Computational Quantification of Collagen Density and Ki67-Positive Cells in a Forensic Porcine Wound Model

Obtaining an accurate age of skin wounds is a diagnostic challenge in forensic pathology. This study aimed to quantify collagen density and proliferation activity in porcine experimental wounds over time using machine learning-based segmentation to provide an objective method for wound age estimation. Tissue sections from porcine experimental wounds (n = 68) were stained with Masson's trichrome stain or immunohistochemically labeled for proliferation activity by Ki67. The experimental wounds were located on the back and 5-35 days old. Collagen and proliferation activity in the wounds were quantified by training and application of neural network pixel and random trees object classifiers. The relative collagen fraction and the collagen ratio between lower and upper wound regions displayed significant time-dependent patterns. The proliferative activity, assessed by the percentage of Ki67-positive cells, was not suitable for age assessment. In conclusion, the application of a neural network pixel classifier trained to differentiate between collagen and cellular components is an objective method for forensic wound age assessment. However, to obtain a higher precision, the method should be used as a supportive tool in combination with other time-dependent markers.