The place of AI in radiology today

The place of AI in radiology today

By Mohamed Akrout

The arrival of AI in the field of medicine is announced as a revolution, an upheaval of practices that will have a tremendous impact on drug development, wearable devices, and radiology. The latter field already uses many computer tools for image processing, and while they are certainly very powerful, they are not designed to be intelligent. These tools include computer-aided detection/diagnosis (CAD), which automatically analyzes images for the detection of pulmonary nodules, microcalcifications in mammography, brain lesions, and a variety of other abnormalities.

The rapid progress of AI is being used in radiology to outperform the CAD approach. In this article, we present different use cases where AI has been applied successfully and discuss how the radiologist of the future might utilize AI.

Automated interpretation of examinations:

The first use of AI in radiology is the automated interpretation of exams. By drawing on text and other data in a patient’s medical record,  possible treatments can be suggested. Avicenna, an AI-powered software developed by IBM, uses structured data and electronic health records (i.e., disease history) from predefined content in images. This technology was made possible thanks to the rapid progress of the natural language processing (NLP) domain. By integrating the patient’s antecedents, the history of their illness, the clinical setting, biological parameters, data from other paraclinical examinations, and comparing with previous examinations of the patient, the AI technology can reason like a radiologist.

Robots and chatbots in interventional radiology:

In interventional radiology, robots and chatbots have been developed to facilitate the role of the radiologist and provide real-time information to the patient about the next phase of treatment. By using the same technology found in self-driving cars, researchers and radiologists at the University of California at Los Angeles (UCLA) are using AI-powered chatbot applications to quickly get answers to frequently asked questions. This technology uses IBM’s Watson AI, which can answer questions posed in natural language and perform other machine learning functions.

Do better than a radiologist:

AI can also outperform a radiologist by using deep learning, a subdomain of AI, to make predictions based upon the results of diagnostic experiments that have accumulated in massive databases. In December 2017, Stanford researchers published an AI model called CheXNet which detects pneumonia from chest X-rays at a level exceeding practicing radiologists. After being trained using 100,000 frontal view X-ray images of 14 diseases, the AI model distinguished more than 200 levels of gray (versus the 16 to 20 that the human eye can) which allowed the AI to see anomalies that would be invisible to the human eye. The final accuracy of this model outperformed the average radiologist’s performance on the pneumonia detection task.

Will AI take over radiology?

These technological advances are currently changing radiology, but AI will not replace radiologists entirely. Radiologists have a history of embracing technological innovations and successfully integrating them into their workflows. AI raises many hopes for improvements in diagnostics that could be made in the years to come. However, many radiologists fear that its use will diminish their role, pushing them into the background during the diagnostic process.

In May 2017, at the GPU Tech Conference in San Jose, to the question of whether AI could one day replace radiologists, Dr. Curtis Langlotz, Professor of Radiology at Stanford University, provided a clear answer:

“No. But radiologists who use artificial intelligence will replace those who do not use it.”

Automation in radiology enabled by AI can help to limit errors and allow repetitive tasks to be performed more easily so that radiologists can focus on diagnostic activities and improve the quality of their relationships with their patients by spending more time with them face-to-face during consultations.


If you have any questions or would like to know if we can help you with your innovation challenge, please contact our Life Sciences lead, Jeremy Schmerer at jschmerer@prescouter.com.

More of Our Insights & Work