Researchers at Monash University have developed a novel co-training AI algorithm for medical imaging that emulates the process of seeking a second opinion. The research, recently published in Nature Machine Intelligence, addresses the scarcity of human annotated medical images by utilizing adversarial learning against unlabelled data. This advancement is expected to benefit radiologists and other healthcare professionals involved in medical image analysis. The algorithm works on a “dual-view” AI system, where one part imitates how radiologists label images, while the other evaluates the quality of AI-generated labels against limited human-provided labels. The algorithm demonstrated significant improvements in semi-supervised learning, achieving a 3% average improvement over the current state-of-the-art approach in publicly accessible medical datasets with only 10% labelled data. The researchers plan to expand its application to different medical image types and develop a comprehensive end-to-end product for radiologists’ use.
AI Can Ask for a Second Opinion on Medical Scans
Join the AGEIST movement!
Sign-up for our weekly magazine.
See medical disclaimer below.
The ideas expressed here are solely the opinions of the author and are not researched or verified by AGEIST LLC, or anyone associated with AGEIST LLC. This material should not be construed as medical advice or recommendation, it is for informational use only. We encourage all readers to discuss with your qualified practitioners the relevance of the application of any of these ideas to your life. The recommendations contained herein are not intended to diagnose, treat, cure or prevent any disease. You should always consult your physician or other qualified health provider before starting any new treatment or stopping any treatment that has been prescribed for you by your physician or other qualified health provider. Please call your doctor or 911 immediately if you think you may have a medical or psychiatric emergency.