LightMIS：无需逐阶段解码器的超轻量级医学图像分割

中文摘要 AI 辅助

我们提出LightMIS，一个可扩展的超轻量级卷积网络家族，用于二维二值医学图像分割，无需学习逐阶段解码器。LightMIS使用尺度对齐投影块将五级编码器的输出对齐到公共分辨率，一次性聚合它们，并使用自适应融合级联细化融合表示。该级联将自适应核融合与所提出的渐进式感受野融合模块相结合，该模块利用临时通道扩展、互补深度可分离感受野和渐进式跨分支信息传递。我们在DRIVE、Kvasir-SEG、DSB18、BUSI、ISIC-2017和ISIC-2018上，使用共同的nnU-Net v2.3.1协议，通过五折交叉验证评估了LightMIS-T、LightMIS-S和LightMIS。完整的LightMIS包含0.131百万参数，对于3×256×256的输入需要0.575 GFLOPs，分别实现了模态宏观Dice和IoU分数86.71%和78.99%。Mobile U-ViT获得86.75%的Dice和79.07%的IoU，因此观察到的差异分别为0.04和0.08个百分点。相对于Mobile U-ViT、nnWNet和nnU-Net，LightMIS将参数数量减少了90.58%至99.61%，GFLOPs减少了82.54%至96.14%。在Arm Mali-G52 MC2 GPU上，所有LightMIS变体实现了完全GPU委派，中位委派延迟范围从LightMIS-T的53.31毫秒到LightMIS的138.31毫秒。这些结果展示了在所评估任务中有利的精度-复杂度权衡和片上执行可行性。代码可在https://github.com/AndreiiArhire/LightMIS公开获取。

英文摘要

We present LightMIS, a scalable family of ultra-lightweight convolutional networks for 2D binary medical image segmentation without a learned stage-wise decoder. LightMIS aligns the outputs of a five-level encoder to a common resolution using Scale-Aligned Projection blocks, aggregates them once, and refines the fused representation with an Adaptive Fusion Cascade. The cascade combines Adaptive Kernel Fusion with the proposed Progressive Receptive Fusion module, which uses temporary channel expansion, complementary depthwise receptive fields, and progressive cross-branch information transfer. We evaluate LightMIS-T, LightMIS-S, and LightMIS using five-fold cross-validation under a common nnU-Net v2.3.1 protocol on DRIVE, Kvasir-SEG, DSB18, BUSI, ISIC-2017, and ISIC-2018. Full LightMIS contains 0.131 M parameters and requires 0.575 GFLOPs for a $3\times256\times256$ input, achieving modality-macro Dice and IoU scores of 86.71% and 78.99%, respectively. Mobile U-ViT obtains 86.75% Dice and 79.07% IoU, so the observed differences are 0.04 and 0.08 percentage points. Relative to Mobile U-ViT, nnWNet, and nnU-Net, LightMIS reduces parameter count by 90.58$-$99.61% and GFLOPs by 82.54$-$96.14%. On an Arm Mali-G52 MC2 GPU, all LightMIS variants achieve full GPU delegation, with median delegated latency ranging from 53.31 ms for LightMIS-T to 138.31 ms for LightMIS. These results demonstrate a favorable accuracy-complexity trade-off and on-device execution feasibility for the evaluated tasks. The code is publicly available at https://github.com/AndreiiArhire/LightMIS.

发表机构

• Alexandru Ioan Cuza University of Iasi（亚历山德鲁·伊万·库扎大学）
• University of Wurzburg（维尔茨堡大学）

机构由 AI 辅助整理，请以论文原文为准。