In a study published in the journal Journal of Medical Internet Research, scientists present an AI capable of predicting stroke and facilitating diagnosis. Diagnostic errors represent a major public health problem and contribute to preventable patient harm and excessive health spending, and it was precisely to overcome this obstacle that the team conducted the research.
It is also worth understanding that stroke can be difficult to diagnose, as patients do not always have classic symptoms and other conditions can mimic it. In addition, they may have atypical symptoms. Approximately 25% of stroke patients do not have the usual problems with speech, facial drooping, and limb weakness.
“Machine learning methods have been used to help detect stroke by interpreting detailed data such as clinical notes and diagnostic imaging results. But this information may not be readily available when patients participate in triage in hospital emergency departments,” say the researchers.
To develop their stroke prediction algorithm, the scientists used more than 143,000 individual patient records from Florida critical care hospital admissions between 2012 and 2014. The model they created was able to predict with 84% accuracy.
However, the researchers themselves point out that their algorithm is not intended to be an autonomous model; should be used in conjunction with existing stroke diagnostic models.
Source: Journal of Medical Internet Research via New Atlas
