An AI system developed in Sweden has been shown to match leading urological pathologists at diagnosing prostate cancer in tissue samples.
The researchers wanted to develop an AI-based tool that had clinically acceptable accuracy for detecting prostate cancer, localization, and Gleason grading. The AI system was trained using 6682 digitized slides of needle core biopsies from 976 patients aged 50-69 from Sweden and a further 271 slides from 93 patients outside of Sweden.
The new AI-based system was tested for accuracy by a team of researchers at the Karolinska Institutet in Sweden who evaluated its ability to predict the presence, extent, and Gleason grade of malignant tissue. The accuracy of the AI tool was also compared with the grading performance of 23 experienced urological pathologists. The AI tool was found to detect and grade prostate cancer with a level of accuracy comparable to experienced urological pathologists.
The AI system was developed as part of the OncoWatch project by the same team of researchers who developed the blood-based prostate cancer diagnostic test, Stockholm3, in 2017. Stockholm3 is now used in Sweden, Norway, Finland, and Denmark to screen for prostate cancer and has doubled the number of detected aggressive prostate cancers and has also cut the number of unnecessary biopsies in half.
The new OncoWatch AI tool will be validated later this year in a 9-country multicenter study to assess its performance across a more extensive range of digital pathology scanners.
Each year, around 450,000 new cases of prostate cancer are diagnosed in Europe making it the most commonly diagnosed cancer in men. While prostate cancer is treatable, it is essential for the cancer to be detected quickly when treatment is most likely to be effective. The new tool could help diagnose patients far more quickly, thus helping to improve patient outcomes and reduce the cost of healthcare.
There is currently a global shortage of skilled urological pathologists. The new AI tool could be used on prostate biopsies with the results verified by urological pathologists. That would free up a great deal of time and allow them to concentrate on more difficult cases. Since the AI tool is cloud-based, it can be easily accessed from anywhere. The researchers are expecting a strong uptake for their new tool.
You can read more about the study in the paper – Artificial intelligence for diagnosis and grading of prostate cancer in biopsies: a population-based, diagnostic study – which was recently published in Lancet Oncology. DOI: 10.1016/S1470-2045(19)30738-7