In a new development, researchers have developed an AI-based algorithm that can detect different types of brain injuries. A team of researchers at Imperial College and the University of Cambridge clinically validated and tested the algorithm for large sets of CT scans. The algorithm successfully detects, segments, quantifies, and differentiate different types of brain lesions.
The results of animal testing of the algorithm are published in The Lancet Digital Health, and could be more useful for large-scale research studies. This could be to develop more personalized treatment for head injuries. Also, with further validation, the algorithm could be useful for certain clinical scenarios, such as where radiological expertise involves hefty costs.
CT scan findings often missed, Estimates made to Chart line of Treatment
Meanwhile, worldwide, head injury is a huge public health burden and affects up to 60 million people each year. Besides, it is also the leading cause of mortality among young adults. For example, in the event of a head injury, the patient is sent for CT scan to check for blood clots in and around the brain, and to help find if surgery is required.
“CT scan is a very important tool, but it’s mostly not used quantitatively,” said co-lead author of the study from Cambridge’s Department of Medicine. Also, often, much of the information generated via CT scan is missed. It is researchers who know that the type, volume, and location of a lesion in the brain are important for patient outcomes.
Clinically, different types of blood in or around brain can result into different patient outcomes. And, radiologists often make estimates in order to determine the best line of treatment. Meanwhile, detailed analysis of a CT scan with annotations can take a long time, especially among patients with more severe injuries.