The Internet of Medical Things (IoMT) is network of interconnected medical devices which enable real-time monitoring, diagnosis, and treatment in healthcare. IoMT can fundamentally reshape patient care facilitating remote monitoring and personalized treatments. However, heterogeneous nature of IoMT exposes networks to a number of security issues which have consequences. Consequently, there is a pressing need to shift toward intelligent, adaptive solutions such as machine learning models which can be trained to recognize subtle patterns and anomalies in network traffic. While numerous studies have applied ML techniques to general network security, there remains a notable gap in research focusing specifically on IoMT environments. High accuracy might not necessarily translate to effective threat detection if the model fails to adequately capture the minority class representing attack traffic in skewed datasets like BoT-IoT. To address this present study incorporates rigorous evaluations, including k-fold cross-validation and fold-by evaluation covering not only accuracy but precision, recall, F1-score, and macro-averages of five ML algorithms specifically: Gaussian Naïve Bayes, multi-layer perceptron, decision tree, k-nearest neighbors, and gradient boosting trees. Additionally, we examine these ML models against state of the art to identify future directions.