AI Screening for Brain Tumours

The new technology presents a novel application for AI within the medical field and could lead to increased clinical trials for diffuse gliomas.

A new AI diagnostic screening system, called DeepGlioma, has shown to rapidly detect and identify diffuse gliomas in under 90 seconds. The technology was developed in an effort to revolutionise diagnostic testing and came about as a result of collaborative efforts between the engineers and neuroscientists at Michigan Medicine. Contributing investigations from other research teams also helped in the development of the DeepGlioma rapid imaging system, which analyses tumour biopsies and diagnoses any genetic mutations which may result in gliomas.

Diffuse Gliomas remain the most common and deadly brain tumours, often advancing rapidly. They can be identified by defined molecular markers, however diagnosis requires complex and lengthy molecular testing, which is unfavourable for diagnosis and can result in sub-optimal care and treatment.

Development of a suitable treatment is also limited by the lack of accurate diagnoses. DeepGlioma uses stimulated Raman histology (an optical imaging technique) combined with deep neural networks to provide real-time images of brain tumour tissue.

The system identified defined molecular markers within a study of 153 patients, differentiating between types of diffuse gliomas without the need for tissue processing. This novel solution for rapid, accurate diagnosis presents the possibility of developing tailored treatments with more accurate patient prognosis in future. This also facilitates increased enrolment in clinical trials, as patient mutations can be more readily defined.