The increasing complexity and volume of health data require sophisticated computational approaches for processing and extracting meaningful insights. This paper presents a systematic review of bio-inspired algorithms applied to feature selection in multimodal machine learning contexts specific to healthcare applications. We analyze the main bio-inspired algorithms used in this domain, including Genetic Algorithms (GA), Salp Swarm Algorithm (SSA), and Bacterial Foraging Optimization (BFO), and the evaluation metrics employed to measure their performance, such as precision, sensitivity, and area under the curve (AUC). We identify significant challenges faced by researchers, including limited samples combined with high-dimensional feature spaces, incompatibility between features from different modalities, noise and uncertainty in the data, and the need for interpretability of results in clinical contexts. Our analysis shows that, although bio-inspired algorithms demonstrate considerable potential for feature selection in multimodal health data, their effective implementation requires balancing algorithmic performance with practical feasibility, considering computational limitations and interpretability requirements specific to the healthcare domain.