Is AI Really Helping Doctors?

The incorporation of artificial intelligence (AI) into medical image interpretation has shown great potential for improving diagnostic accuracy and efficiency. Exploiting the advantages of artificial intelligence while taking advantage of the expertise of clinicians will bring profit in terms of patient care… Thus, for several months, for example, various Public Aid institutions – Paris hospitals are equipped with automatic image software analysis. Does the latter indicate a bone fracture or a pulmonary nodule? The human arrives afterwards to confirm, deny or correct the software analysis which sometimes does not highlight the expected image.

In order to optimize the application of AI in clinical practice, it is essential to have a comprehensive understanding of the impact of AI assistance on physicians. “Clinicians have different levels of expertise, experience and decision-making stylessays Feiyang Yu, from the computer science department at Stanford University, in the United States. Ensuring that AI support accommodates this heterogeneity is critical for targeted implementation and maximizing positive impact on patient care. » His team conducted a study 1the results of which have just been published. The experiment examines the performance of 140 radiologists with and without AI assistance in 15 diagnostic tasks ranging from fractures and pulmonary edema to cardiac pathologies and infections. Doctors had to examine 324 “cases”.

Predictive capabilities?

“Participating radiologists were trained in the AI ​​assistance system before starting the experiment. They also saw examples of this AI's predictions, which will help calibrate their interpretation and inform the use of AI in observation.” continues Feiyang Yu. To assess the impact of AI on doctors' ability to identify and correctly diagnose problems, researchers used advanced computational techniques to measure changes in performance when AI was used or not in practitioners' practice. The results revealed a variety of effects of AI assistance, with improvements in some radiologists and degradation in others. Results showing the inconsistent overall effect of using the tool.

The effects of artificial intelligence on the skills of human radiologists have often shown surprising variability. “Contrary to our expectations, professional experience, expertise in thoracic radiology, or prior use of AI tools by a radiologist do not reliably predict the impact of an AI tool on performance2 medicine » the researchers explain. Initially, lower performing clinicians did not benefit from AI assistance. “Some benefited, some less and some not at all. So overall, radiologists have not advanced their performance, whether they use artificial intelligence or not.”, says Feiyang Yu. A trend that was also true among those who performed better initially. Their good performance was maintained, with or without AI help.

Although the researchers were unable to determine the exact reasons for these results, they highlight the critical importance of testing and validating the performance of AI tools before their clinical deployment. Trials that could ensure that less capable AI does not hinder the skills of human clinicians and, by extension, the quality of patient care. “Our findings highlight the inadequacy of a one-size-fits-all approach to AI assistance and highlight the importance of individualized strategies to maximize benefits and minimize errors. » In conclusion, the research team underlines that in addition to improving the accuracy of AI tools, it is important to train radiologists to identify false predictions and challenge the diagnoses provided by these tools. The next step is that of reasoning… the developers of these new tools must ensure that they design models capable of explaining their decisions. At this price artificial intelligence will be able to take a new step in medical assistance and effectively guide professionals.

The Smart Ultrasound in Obstetrics and Gynecology (Suog) project aims to use artificial intelligence in all its forms to improve ultrasound pregnancy monitoring. It is a helpful tool that provides intelligent, iterative guidance when an operator notices an unusual appearance in the fetus. This tool proposes, in real time and during the ultrasound, to obtain the ultrasound images to locate the necessary points for the diagnosis. These images then make it possible to communicate with the specialist with all the necessary information to coordinate care: organizing a check-up by the specialist, possibility of tele-expertise, referral to a center specializing in fetal medicine, adjustment of deadlines. Suog incorporates a complex form of artificial intelligence, reasoning on a knowledge base and combining image recognition.

Source: Health Sorbonne University

The Smart Ultrasound in Obstetrics and Gynecology (Suog) project aims to use artificial intelligence in all its forms to improve ultrasound pregnancy monitoring. It is a helpful tool that provides intelligent, iterative guidance when an operator notices an unusual appearance in the fetus. This tool proposes, in real time and during the ultrasound, to obtain the ultrasound images to locate the necessary points for the diagnosis. These images then make it possible to communicate with the specialist with all the necessary information to coordinate care: organizing a check-up by the specialist, possibility of tele-expertise, referral to a center specializing in fetal medicine, adjustment of deadlines. Suog incorporates a complex form of artificial intelligence, reasoning on a knowledge base and combining image recognition.

Source: Health Sorbonne University

The Smart Ultrasound in Obstetrics and Gynecology (Suog) project aims to use artificial intelligence in all its forms to improve ultrasound pregnancy monitoring. It is a helpful tool that provides intelligent, iterative guidance when an operator notices an unusual appearance in the fetus. This tool suggests, in real time and during the ultrasound, the ultrasound images to be taken to locate the necessary points for diagnosis. These images then make it possible to communicate with the specialist with all the necessary information to coordinate care: organizing a check-up by the specialist, possibility of tele-expertise, referral to a center specializing in fetal medicine, adjustment of deadlines. Suog incorporates a complex form of artificial intelligence, reasoning on a knowledge base and combining image recognition.

Source: Health Sorbonne University

Leave a Reply

Your email address will not be published. Required fields are marked *