Near-infrared optical spectroscopy for pancreas shrinkage estimation with multi synchrosqueezing transform and multivariate regression model.

Abstract

In this article, we proposed a method to estimate pancreas shrinkage with pancreas β cell insulin secretion. The β cells in the pancreas secrete insulin and digestive enzymes after food consumption. Conventionally, the pancreas structure estimation is done with magnetic resonance imaging (MRI) and ultrasound imaging techniques. However, the structure of the pancreas changes due to islet cell death. The presence of islet cells is detected through near infrared (NIR) spectroscopy signal acquired from the epigastric region (pancreas) of the abdomen. Subsequently, the NIR spectroscopy signal from the pancreas is analyzed with multi synchrosqueezing transform (MSST); whereas, the β cell insulin secretion varies for diabetic and nondiabetic persons. The existence of β cell and insulin secretion correlates with Root Mean Square (RMS) and kurtosis via a multivariate regression model to evaluate pancreas shrinkage. In terms of numerical results, NIR spectroscopy signal from the pancreas was obtained for about 20 nondiabetic and 20 diabetic persons. The pancreas shrinkage was estimated with 88% accuracy. The results are validated with MRI pancreas images for earlier detection of the apoptotic pancreas. The pancreas shrinkage causes lower insulin emission and unpredictable blood glucose in diabetic patients. Analysis of NIR spectroscopy signals of the pancreas with MSST was done to obtain higher-order and lower-order frequency components.