This morning, Ezra Biomedical Engineer Helen Xu gave an exciting presentation at the 2019 Computer Assisted Radiology and Surgery (CARS) conference concerning the latest data on our artificial intelligence (AI)-based lesion segmentation algorithm. Held in Rennes, France, the CARS conference features presentations on a host of the newest and most impressive patient-minded technologies. Our AI technology is now able to detect prostate cancer lesions on multi-parametric MR images (mp-MRI) with 93% accuracy.
Ezra’s engineers have developed an AI algorithm using MRIs from 346 patients obtained from the Cancer Imaging Archive data portal, which had each been evaluated by three radiologists. The segmentations made by the most senior radiologist – who has more than 15 years of experience – were used as the “ground truth” for AI network training and validation. The segmentations made by the other two radiologists were used to establish a baseline comparison of the network performance.
The AI uses an advanced neural network architecture to learn from the radiologist’s assessment of prostate lesions on the mp-MRIs. It is then able to simulate the radiologist and predict lesion locations and boundaries that could be clinically significant cancers on new images that the AI has never before seen. The AI has demonstrated its ability to learn features useful in prostate lesion detection, and has achieved results similar to the state-of-the-art studies. It has the potential to help radiologists identify abnormal cancerous lesions faster, and with more accuracy.
“Our goal is to improve the speed, accuracy, and affordability of cancer screening, and I’m thrilled to see the work our AI team is doing to accomplish that,” says Emi Gal, our CEO & co-founder. “We look forward to integrating our exciting new technology into clinical workflows once we obtain FDA clearance, which we expect later this year.”
We plan to expand our Prostate AI to all other organs, in order to make radiologists more accurate and more efficient.