The Internet of Medical Things (IoMT) has revolutionized modern healthcare by enabling the collection and analysis of real-time data. However, this interconnected ecosystem also introduces significant security risks, particularly malware attacks that compromise patient safety and data privacy. Traditional security measures are often insufficient because of resource constraints and the real-time operational demands of IoMT devices. This research proposes an optimized hybrid machine learning framework that integrates convolutional neural networks (CNN), long short-term memory (LSTM), random forest (RF), and principal component analysis (PCA) to enhance malware detection in IoMT environments. The proposed method includes an adaptive feature selection mechanism, a resource-efficient architecture, and an ensemble learning model with machine learning capabilities. Validation through experimentation using the CIC-MalMem-2022 dataset, which comprises labeled memory dumps from benign and various malware processes, demonstrated that the proposed framework outperformed current hybrid models while reducing computational costs, achieving a detection accuracy of 99.59%. This study presents a scalable and efficient security solution designed to address the constraints of IoMT devices, addressing critical challenges in healthcare cybersecurity. Copyright (c) 2025 IIUM Press. This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.