Smart IoT with the hybrid evolutionary method and image processing for tumor detection.

The primary objective of modern healthcare systems is to enhance public health by providing efficient, reliable, and well-structured solutions. Improving patient satisfaction through tailored medical services has driven rapid advancements in healthcare, leading to increased competition and system complexity. However, the expansion of healthcare services introduces challenges such as high data volume, latency, response time constraints, and security vulnerabilities. To address these issues, fog computing offers an effective solution by processing data closer to end devices, reducing latency, and enabling real-time responses. This research proposes a robust brain tumor detection framework within a fog-based smart healthcare infrastructure. The process begins with data placement leveraging an improved evolutionary technique for Image Processing (HETS-IP) to optimize fog node placement based on key parameters such as energy efficiency and latency. Specifically, the Particle Swarm Optimization (PSO) algorithm is enhanced with a direct binary encoding technique, in which solutions are represented as binary strings, making it suitable for problems where decisions are discrete. This approach allows efficient optimization in binary decision spaces and improves adaptability for complex placement problems. Once data placement is committed, the tumor detection framework is performed directly at fog nodes to enhance real-time processing. This phase will begin with preprocessing, where a bilateral filter is applied to reduce noise while preserving critical edge details. Next, feature extraction is utilized to derive statistical texture features, which capture diagnostic information essential for distinguishing between tumor types. The process continues by classification using a deep Convolutional Neural Network (CNN) with sequential architecture to classify tumors. Simulation results demonstrate that HETS-IP outperforms traditional evolutionary algorithms, including Ant Colony Optimization (ACO), Genetic Algorithm-Simulated Annealing (GASA), and Genetic Algorithm (GA). On average, HETS-IP reduces energy consumption by 5%, 9%, and 14% and decreases makespan by 4%, 6%, and 11%, respectively. Additionally, the proposed approach achieves an accuracy of 97% and a precision of 96%, ensuring highly reliable brain tumor detection.