Accelerated Cancer Prediction in WBAN with Markov Transition Field and U-net+ Collaboration

Background-In the field of cancer prediction and surveillance, the combination of Wireless Body Area Networks (WBANs) and advanced machine learning techniques has shown significant promise recently. Comprising an array of wearable sensors, WBANs continuously collect physiological and activity data from patients in a non-invasive manner. However, there is a significant issue in converting these univariate time sequence data into a format that is simpler to comprehend and examine. Methods-To address the issue of transforming univariate to bivariate time sequence data, we built new model termed as MTF-UX model. We apply the Markov Transition Field (MTF) technique to transform the univariate time sequence data collected by WBANs into bivariate time sequence pictures. These MTF-generated images are fed into the subsequent deep learning system. For the predictive modeling, we make use of an improved U-Net architecture known as U-Net+.Results-With additional layers and modified settings, an improved version of U-Net improves both the accuracy and robustness of cancer prediction. The results demonstrate that a dense layer of [256,128,64] and an image size of 256 x 256 produced the greatest results in terms of accuracy, F1 score, AUC, and MCC_UX, which were 98.46%, 94.26%, 95.00%, and 96.58%, correspondingly.Concluding Remarks-The Markov Transition Field (MTF) approach, enhanced U-Net architecture (U-Net+), and WBANs are utilized in this work to present a novel model for accurate and early cancer prediction called the "MTF-UX model". WBANs, MTF, and U-Net+ together offer a potent framework attaining excellent prediction accuracy, accurate tracking, and providing insightful information about patient health.