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/

Zendy Plus

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

Zendy Tools

Zendy Open

Convolutional neural networks have enabled significant improvements inmedical image-based diagnosis. It is, however, increasingly clear that thesemodels are susceptible to performance degradation when facing spuriouscorrelations and dataset shift, leading, e.g., to underperformance onunderrepresented patient groups. In this paper, we compare two classificationschemes on the ADNI MRI dataset: a simple logistic regression model usingmanually selected volumetric features, and a convolutional neural networktrained on 3D MRI data. We assess the robustness of the trained models in theface of varying dataset splits, training set sex composition, and stage ofdisease. In contrast to earlier work in other imaging modalities, we do notobserve a clear pattern of improved model performance for the majority group inthe training dataset. Instead, while logistic regression is fully robust todataset composition, we find that CNN performance is generally improved forboth male and female subjects when including more female subjects in thetraining dataset. We hypothesize that this might be due to inherent differencesin the pathology of the two sexes. Moreover, in our analysis, the logisticregression model outperforms the 3D CNN, emphasizing the utility of manualfeature specification based on prior knowledge, and the need for more robustautomatic feature selection.

Feature robustness and sex differences in medical imaging: a case study in MRI-based Alzheimer's disease detection