Researchers from the University of Cambridge and Imperial College London have developed Artificial Intelligence technology that can detect, segment, quantify and differentiate different types of brain lesions.
The researchers developed a machine learning tool based on an artificial neural network, training the tool on more than 600 different CT scans, showing brain lesions of different sizes and types. They then validated the tool on an existing large dataset of CT scans. The AI was able to classify individual parts of each image and tell whether it was abnormal.
PhD student Miguel Monteiro, who is co-first author on the study from Imperial’s Department of Computing, was a key developer of the AI tool. “Head injuries are a major health problem worldwide. Our artificial network and algorithm showed us that we can somewhat automate the diagnosis of different injury types, to potentially help inform treatment in the future.”
This advancement of Artificial Intelligence technology in medicine could be useful in large-scale research studies and for developing more personalized treatments for traumatic head injuries, which see a global incidence of more than 60 million cases per year, and are the leading cause of mortality in young adults.
“Often, much of the rich information available in a CT scan is missed, and as researchers, we know that the type, volume and location of a lesion on the brain are important to patient outcomes,” said co-senior author, Professor David Menon from Cambridge’s Department of Medicine. “We hope [the technology] will help us identify which lesions get larger and progress, and understand why they progress so that we can develop more personalized treatment for patients in future.”
Similarly important to accuracy is the speed of diagnosing injuries. “Detailed assessment of a CT scan with annotations can take hours, especially in patients with more severe injuries,” says co-first author Dr Virginia Newcombe, also from the University of Cambridge, noting that a “tool that could automatically identify and quantify the different types of brain lesions” had the potential to accelerate diagnoses in a clinical setting.
The report, published in The Lancet Digital Health, states that work now needs to focus on the incorporation of such algorithms into clinical practice, in order to help facilitate radiology workflows by flagging scans that require urgent attention, and aiding reporting in resource-limited areas where there is little access to radiological expertise.
