Software Provided Diagnoses for 32% of Subjects in Trial
Last fall, SafeMinds Shares reported that the pediatric behavioral health company, Cognoa, was working with the Food and Drug Administration (FDA) for clearance on their new software which uses artificial intelligence to assist in diagnosing autism in young children. Last week, the company was given the green light from the FDA to begin marketing their new product. Cognoa’s ASD Diagnosis Aid uses software loaded on a device which uses a machine learning algorithm to collect input from parents or caregivers, video analysts, and health care providers in order to assist physicians evaluate a patient’s risk of autism. The device consists of three main components: a mobile app for parents to answer questions about behavioral problems and to upload video footage of their child; a video analysis portal which allows specialists to view and assess uploaded videos of patients; and a healthcare provider portal that tracks the information given by parents, video analysts, and other health care providers. After the information from the various components has been assessed, the ASD Diagnosis Aid reports either a positive or negative ASD diagnosis. The FDA evaluated the safety and effectiveness of this product through a study of 425 patients aged 18 months to 5 years. The average age of the subjects was 2.8 years. The study compared assessments made by the device against assessments made by a panel of clinical experts. The device results matched the panel’s conclusions for 81% of patients who tested positive for ASD and 98% for patients who tested negative. Additionally, the device made an accurate ASD determination in 98.4% of patients with the disorder and in 78.9% without the disorder. In the end, 32% of patients who participated in the trial received an autism diagnosis. Cognoa makes clear that their ASD Diagnosis Aid is a tool and should never be a replacement for a clinician. The company plans to begin marketing this software under the name Canvas Dx later this year.
