Medical scan AI that can seek second opinion from other AI developed in Australia

Researchers at Monash University's faculties of Engineering and IT have developed an innovative AI algorithm capable of assessing the annotations or labels of other AI algorithms in medical scans, similar to seeking a second opinion.

The team designed a dual-view AI system, where one component labels medical images while the other evaluates the quality of the AI-generated labels by comparing them to those provided by radiologists. To train their AI, the researchers utilised 10% of labelled data from three publicly accessible medical datasets.

As reported in the journal Nature Machine Intelligence, the AI system showcased a significant 3% improvement "compared to the most recent state-of-the-art approach under identical conditions."

Himashi Peiris, the principal researcher and a PhD candidate from the Faculty of Engineering, remarked, "It demonstrates remarkable performance even with limited annotations, unlike algorithms that rely on large volumes of annotated data."

The primary objective of this research was to overcome the scarcity of human-annotated medical images by using a competitive learning approach against unlabeled data. Traditional manual labelling of medical scans is time-consuming, prone to errors, and subject to individual interpretations, leading to extended waiting periods for patients seeking treatment. Moreover, large-scale annotated medical image datasets are often limited due to the substantial time, effort, and expertise required for manual annotation.

The Monash algorithm enables multiple AI models to leverage the strengths of both labelled and unlabeled data, learning from each other's predictions to enhance overall accuracy. This approach facilitates more informed decisions, validation of initial assessments, and the discovery of more accurate diagnoses and treatment options.

Currently, the researchers are working on expanding their AI system to accommodate different types of medical images and developing an end-to-end product dedicated to medical practices.

In the healthcare sector, AI plays a vital role in supporting clinician decisions and complementing clinical diagnoses. IBM's Watson is a popular example, utilising various AI technologies to analyse information and offer personalised treatment insights and recommendations. Watson has been commercially applied in genomics, drug discovery, healthcare management, and oncology, showcasing the incredible potential of AI in revolutionising healthcare practices.

Click here to read the original news story.