We aim to help doctors, patients, and hospitals through AI. Hence we are interested in all kinds of medical data such as radiology, pathology, cytology, DNA, ultrasound, optic images, ECG, medical records, etc.

This study is a fundamental study on medical AI. Basically, the AI is a black box, so the reason for the AI's outcome is unknown. But explanation or interpretation for AI results is very important in the medical field.

Explainable AI is an alternative to solve the problem. For example, when AI read the Chest X-Ray image and predict some disease, the heatmap obtained by the class activation map highlights the area affected to AI decision like this. If the highlighted area is correct, a doctor can trust AI decisions. Also, when AI reads the input image as some disease, we can search and show the visually or clinically similar images with the input image. If the detected similar images' finding or disease name is the same disease with AI decision, the decision’s reliability will be increased.

Since poor-quality images can affect the reading and analysis of AI, it is necessary to screen images that are difficult to read in advance. This is data clean-up AI that can calculate a quality score for the medical images using AI, and proposed a quality assurance method based on the quality score.