The analysis of digital images using computer algorithms to mine digital data directly is an important and increasingly utilized tool in many different fields of medicine. However, despite the clear value of image analysis to forensics, there is little research and no current practical applications of this technology to forensic pathology.
This paper demonstrates the technique of Spatially Invariant Vector Quantization (SIVQ) in the analysis of gross forensic images. SIVQ is a pattern recognition algorithm that can run on a laptop and be utilized by a forensic pathologist with no programming experience. The user merely points and clicks to the feature of interest and SIVQ searches the entire image for the desired feature. The results can be displayed in a variety of formats, both qualitative and quantitative.
Gross forensic images from gunshot wounds, burns and patterned contusions were analyzed by a forensic pathologist using SIVQ. Based on features selected by the user, the algorithm was able to detect both the gunshot wound and its marginal abrasion, calculate the surface area of a gunshot wound defect, distinguish between a thermal burn and the surrounding areas of healing, analyze a patterned contusion, and both identify and calculate the density of gunpowder stippling.
Even though the work is preliminary, it is obvious from these few examples that SIVQ is easily able to identify features of interest in gross forensic images. Additional work needs to be done to fully explore the potential for this application of technology to the practice of forensic pathology.