A new international study led by researchers at Sweden’s Karolinska Institutet shows that AI-based models can outperform human experts in identifying ovarian cancer in ultrasound images. The study is published in Nature medicine.
“Ovarian tumors are common and often detected by chance,” says Professor Elisabeth Epstein from the Department of Clinical Science and Education at Södersjukhuset (Stockholm Southern General Hospital), Karolinska Institutet and senior consultant at the Department of Obstetrics and Gynecology. from the hospital. “There is a serious shortage of ultrasound experts in many parts of the world, which has raised concerns about unnecessary interventions and delays in cancer diagnoses. Therefore, we wanted to find out if AI can complement human experts.”
AI outperforms experts
The researchers have developed and validated neural network models capable of differentiating between benign and malignant ovarian lesions, having trained and tested the AI on more than 17,000 ultrasound images of 3,652 patients in 20 hospitals in eight countries. They then compared the diagnostic ability of the models with that of a large group of experts and less experienced ultrasound examiners.
The results showed that the AI models outperformed expert and non-expert examiners in identifying ovarian cancer, achieving an accuracy rate of 86.3 percent, compared to 82.6 percent and 77.7 percent. percent of expert and non-expert examiners, respectively.
“This suggests that neural network models may offer valuable support in the diagnosis of ovarian cancer, especially in difficult-to-diagnose cases and in settings where there is a shortage of ultrasound experts,” says Professor Epstein.
Reduce the need for expert referrals
AI models can also reduce the need for expert referrals. In a simulated triage situation, AI support reduced the number of referrals by 63 percent and the misdiagnosis rate by 18 percent. This may lead to faster, more cost-effective care for patients with ovarian injuries.
Despite the promising results, the researchers stress that more studies are needed before fully understanding the full potential of neural network models and their clinical limitations.
“With continued research and development, AI-based tools can be an integral part of tomorrow’s healthcare, freeing up experts and optimizing hospital resources, but we must ensure they can adapt to different clinical settings and patient groups. patients,” says Filip Christiansen. , PhD student in Professor Epstein’s research group at Karolinska Institutet and joint first author with Emir Konuk at KTH Royal Institute of Technology.
AI Support Security Assessment
Researchers are now conducting prospective clinical studies at Södersjukhuset to evaluate the daily clinical safety and utility of the AI tool. Future research will also include a multicenter randomized study to examine its effect on patient treatment and healthcare costs.
The study was carried out in close collaboration with researchers at the KTH Royal Institute of Technology and was funded by grants from the Swedish Research Council, the Swedish Cancer Society, the Stockholm Regional Council, the Radiumhemmet and Wallenberg Cancer Research Funds. AI, Autonomous Systems. and Software Program (WASP).
Elisabeth Epstein, Filip Christiansen and three co-authors applied for a patent through the company Intelligyn for computer-aided diagnostic methods. Elisabeth Epstein, Filip Christiansen and Kevin Smith, a researcher at the KTH Royal Institute of Technology, also own shares in Intelligyn, of which Professor Epstein is a non-salaried manager.