På svenska

Presentation Information     2017-08-28 (14:15)   •  4307

Speaker Jian FransÚn  (CBA)
Comment Department of Surgical Sciences, plastic surgery Uppsala University
Type External presentation
Title Machine learning and computer vision in diagnostics: Applications to burn wounds and skin lesions assessment
Abstract Background: Burn care can be costly and burn experts are sparse especially in rural areas and the disease often requires prompt decision-making. Skin lesions and cancer are a public health issue relying on self reporting and early detection. Automated diagnosis and predictions of skin images using advanced algorithms through carriable mobile phone apps might disseminate expert advice to the public and point-of-care diagnosis. Aim: To study the use of artificial intelligence (AI) and imaging for diagnostic support for burn wounds and skin lesions. Methods: Data will be gathered from health records in Sweden and South Africa. Algorithms will be trained using one set of training data with the outcomes, validated/optimized using a second set and then tested on a third independent set until the algorithms reaches aspired accuracy for diagnosis. Preliminary results: Ethical permission has been granted for image gathering from South Africa and Uppsala University Hospital in Sweden. Pictures has been gathered. Specific machine learning algorithms and assisting software has been selected. Initial training of algorithm has begun.