English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Deep learning techniques have shown great potential in medical imageprocessing, particularly through accurate and reliable image segmentation onmagnetic resonance imaging (MRI) scans or computed tomography (CT) scans, whichallow the localization and diagnosis of lesions. However, training thesesegmentation models requires a large number of manually annotated pixel-levellabels, which are time-consuming and labor-intensive, in contrast toimage-level labels that are easier to obtain. It is imperative to resolve thisproblem through weakly-supervised semantic segmentation models usingimage-level labels as supervision since it can significantly reduce humanannotation efforts. Most of the advanced solutions exploit class activationmapping (CAM). However, the original CAMs rarely capture the precise boundariesof lesions. In this study, we propose the strategy of multi-scale inference torefine CAMs by reducing the detail loss in single-scale reasoning. Forsegmentation, we develop a novel model named Mixed-UNet, which has two parallelbranches in the decoding phase. The results can be obtained after fusing theextracted features from two branches. We evaluate the designed Mixed-UNetagainst several prevalent deep learning-based segmentation approaches on ourdataset collected from the local hospital and public datasets. The validationresults demonstrate that our model surpasses available methods under the samesupervision level in the segmentation of various lesions from brain imaging.

Mixed-UNet: Refined Class Activation Mapping for Weakly-Supervised Semantic Segmentation with Multi-scale Inference