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

Convolutional Neural Networks (CNNs) are frequently and successfully used inmedical prediction tasks. They are often used in combination with transferlearning, leading to improved performance when training data for the task arescarce. The resulting models are highly complex and typically do not provideany insight into their predictive mechanisms, motivating the field of'explainable' artificial intelligence (XAI). However, previous studies haverarely quantitatively evaluated the 'explanation performance' of XAI methodsagainst ground-truth data, and transfer learning and its influence on objectivemeasures of explanation performance has not been investigated. Here, we proposea benchmark dataset that allows for quantifying explanation performance in arealistic magnetic resonance imaging (MRI) classification task. We employ thisbenchmark to understand the influence of transfer learning on the quality ofexplanations. Experimental results show that popular XAI methods applied to thesame underlying model differ vastly in performance, even when considering onlycorrectly classified examples. We further observe that explanation performancestrongly depends on the task used for pre-training and the number of CNN layerspre-trained. These results hold after correcting for a substantial correlationbetween explanation and classification performance.

Benchmark data to study the influence of pre-training on explanation performance in MR image classification