The first AI-based pathology diagnostic system is up and running in a live clinical setting to assist in the confirmation of prostate cancer.
The Institute of Pathology at Maccabi Healthcare Services in Israel has deployed Ibex Medical Analytics’ Second Read (SR) System to identify various cell types and features within whole slide images of prostate core needle biopsies (PCNBs), including grading cancerous glands and other clinically significant features. Its use follows a pilot period in which it identified isolated major errors in retrospective PCNBs that were previously diagnosed as benign by pathologists.
“Prostate cancer diagnosis is labor intensive,” Joseph Mossel, co-founder and CEO of Ibex, the strategic partner of Maccabi, told HCB News. “It requires reviewing a large number of biopsy slides with potential for a human pathologist to miss relatively small cancer foci. Furthermore, there is potential for more mundane human error such as missing a slide within a case, typing errors and so forth. These risks have become more pronounced in recent years with an ever-increasing workload and a global shortage of pathologists.”
The SR system analyzes the entire case and alerts users to discrepancies found with the original diagnosis.
Following its deployment, the solution utilized AI and machine learning techniques to identify a suspicious PCNB that was diagnosed earlier in the day as benign by a pathologist at the institute, which handles approximately 700 PCNBs out of 160,000 histology accessions annually.
Staff re-examined the PCNB and confirmed the presence of low-grade prostate cancer, demonstrating the higher efficiency and accuracy of the platform.
System training is derived from thousands of image samples of hundreds of PCNBs in multiple institutes. Digital images used for this initiative were produced by The Maccabi Pathology Institute’s Philips IntelliSite Pathology solution.
Training is currently limited to the prostate with the potential to teach the system how to diagnose other forms of cancer provided a large enough training set is available. This, however, could be a challenge for very rare types of cancer.
Mossel says, however, that the deployment of the system provides faster diagnoses and better management of population health in a cost-effective manner.
“The system allows for better management and improvement of population health while controlling for costs,” he said. “The deployment of AI-based tools for cancer diagnostics in pathology leads to faster diagnosis and associated treatment, increased efficiency, and increased accuracy.”
Ibex plans to expand the number of cancer types and clinical use cases the solution can address as part of its focus to make the SR system a commercially available product.
Availability of the platform is currently limited.