Measurement of Skin Test Wheals Using Image Segmentation Approaches

Medical image segmentation is a widely used approach for skin image analysis ranging from detection of skin wheals, characterization of skin lesions and measurement of tumor-size. The segmentation approaches used in this work focus on the outcome of the tuberculin skin test or Mantaux test. In a skin test, the outcome is a wheal (localized swelling), which usually has a particular shape such as a circle or an ellipse. Automatic wheal detection and measurement is an important step to precisely estimate the wheal dimensions, and thereby estimate the extent and intensity of infection. This work presents an analysis of segmentation approaches on images captured using simple photography for estimating the wheal area and perimeter in terms of number of pixels. Four segmentation approaches are used; namely, Otsu's, region-growing, watershed and region watershed. From the experimental results, it is observed that the average relative error in computing the area and perimeter using the four approaches is found to be 50.77% and 56.35%, 24% and 12.19%, 23.34% and 16.15%, and 30.61% and 20.42% respectively. These automatic wheal detection methods work sufficiently well as compared to the manual selection method. The approach presented in this work minimizes the errors and inconsistencies involved in the reading and interpretation of skin test outcomes; moreover, using computer aided diagnosis would enhance automation of healthcare informatics at clinics and hospitals.