“…The sample sizes ranged from 10 to 2062 patients [121][122][123][124][125][126] and 396 images of PI. 127 Various data science methods have been used to detect or predict PIs, including logistic regression, 121,122 nonlinear regression, 124 a univariate Cox regression, 125 and three data mining algorithms (decision trees, neural networks, and SVMs). 121 The significant predictive factors comprised PI history, 121 without cancer, 121 excretion, 121 activity/mobility, 121 skin condition/circulation, 121 estimated surgery time, 122 serum albumin level, 122,125 multiple ulcers, 125 and presence of a single caregiver.…”