A new study from Sweden’s Karolinska Institutet reveals that AI models can identify ovarian cancer in ultrasound images better than human experts. Published in Nature Medicine, the research shows how AI can help diagnose ovarian tumors more accurately and efficiently.
“Ovarian tumors are often found by chance,” says Professor Elisabeth Epstein of Karolinska Institutet. “Many regions lack ultrasound experts, leading to delays in diagnosis and unnecessary procedures. We wanted to see if AI could help.”
AI Accuracy Outshines Experts
Researchers trained neural network models using over 17,000 ultrasound images from 3,652 patients across eight countries. The AI achieved an accuracy rate of 86.3%, outperforming both expert examiners (82.6%) and less experienced ones (77.7%). This shows AI’s potential to assist in diagnosing ovarian cancer, especially in complex cases or regions with limited expertise.
Fewer Referrals, Faster Care
In tests, the AI reduced unnecessary expert referrals by 63% and misdiagnoses by 18%. This means faster, more cost-effective care for patients with ovarian lesions.
Future Research and Clinical Testing
While the results are promising, further studies are needed to confirm AI’s safety and effectiveness in real-world settings. Clinical trials are underway at Stockholm South General Hospital to evaluate the tool’s impact on patient care and healthcare costs.
The study involved researchers from Karolinska Institutet and KTH Royal Institute of Technology and was funded by the Swedish Research Council and other organizations. Some researchers involved have applied for patents related to the technology.
With continued development, AI tools like this could revolutionize healthcare, easing the workload on experts and improving hospital efficiency.
Source: news-medical
