The new AI system was developed and validated using data sets from 16,000 patients who underwent a scintigraphy imaging examination at nine institutions in Europe and Asia, including Vienna General Hospital, between 2010 and 2020. Scintigraphy is a nuclear medicine procedure that is used to diagnose cancer, thyroid, kidney and heart disease, among other things. The new AI tool created as part of the research work led by Christian Nitsche (MedUni Vienna's Department of Medicine II) and Marcus Hacker (MedUni Vienna's Department of Biomedical Imaging and Image-guided Therapy) can automatically recognise cardiac amyloidosis during a scintigraphy, thereby significantly speeding up the diagnosis.
At least as reliable as doctors
The AI application was not only developed as part of the large-scale study, but also tested for accuracy compared to the diagnostic performance of doctors. "We found that the system can consistently recognise cardiac amyloidosis at least as reliably as medical experts," report lead authors Clemens Spielvogel and David Haberl from MedUni Vienna's Department of of Biomedical Imaging and Image-guided Therapy. The research team also analysed possible correlations between the diagnoses made by the AI system and the occurrence of heart failure and the risk of death. It was found that scintigraphy patients in whom the AI system predicts cardiac amyloidosis have twice the risk of death and a more than 17-fold higher risk of heart failure than patients without such a result.
Cardiac amyloidosis is a rare and often late-diagnosed but serious disease that can hide behind heart insufficiency, for example, more often than previously thought. In 2020, disease-modifying therapies that can halt the progression of cardiac amyloidosis were approved in the European Union for the first time. However, since existing protein deposits and thus the disease cannot be reversed, early and precise diagnosis plays an important role for patients. "In the future, our findings and the technology we have developed could enable screening for cardiac amyloidosis among all scintigraphy patients, with the AI system evaluating the image data in parallel with doctors," says Clemens Spielvogel, summarising the enormous relevance of the study results.
Publication: The Lancet Digital Health
Diagnosis and prognosis of abnormal cardiac scintigraphy uptake at risk for cardiac amyloidosis using artificial intelligence: An international, multi-center, multi-tracer development and validation study;
Clemens P. Spielvogel, David Haberl, Katharina Mascherbauer, Jing Ning, Kilian Kluge, Tatjana Traub-Weidinger, Rhodri H. Davies, Iain Pierce, Kush Patel, Thomas Nakuz, Adelina Göllner, Dominik Amereller, Maria Starace, Alice Monaci, Michael Weber, Xiang Li, Alexander R. Haug, Raffaella Calabretta, Xiaowei Ma, Min Zhao, Julia Mascherbauer, Andreas Kammerlander, Christian Hengstenberg, Leon J. Menezes, Roberto Sciagra, Thomas A. Treibel, Marcus Hacker and Christian Nitsche;
https://www.thelancet.com/journals/landig/article/PIIS2589-7500(23)00265-0/fulltext