Improved diagnosis by automated macro- and micro-anatomical region mapping of skin photographs.

BACKGROUND: The exact location of skin lesions is key in clinical dermatology. On one hand, it supports differential diagnosis (DD) since most skin conditions have specific predilection sites. On the other hand, location matters for dermatosurgical interventions. In practice, lesion evaluation is not well standardized and anatomical descriptions vary or lack altogether. Automated determination of anatomical location could benefit both situations. OBJECTIVE: Establish an automated method to determine anatomical regions in clinical patient pictures and evaluate the gain in DD performance of a deep learning model (DLM) when trained with lesion locations and images. METHODS: Retrospective study based on three datasets: macro-anatomy for the main body regions with 6000 patient pictures partially labelled by a student, micro-anatomy for the ear region with 182 pictures labelled by a student and DD with 3347 pictures of 16 diseases determined by dermatologists in clinical settings. For each dataset, a DLM was trained and evaluated on an independent test set. The primary outcome measures were the precision and sensitivity with 95% CI. For DD, we compared the performance of a DLM trained with lesion pictures only with a DLM trained with both pictures and locations. RESULTS: The average precision and sensitivity were 85% (CI 84-86), 84% (CI 83-85) for macro-anatomy, 81% (CI 80-83), 80% (CI 77-83) for micro-anatomy and 82% (CI 78-85), 81% (CI 77-84) for DD. We observed an improvement in DD performance of 6% (McNemar test P-value 0.0009) for both average precision and sensitivity when training with both lesion pictures and locations. CONCLUSION: Including location can be beneficial for DD DLM performance. The proposed method can generate body region maps from patient pictures and even reach surgery relevant anatomical precision, e.g. the ear region. Our method enables automated search of large clinical databases and make targeted anatomical image retrieval possible.

Observer-independent assessment of psoriasis-affected area using machine learning.

BACKGROUND: Assessment of psoriasis severity is strongly observer-dependent, and objective assessment tools are largely missing. The increasing number of patients receiving highly expensive therapies that are reimbursed only for moderate-to-severe psoriasis motivates the development of higher quality assessment tools. OBJECTIVE: To establish an accurate and objective psoriasis assessment method based on segmenting images by machine learning technology. METHODS: In this retrospective, non-interventional, single-centred, interdisciplinary study of diagnostic accuracy, 259 standardized photographs of Caucasian patients were assessed and typical psoriatic lesions were labelled. Two hundred and three of those were used to train and validate an assessment algorithm which was then tested on the remaining 56 photographs. The results of the algorithm assessment were compared with manually marked area, as well as with the affected area determined by trained dermatologists. RESULTS: Algorithm assessment achieved accuracy of more than 90% in 77% of the images and differed on average 5.9% from manually marked areas. The difference between algorithm-predicted and photograph-based estimated areas by physicians was 8.1% on average. CONCLUSION: The study shows the potential of the evaluated technology. In contrast to the Psoriasis Area and Severity Index (PASI), it allows for objective evaluation and should therefore be developed further as an alternative method to human assessment.