Few-shot skin lesion classification with Adaptive Multi-Scale Convolutional Attention Network

HuiYing Jin, E Liu, Qin Xu, YuChen Li, XiaoLiang Chen, WeiHua Chen · 2026

Read Few-shot skin lesion classification with Adaptive Multi-Scale Convolutional Attention Network by HuiYing Jin, E Liu, Qin Xu, YuChen Li, XiaoLiang Chen, WeiHua Chen (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0351318.txt.

Description

HuiYing Jin and five colleagues build a skin lesion classifier designed to learn from very few examples. Their 2026 study addresses a real bottleneck in medical imaging. Rare lesions offer few training pictures, boundaries blur, and sizes vary wildly. Their network carries three cooperating modules. One adjusts its viewing window to each lesion's scale. Another blends information across resolutions, weighting the most useful channels. A third uses image gradients to sharpen attention on borders and texture.

Few-shot experiments ran on two public datasets of skin images. The system beat existing baseline models across several metrics. It also generalised promisingly between the datasets. Visual checks confirmed the model genuinely looks at the lesion, not the surrounding skin. Dermatology AI usually starves on rare conditions. A classifier built for scarcity could extend computer-aided diagnosis to the lesions doctors see least and worry about most.

Catalog record

CREATORHuiYing Jin, E Liu, Qin Xu, YuChen Li, XiaoLiang Chen, WeiHua Chen
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE69187
LICENSECC_BY
RIGHTSCC BY
SOURCEPLOS

Source: https://doi.org/10.1371/journal.pone.0351318

Browse the full library · About the Tower of Babel