Explainable artificial intelligence (XAI) in deep learning-based medical image analysis

The Image Sciences Institute in Utrecht, the Netherlands, offers an insightful overview of explainable artificial intelligence (XAI) as applied to deep learning-based medical image analysis. The advent of deep learning has ushered in significant advancements in the realm of automated image analysis, a task traditionally reliant on human. In the context of medical decision-making, where the stakes are often exceptionally high, concerns among medical experts regarding the nature of deep learning are entirely understandable.
One of the prevalent forms of XAI in medical image analysis is visual explanation, commonly referred to as saliency mapping. Saliency maps serve as a valuable tool by highlighting the crucial regions within an image that play a pivotal role in decision-making. Another noteworthy example of XAI is textual explanation, a method that creates textual descriptions. These descriptions can range from relatively straightforward characteristics to comprehensive medical reports.
Given the intrinsic connection between high-stakes decision-making and the field of medicine, it is evident that XAI is poised to become increasingly indispensable. The future trajectory of XAI in medical image analysis may encompass biological explanations. Several researchers have already successfully predicted biological processes based on imaging features utilizing deep learning. Additionally, XAI holds the potential to assist physicians in the diagnostic process and uncover concealed information within medical images, thereby contributing to more informed medical practices.