Abstract : Reinforcement learning has been applied in situations where an appropriate algorithm is lacking to address an issue. It highlights how a person learns via interactions with their surroundings. Reinforcement learning has been applied in the machine learning field to handle a wide range of challenging problems that are typically regarded as highly cognitive. The goal of this work is to demonstrate how well the reinforcement learning approach can identify and categorise cancer from a variety of medical image types, including CT (Computerised Tomography), MRI (Magnetic Resonance Imaging), USG (Ultra Sound SonoGraphy), and others. These days, a wide range of results in learning policies across numerous domains can be attributed to the combination of reinforcement learning and neural networks. It has made it possible to complete a task with complete impartiality by doing away with human interpretation and prejudice. In this article, we've concentrated on the state of reinforcement learning algorithms as they apply to a variety of domains, including gaming, robotics, skin, organ, and lesion detection, as well as the identification of cancer in different organs.  This review study has addressed the essential features and theoretical perspective of the present algorithms, as well as the primary concerns that limit the uses of reinforcement learning algorithms in the health sector, particularly in the area of cancer diagnosis.  Our aim is to investigate a select few current cancer detection approach algorithms.