Big Self-Supervised Models Advance Medical Image Classification

January 2021 in “ arXiv (Cornell University) ”

Shekoofeh Azizi, Basil Mustafa, Fiona Ryan, Zachary Beaver, Jan Freyberg, Jonathan Deaton, Aaron Loh, Alan Karthikesalingam, Simon Kornblith, Ting Chen, Vivek Natarajan, Mohammad Norouzi

Preprint — not peer reviewed. This was posted to a preprint server or data repository. It has not been through a journal's review process, and its findings may change or not hold up.

Studysummary This study found that self-supervised pretraining significantly improves accuracy in medical image classifiers for dermatology and chest X-ray tasks, outperforming supervised baselines and showing robustness to distribution shifts with limited labeled data.

Automatically generated from the study's abstract, not written by a person, and not a review of the full paper. Not medical advice or a treatment recommendation. Read the original study, and consult a qualified healthcare professional before changing treatment. Full disclaimer ·

View this study on arxiv.org →

Discuss this study in the Community →

Similar Research

5 / 1000+ results

research Bias Assessment and Data Drift Detection in Medical Image Analysis: A Survey

1 citations , September 2024 in “arXiv (Cornell University)”

This study reviews methods to ensure the reliability of machine learning models in medical imaging, focusing on bias detection, data drift assessment, and accuracy estimation without ground truth labels to enhance integration into clinical settings.

research A Comprehensive Review of Alopecia Areata Detection Using Self-Supervised Auto ML Models: Systematic Insights and Emerging Challenges

September 2025

This study reviewed the use of self-supervised Auto ML models for detecting alopecia areata, finding significant advancements in automated diagnosis but also challenges such as model explainability and data bias, which may guide future AI-driven dermatological diagnostics.

research Advancing Hair Disease Diagnostics: A Deep Learning Approach Using Inception-ResNet v2 for Multi-Class Classification

December 2024

This study developed a high-performance deep learning model using the Inception-ResNet v2 architecture to classify 10 hair disease classes, achieving an accuracy of 94.7% and balanced precision, recall, and F1-scores of 0.94, suggesting reliability for automated dermatology diagnostics.

research Optimizing skin disease diagnosis: harnessing online community data with contrastive learning and clustering techniques

1 citations , February 2024 in “npj digital medicine”

This study developed a deep-learning model using unannotated dermatology images from online forums, achieving 49.64% accuracy in classifying 22 skin diseases and 61.76% accuracy in detecting monkeypox, highlighting the potential of these images for skin disease diagnostics in China.

research AutoMIR: Effective Zero-Shot Medical Information Retrieval without Relevance Labels

January 2025

This study introduced a novel framework called SL-HyDE that significantly improved zero-shot dense retrieval accuracy in medical information retrieval without relying on labeled data.