Two new artificial intelligence tools are under investigation that may change the way we think about common, serious medical conditions.
Researchers at McGill University in Montreal have shown that they can use advanced algorithms to detect dementia two years before the onset of symptoms, while a separate and unrelated group of researchers in Ontario have used machine-learning to assess whether or not a skin lesion is a candidate for melanoma biopsy, more accurately than a conventional dermatology exam.
Predicting dementia using hundreds of PET scans It’s well-known that amyloid accumulates in patients with mild cognitive impairment, but that occurs decades before dementia symptoms appear. According to the team at McGill, which published its findings in the journal Neurobiology of Aging, amyloid isn’t a reliable predictive biomarker because not all MCI patients develop Alzheimer’s disease.
The team trained the AI tool using hundreds of amyloid PET scans of MCI patients taken from the Alzheimer’s Disease Neuroimaging Initiative database.
The tool was shown to identify which patients would develop dementia with 84 percent accuracy before symptom onset.
The team is conducting ongoing research to find other biomarkers for dementia that could be incorporated into the algorithm to improve the software’s prediction capabilities. They are also testing the algorithm to validate it in different patient cohorts such as those with small strokes.
"By using this tool, clinical trials could focus only on individuals with a higher likelihood of progressing to dementia within the time frame of the study,” Dr. Serge Gauthier, co-lead author and professor of neurology, neurosurgery and psychiatry at McGill, said in a statement. “This will greatly reduce the cost and the time necessary to conduct these studies.”
Melanoma — sometimes more than meets the eye An unrelated research team based at the University of Waterloo and the Sunnybrook Research Institute developed their own AI technology that evaluates images of skin lesions and provides objective data on the known biomarkers of melanoma.
Dermatologists typically rely on subjective visual exams to decide if a patient requires a biopsy to diagnose the disease. Drawing insight from tens-of-thousands of skin images, the new tool looks for changes in the concentration and distribution of a chemical that gives skin its color and hemoglobin, which are strong indicators of melanoma.
The skin cancer detecting tool could be available for clinical use as early as next year.
