The reality of artificial intelligence (AI) in medicine is far more mundane than what is portrayed in Hollywood, where AI is often depicted as taking in an entire medical chart and making a diagnosis. In fact, AI does not diagnose cancer like a human doctor does. An AI takes in a static image and generates a prediction based on mathematical patterns found in the AI’s training data. This prediction is different from a diagnosis.
Humans use a series of standard tests, such as self-exams, mammography, ultrasound, needle biopsy, genetic testing, or surgical biopsy, to generate a diagnosis. Treatment options for cancer include surgery, radiation, chemotherapy, and maintenance drugs. The tests, treatment, and drugs available today in US hospitals are the best available in the history of the world.
When using AI for medical diagnosis, it is necessary to keep the images at high-resolution. A low-resolution image may result in a false negative. Another problem that can impact AI’s ability to accurately analyze medical images is the use of a color image file. While X-ray images appear black and white to the human eye, the computer may represent the image as a full color image, which can affect the AI’s evaluation of the image.
In summary, while AI has potential to improve medical diagnosis and treatment, it is important to have a realistic understanding of its limitations and how it works in practice. AI is built on top of the diagnostic processes that human doctors use, and it is not a replacement for human expertise.