AI can improve ovarian cancer diagnoses

“Ovarian tumors are common and are often discovered incidentally,” says Professor Elisabeth Epstein from the Department of Clinical Sciences and Education at Karolinska Institutet Södersjukhuset (Stockholm Southern General Hospital) and senior consultant at the hospital’s Department of Obstetrics and Gynecology. “World There is a severe shortage of ultrasound experts in many places, raising concerns about unnecessary intervention and delayed cancer diagnosis, so we wondered whether artificial intelligence could complement human experts.”

Artificial intelligence outperforms experts

The researchers developed and validated a neural network model capable of distinguishing benign from malignant ovarian lesions and trained and tested it on more than 17,000 ultrasound images from 3,652 patients at 20 hospitals in eight countries. They then compared the model’s diagnostic capabilities to a large group of experts and less experienced sonographers.

The results showed that the artificial intelligence model was better than expert and non-expert examiners in identifying ovarian cancer, with an accuracy of 86.3%, while the accuracy of expert and non-expert examiners was 82.6% and 77.7% respectively.

Professor Epstein said: “This shows that neural network models can provide valuable support in the diagnosis of ovarian cancer, especially in difficult-to-diagnose cases and where ultrasound experts are lacking.”

Reduce the need for expert recommendations

AI models can also reduce the need for expert recommendations. In simulated triage situations, artificial intelligence support reduced the number of referrals by 63% and the misdiagnosis rate by 18%. This could provide faster, more cost-effective care for patients with ovarian pathology.

Although the results are encouraging, the researchers stress that further research is needed before the full potential of the neural network model and its clinical limitations can be fully understood.

Filip Christiansen said: “Through continuous research and development, artificial intelligence-based tools can become an integral part of the future of healthcare, reducing the burden on specialists and optimizing hospital resources, but we need to ensure that they can adapt to different clinical settings and Patient groups.

Assessing the security of artificial intelligence support

The researchers are currently conducting a prospective clinical study in Södersjukhuset to evaluate the daily clinical safety and usefulness of the artificial intelligence tool. Future research will also include a randomized multicenter study to examine the impact on patient management and healthcare costs.

The study was conducted in close collaboration with researchers at KTH Royal Institute of Technology and was funded by the Swedish Research Council, the Swedish Cancer Society, the Stockholm Regional Council, the Radiumhemmet Cancer Research Fund and the Wallenberg AI, Autonomous Systems and Software Programs (WASP).

Elisabeth Epstein, Filip Christiansen and three co-authors have patented a computer-supported diagnostic method through Intelligyn. Elisabeth Epstein, Filip Christiansen and KTH Royal Institute of Technology researcher Kevin Smith also hold shares in Intelligyn, where Professor Epstein is an unpaid manager .