The principles of evidence-based medicine (EBM), originally defined as the honest, clear, and fair use of the best available evidence in making decisions about individual patient care, have shaped modern medicine and health care.
Indeed, despite the great success of EBM over the years, there remains a large gap between theory and practice, as evidenced by various issues and concerns about the quality, efficiency, and cost of current clinical practice.
Input and output data from the same domain
This is probably the simplest situation for DL since there is no change in domain between input and output data. Many of AIM’s functions fall into this category. Indeed, the list of DL applications in this category can be quite extensive and new types of applications continue to appear in the literature.
To name a few, we provide bidirectional transformers for super-resolution imaging, super-resolution dose calculation, microcopy image processing and denoising, genomic data simulation, image in painting, image registration, and natural language processing (NLP).
Image quality in medical imaging is a trade-off between imaging time, spatial resolution, temporal resolution, and patient dose or photodamage when x-rays or optical photons are involved.
In the present practice, each imaging event is performed independently. DL provides an effective way to incorporate information obtained from previous imaging studies to obtain better images.
Empirical evidence or measurements
Most clinical decision-making models belong to this category. In practice, AI and DL are particularly useful in simplifying a variety of decision-making processes by learning complex relationships and incorporating existing knowledge into an inference model.
Machine learning models are trained with measurement data defined by experienced professionals like doctors.
Due to its ability to learn from big data in different domains and integrate knowledge from different sources and diverse fields, AI models can potentially surpass any human understanding and provide healthcare with significantly better decisions.
Applications beyond traditional indications
Some healthcare applications of AI have been highlighted in previous chapters. The application of AI certainly goes beyond that. In practice, although some are still in their early stages, AI is applied to a variety of clinical decision-making tasks in almost all areas of medicine.