Liver Tumor Segmentation and Classification Model Using HDFOA-Based Deep Learning Model in Smart 5G Health Monitoring
Arunadevi Thirumalraj, Swathi Baswaraju, Vijilius Helena Raj, S. Stephe · Apple Academic Press eBooks · 2025
Provisioning of health services such as care, monitoring, and remote surgery is being improved thanks to 5G cellular technology. The emergence of connected, smart healthcare is one of the most significant ways that the IoT has changed information technology. One of the most important aspects of this topic is the potential application of IoT technology to enhance healthcare, particularly in providing effective health monitoring to those who may not have immediate access to such systems. The utilization of improved deep learning (DL) algorithm (alg) in IoT liver cancer monitoring is the main topic of this work. It may be especially 52 challenging to identify liver tumors since they often exhibit the same contrast and intensity levels as the neighboring healthy tissues. Additionally, the irregular shapes of tumors may be impacted by the kind and stage of the cancer. Typically, the two basic steps in the segmentation of liver tumors are the identification of the liver region and the segmentation of the cancers inside it. Furthermore, automatically classifying these cancers is a challenging task. This study proposes a novel method that combines DL with CT data to enhance the efficiency of liver cancer identification. The research’s first preprocessing steps, which include bilateral filtering and histogram equalization, are performed on benchmark datasets. The liver is subsequently divided using the grab cut segmentation method. Using the Capsule DL method, cancer areas are retrieved once the liver area has been appropriately segmented. To lessen the complexity brought on by a large number of features, a multi-objective feature selection technique is used, based on the Hybrid Dragonfly Optimization Algorithm (HDFOA). Classification is performed utilizing MobileNetV2 and U-Net. These carefully selected features, which also make use of HDFOA, are then used to enhance the hyperparameter tuning of the Improved DL model. By conducting a thorough experimental analysis, this study demonstrates the usefulness of the suggested model and shows its ability to achieve high classification accuracy (ACC) of 98.97% when compared to other methodologies currently used in the field.