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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Deep learning in medical imaging has the potential to minimize the risk ofdiagnostic errors, reduce radiologist workload, and accelerate diagnosis.Training such deep learning models requires large and accurate datasets, withannotations for all training samples. However, in the medical imaging domain,annotated datasets for specific tasks are often small due to the highcomplexity of annotations, limited access, or the rarity of diseases. Toaddress this challenge, deep learning models can be pre-trained on large imagedatasets without annotations using methods from the field of self-supervisedlearning. After pre-training, small annotated datasets are sufficient tofine-tune the models for a specific task. The most popular self-supervisedpre-training approaches in medical imaging are based on contrastive learning.However, recent studies in natural image processing indicate a strong potentialfor masked autoencoder approaches. Our work compares state-of-the-artcontrastive learning methods with the recently introduced masked autoencoderapproach "SparK" for convolutional neural networks (CNNs) on medical images.Therefore we pre-train on a large unannotated CT image dataset and fine-tune onseveral CT classification tasks. Due to the challenge of obtaining sufficientannotated training data in medical imaging, it is of particular interest toevaluate how the self-supervised pre-training methods perform when fine-tuningon small datasets. By experimenting with gradually reducing the trainingdataset size for fine-tuning, we find that the reduction has different effectsdepending on the type of pre-training chosen. The SparK pre-training method ismore robust to the training dataset size than the contrastive methods. Based onour results, we propose the SparK pre-training for medical imaging tasks withonly small annotated datasets.

Self-Supervised Pre-Training with Contrastive and Masked Autoencoder Methods for Dealing with Small Datasets in Deep Learning for Medical Imaging