ofblued boundaries and background noisein lumbar MRIimages，this paper proposesDPS-UNet，segmentationnetworkasedonreceptivefieldamplificationandpathaugmentation.Firstly，a7-layedep encoderisconstructedtoamplifytheglobalceptiefeldectivelyapturingcomplextopologicalaturesofthebaine andovercomingthelimitations inreceptivefeldoftraditionalnetworks.Secondlyabotom-uppathaugmentationstructueis introducedtoientlytrnsmitshallowspatiallocalzationiformationtodelayers，eaning teexpressvepowerofeature pyramid.Simultaneouslyaparameter-freeSimAMatention module isembedded toadaptivelysuppress soft issuenoiseand reinforceedgeresponses.ExperimentalresultsdmonstratethatD-UNetignificantlyoutperformsmainstreammethods，withthe HD95 distance decreasing from 7.21 to 2.78 and mIoU reaching 86.45% .The ablation experiments further validate the synergistic effectiveness ofthe deep encoderand feature enhancement strategies， indicating promising clinical application potential.

Keywords： lumbar image segmentation; U-Net; SimAM; path augmentation; receptive field amplification

0 引言

随着医学影像技术的飞速发展，腰椎磁共振成像（MRI）已成为诊断椎间盘突出、椎管狭窄及脊柱侧弯等腰椎疾病的“金标准”。（剩余10202字）