Research centers are incessantly developing new algorithms of artificial intelligence (AI) that are oriented at supporting medicine.It is already possible to diagnose pneumonia this way, among others.
Artificial Intelligence (AI) stands ready to develop the medical sector by improving the effectiveness and accuracy of diagnosis with regard to specialties relying on images.This is predominantly about radiology and pathology.Yet as technology progresses, experts are facing potential downsides of such solutions.
Albert Hsiao, a radiologist at the University of California, San Diego, who has developed algorithms for reading cardiac images, says that in the course of his AI-supported work he can see a lot of risks that may bring about problems.The researcher describes one of the most serious problems as trivial, as most AI software had been developed and subject to testing in a single specific hospital which, in turn, adversely affects the functioning of a given algorithm after it has been placed in another.
Last month, American scientists and doctors published a scheme in the Journal of American College of Radiology. It described how to transform AI based on research into new and improved medical imaging.The authors insisted that there should be closer interdisciplinary cooperation in building and testing AI algorithms, as well as intensive validation of these before they reach patients, among other things.
Algorithms learn like scientists.They are "fed" with hundreds of thousands of images (e.g. mammograms), getting trained to recognise patterns in a faster and more accurate way than humans.
Krishna Kandarpa, an interventional radiologist at the National Institute of Biomedical Imaging in Bethseda, Maryland, says that while he is doing an MRI of a heart that is moving, the computer may predict the moment the heart is going to beat and, thus, capture a better image than a blurry one.
AI can also analyse computer tomography scans of patients with suspected brain strokes, identify those being at a greater risk for cerebral bleeding and label them as priority ones so that the radiologist can subject them to examinations prior to others.An algorithm can also be capable of detecting barely visible breast tumours during mammography or pneumonia - in the case of a radiograph.
Despite the fact that technology is developing, most patients are currently of the same opinion: I would like a human doctor, regardless of all.The machine can accompany them.