Automated Measurement of Pressure Injury/Ulcer through Image Processing

Objective and accurate assessment of pressure injury/ulcer (PI/U) healing is needed to deliver better wound care to patients. Progress in wound healing is primarily quantified by the rate of change of PI/U’s dimensions. However, accurate measurement of PI/U dimension is challenging due to the complexities of PI/U itself and clinical environment. Photographing PIs/Us has become a standard practice in nursing documentation at many hospitals today. With special software and tool’s help, PI/U dimension can be estimated from PI/U images. Performed manually, this process is very time-consuming and subject to intra- and inter-reader variability. In this paper, we present our methodology to segment and measure the enclosed PI/U area from photographic PI/U images at clinical settings automatically. The first step of our method is to transform the images with RGB (Red, Green, and Blue) color space to YCbCr color space which help us eliminate the inferences from light and skin colors. A probability map, generated by skin color Gaussian model, guides the PI/U segmentation process using Support Vector Machine classifier. After PI/U is segmented from the images, the reference ruler helps complete perspective transformation and determine the size of PI/U. A total 32 PI/U images measured by WOCNs was used to validate the PI/U measurement from the image processing technologies. The results showed that intra-rater reliability of the measurements of length, width and surface area were all 0.89. The inter-rater reliability of length, width, and surface area were 0.89, 0.87 and 0.88, respectively. The innovative aspects of this work include defining a probability map specific to healthy skin and PI/U characteristics, a computationally efficient method to segment PI/U images utilizing the probability map, and computerized PI/U measurements with consideration of perspective transformation. In addition, a high accuracy on PI/U measurement was achieved by our method through comparisons with WOCNs’ measurement.