StitchPro for Computational Pathology Stitching in Patients with Prostate Cancer
Ana Sofia Castro Verde, José Guilherme de Almeida, Jorge Fonseca, Celso Matos, Raquel C. Conceição, Nickolas K. Papanikolaou · 2024
A survey by the International Society of Urological Pathology indicates that only a limited number of respondents implement whole-mount histopathology in their clinical routines. Prostate fragments need to be reconstructed into a pseudo-whole-mount histopathology slide using a stitching solution that does not rely on overlap, completeness, and interlocking edges assumptions. Our goal was to develop a methodology to perform automatic stitching of prostate fragments. StitchPro, a Python-based stitching algorithm was developed, consisting of boundary detection, histogram matching, and differential evolution optimization. Good qualitative reconstruction of pseudo-whole mount histopathology slides was completed on the following datasets: AutoStitcher, PROSTATE-MRI, and 35 local radical prostatectomy slides. The quantitative stitching accuracy measured by the average Euclidean distance between corner points was 0.35 mm on 37 reconstructions. The stitching was possible when the fragments were provided with small rotations or in an incorrect orientation. StitchPro may have applications in computational pathology for other tissue types.