A recent international study led by researchers at Karolinska Institutet in Sweden has demonstrated that artificial intelligence (AI) models can outperform human experts in identifying ovarian cancer in ultrasound images. The study, published in Nature Medicine, aimed to address the shortage of ultrasound experts in many parts of the world, which can result in unnecessary interventions and delayed cancer diagnoses. By developing and validating neural network models trained on over 17,000 ultrasound images from 3,652 patients across 20 hospitals in eight countries, the researchers found that the AI models were more accurate at differentiating between benign and malignant ovarian lesions compared to both expert and non-expert examiners.
The results of the study showed that the AI models achieved an accuracy rate of 86.3 percent, surpassing the 82.6 percent accuracy of expert examiners and the 77.7 percent accuracy of non-expert examiners. This suggests that AI can provide valuable support in the diagnosis of ovarian cancer, particularly in challenging cases and environments where ultrasound experts are scarce. Additionally, the AI models were found to reduce the need for expert referrals, with a simulated triage situation showing a 63 percent reduction in referrals and an 18 percent decrease in misdiagnosis rates. This has the potential to improve the efficiency and cost-effectiveness of care for patients with ovarian lesions.
Despite the promising results, the researchers stress the importance of further studies to fully understand the potential and limitations of the neural network models in clinical practice. Ongoing research includes prospective clinical studies at Södersjukhuset to evaluate the safety and usefulness of the AI tool in everyday clinical settings. Future studies will also involve a randomised multicentre study to assess the impact of AI support on patient management and healthcare costs. The study was conducted in collaboration with researchers at the KTH Royal Institute of Technology and was supported by grants from various funding sources.
The researchers at Karolinska Institutet, along with their collaborators, are continuing to explore the applications of AI-based tools in healthcare. They believe that with further research and development, AI can become an essential component of future healthcare systems, helping to alleviate the burden on experts and optimize hospital resources. It is essential to ensure that these AI tools can be adapted to different clinical environments and patient populations in order to maximize their benefits. The team has also applied for a patent for methods of computer-supported diagnostics through the company Intelligyn, with several researchers holding shares in the company.
Overall, the study demonstrates the potential of AI models to enhance the diagnosis of ovarian cancer through their ability to outperform human experts in identifying malignant lesions on ultrasound images. The findings pave the way for further research into the clinical applications of AI support in healthcare settings, with the goal of improving patient outcomes and healthcare efficiency.