Despite the fact that the use of artificial intelligence in image diagnostics is still quite incipient in Spain, compared to similar countries, AI is the star theme of the congress of the Spanish Society of Medical Radiology (Seram), which will end today in Barcelona with the participation of around 2,400 professionals. Radiologists had not met in the Catalan capital for 20 years. No one imagined then that algorithms should contribute to diagnosis. All professionals are now convinced that AI will be universally used in different fields of medical care in the coming years, and especially in radiology.
The revolution has taken the professionals by no surprise. “We are a clinical specialty, but also a technological one,” argues Josep Munuera, Seram’s scientific manager and director of radiodiagnosis at Sant Pau Hospital (Barcelona). “It is impossible not to talk about AI. We are the most advanced medical specialty in the incorporation of AI in health and we have seen it in most of the tables and meetings these days”, he adds.
80% of medical diagnoses are carried out through an image. AI means applying algorithms and learning techniques to perform the test, interpret radiological images and acquire the images with the highest quality in less time. Finally, the algorithms help identify the pathology and make the differential diagnosis. “We will demonstrate the value of AI in clinical practice by shortening the time of diagnosis or being more precise when classifying diseases”, says Dr. Munuera.
The main current applications of AI include bone fractures, especially in emergency department x-rays; bone and soft tissue tumors on X-rays and CT and MRI images; help in the early detection and follow-up of arthritis and joint diseases, or in the assessment of the situation of prostheses. Algorithms are even used to quantify body fat for cardiovascular risk factors. All this in the interpretive intelligence section, based on algorithms that facilitate the reading of the tests. On the other hand, “non-interpretive intelligence has a great impact on the patient’s experience”, remarks Munuera: “Algorithms have been introduced that reduce the duration of a resonance from 40 to 15 or 20 minutes standard, maintaining or improving the quality of the image, with what this means for a patient who has to be inside the machine”.
There is a different algorithm for each pathology. The company Incepto, which was at the congress, has developed 25. This French-based technology deals with breast cancer screening in Denmark, lung tumor screening for the UK NHS and its solutions are in hundreds of European hospitals. Alfonso Martínez, general manager of the firm in Spain and Portugal, predicts that in the next 5 or 10 years, diagnosis with AI will be used massively because there is no doubt about its efficiency: “All Spanish hospitals will use it”.
In his opinion, the lack of radiologists and the benefits of algorithms, together with the needs of patients, with increasingly advanced ages, are the factors that will trigger this technology. Incepto trains the AI with thousands of images, more than a million in some of the algorithms. Before commercialization, CE marking is required after demonstration through clinical trials of the benefit presented by each algorithm. “In terms of breast cancer screening, we have just published a report that demonstrates a 20% improvement in diagnosis and a 40% improvement in productivity.”
