Deep Reinforcement Learning (DRL), a paradigm inspired by the principles of Reinforcement Learning (RL) and deep neural networks (DNNs), has emerged as a transformative approach in medical science, addressing challenges in diagnosis, treatment planning, and personalized care. By leveraging the ability of DRL to learn from dynamic, high-dimensional, and sparse data, researchers have developed innovative solutions for areas such as cancer detection, drug discovery, robotic surgery, and chronic disease management. However, the implementation of DRL in medical science is not without challenges. Issues such as the scarcity of annotated datasets, ethical concerns regarding decision-making transparency, and the computational demands of DRL algorithms underscore the need for rigorous evaluation and optimization. Furthermore, the interpretability of DRL models remains a critical bottleneck, particularly in medical domains where human oversight and accountability are paramount. This research study comprehensively analyzes the state-of-the-art DRL methodologies, highlighting their efficacy and limitations in solving real-world medical problems. Additionally, it discusses the integration of DRL with complementary techniques such as transfer learning, multi-agent systems, and explainable AI to enhance reliability and generalizability. By critically examining current applications and identifying research gaps, this work aims to provide a foundation for the next generation of AI-driven medical solutions, ultimately paving the way for more precise, efficient, and equitable healthcare.