“We Don’t Need Big Data, We Need Good Data”
Artificial Intelligence and radiologist expertise might merge in the future, ideally creating better service at lower cost. ECR 18 served as a platform to outline potential steps of this process.
Marc Dewey, Heisenberg Professor at Charité, Berlin, Germany, commented on the worldwide trend in medicine to move away from fee-for-service to pay-for-performance. „This ideally leads to improving outcome at a lower cost,“ he said, which is the main aim of „value-based medicine.“
Dewey defined three major areas in this field, which may hold opportunities for radiology.
1. Personalized Decision-Making
“We need clinical prediction rules to personalize decisions,” said Dewey. He mentioned coronary artery disease calculators as one example. In general, personalized decision-making about the necessity of any diagnostic procedure should be improved. He mentioned the European Society of Radiology’s iGUIDE and eGUIDE that integrate patient history into decision support modules as one step in this direction.
2. The “Bionic” Radiologist
A paradigm shift towards the “bionic radiologist” will be need, said Dewey. This means that artificial intelligence (AI) should be combined with human image analysis. AI, which will be seamlessly integrated in medical processes, will make invisible image data structures visible, said Dewey. The radiologist will supervise the results generated by algorithms, integrate them with other clinical data and be responsible for the final interpretation. Dewey believes this will lead to increased consistency and accuracy in image analysis. He compared the work to the situation of a pilot in an airplane cockpit.
A new and yet unpublished study with 103 patients at Dewey’s institution shows that 84% of patients clinically referred to CT preferred this hybrid approach the analysis by radiologists or computers alone.
3. Link Between Tests and Treatments
Developing structured reports that may enhance the link between test results and treatment recommendations is the third area of opportunity. “So far, the key information is hidden somewhere in the reports,” said Dewey. Structured reports would provide pre-defined descriptions. This could maximize objectivity and reduces variability of prose text often used in reports.
Structured data could be leveraged as a mine for prediction tools. He shared his experience about clinical data-half life for these tools, saying that data older than four months is usually useless. „Importantly and against general expectations that still persist, I strongly believe that for value-based radiology, we don’t need big data, we need good data,“ he emphasized.
Getting Into Nuances of Patient Care
Success in the three areas of opportunity might free radiologists from routine work to become a more active participant in the nuances of patient care, believes Dewey. He defined nuances as patient-centered decision making that takes psychological situation, worries, beliefs and soft factors in general into account. „Those nuances automatic systems are not good and can only emerge when radiologists spend more time talking to patients and other clinicians,“ he said.