AI Briefing
KOSign in

LightMIS achieves near-nnWNet accuracy with 98.14% fewer parameters for edge medical imaging

·2026.11.03 05:22

Key point

The 131K-parameter model runs 10.5x faster than nnWNet on Arm Mali-G52 MC2 GPUs with 100% operator coverage.

Details

  • LightMIS is a 131K-parameter 2D binary medical image segmentation model that aggregates multi-scale encoder features directly without a learned stage-wise decoder.
  • The model achieves close observed modality-macro Dice and IoU averages to nnWNet while using 98.14% fewer parameters and 94.58% fewer GFLOPs.
  • Testing on a real edge device with an Arm Mali-G52 MC2 GPU showed LightMIS running around 10.5x faster than nnWNet with 100% GPU operator coverage and no CPU fallback.
  • Experiments covered six datasets and five imaging modalities, with all models trained and evaluated under the same nnU-Net v2.3.1 protocol.
  • Source code, checkpoints, and reproduction details are publicly available on GitHub and arXiv.
open-sourcesegmentationedge-computingmedical-imaginglightweight-models
Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.

Related

GeekNewsHerdr releases native desktop app for managing terminals and coding agentsRedditOpen-source research demonstrates sparse attention for Qwen 3.8 27B with 10M token contextRedditQwen Head Confirms Qwen 4.0 and 5.0 Will Remain Open SourcePyTorchKRAnySpeechRecognizer: Swift package supporting open models compatible with Apple Speech APIPyTorchKRKOMOS Achieves SRCC of 0.900 in Korean Speech MOS PredictionGeekNewsOpenQodex: Open-Source AI Code Review Tool Verifies Agent Changes with 98% Issue Recall