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

The manifold hypothesis is a core mechanism behind the success of deeplearning, so understanding the intrinsic manifold structure of image data iscentral to studying how neural networks learn from the data. Intrinsic datasetmanifolds and their relationship to learning difficulty have recently begun tobe studied for the common domain of natural images, but little such researchhas been attempted for radiological images. We address this here. First, wecompare the intrinsic manifold dimensionality of radiological and naturalimages. We also investigate the relationship between intrinsic dimensionalityand generalization ability over a wide range of datasets. Our analysis showsthat natural image datasets generally have a higher number of intrinsicdimensions than radiological images. However, the relationship betweengeneralization ability and intrinsic dimensionality is much stronger formedical images, which could be explained as radiological images havingintrinsic features that are more difficult to learn. These results give a moreprincipled underpinning for the intuition that radiological images can be morechallenging to apply deep learning to than natural image datasets common tomachine learning research. We believe rather than directly applying modelsdeveloped for natural images to the radiological imaging domain, more careshould be taken to developing architectures and algorithms that are moretailored to the specific characteristics of this domain. The research shown inour paper, demonstrating these characteristics and the differences from naturalimages, is an important first step in this direction.

The Intrinsic Manifolds of Radiological Images and their Role in Deep Learning