Deep Bio’s Prostate Cancer Diagnostic Software Demonstrates High Concordance with Uropathologists at AUA 2021

Share Article

Two posters featuring analyses by Deep Bio’s AI algorithm for prostate cancer diagnosis, DeepDx®, were highlighted at AUA 2021

Deep Bio, a pioneer in medical AI specialized for cancer diagnostics, announced that new study results on DeepDx® Prostate, Deep Bio’s deep learning-based prostate cancer diagnosis software, were presented by researchers from the Stanford University School of Medicine at the 2021 annual meeting of American Urological Association (AUA 2021). Findings describe robust performance of the algorithm in diagnosing and grading cancer on prostatectomy specimens.

In a study, researchers aimed to externally validate the diagnostic performance of DeepDx® Prostate on prostate cancer detection and Gleason grading. The algorithm and two expert uropathologists analyzed 500 tiles created from 150 whole mount prostatectomy specimens by establishing Gleason grade, amount of cancer, and percentage of Gleason pattern 4 and 5 in the tiles. Despite being trained on a different patient population and on needle biopsy cores instead of prostatectomy specimens, DeepDx® Prostate achieved overall high agreement with the reference standard created by consensus of the two uropathologists, demonstrating Cohen’s kappa score of 0.79 (95% CI 0.75 - 0.82). It performed especially well at clinically meaningful thresholds in evaluating benign vs malignant (κ 0.927), and clinically low risk (benign, GG 1, or GG2) versus high risk (GG 3-5) disease (κ 0.858).

DeepDx® Prostate was also utilized in another Stanford Medicine study presented at the conference. This study sought to measure concordance in tumor detection between preoperative MRI and prostatectomy histopathology. Data was sampled from 30 men following prostatectomy, and whole mount specimens were registered to MRI images using a state-of-the-art 3D approach. DeepDx® Prostate was used to identify and grade all the cancerous lesions. Compared to the algorithm that detected all lesions and graded severity regardless of size, MRI alone missed 66% of the cancers, 37% of which were clinically significant tumors. Additionally, using the algorithm significantly reduced time spent for annotating and grading, while it took an average of 45 minutes per specimen for pathologists just to identify tumor areas.

“We are excited to see clinical research featuring use of DeepDx® Prostate presented at AUA 2021 among all the other leading urology research,” said Sun Woo Kim, the CEO of Deep Bio. “We hope this external validation of our algorithm and its application in research reinforces our algorithm’s performance and versatile use cases.”

Deep Bio has been participating in research collaborations with Stanford University School of Medicine utilizing DeepDx® Prostate since 2019. DeepDx® Prostate was trained on prostate core needle biopsy samples and analyzes whole slide images (WSIs) of the biopsy specimens to detect cancer and grade cancer severity based on the Gleason scoring system. The algorithm automatically analyzes other critical values such as the percentage of each Gleason pattern, tumor-to-tissue ratio and the total tumor/tissue length.

About Deep Bio
Deep Bio Inc. is an AI biotech company with in-house expertise in deep learning, pathology, life sciences, and pharmacotherapeutics. As the country’s first to obtain Korea’s MFDS (Ministry of Food and Drug Safety) approval of an AI-based cancer diagnostic support solution, Deep Bio envisions a suite of AI-based IVD SaMDs (In Vitro Diagnostics Software as a Medical Device) for diagnosis and prognosis of multiple cancers. Deep Bio is actively engaged in the research space and participating in ongoing collaborations with top US medical centers. To learn more, visit

Share article on social media or email:

View article via:

Pdf Print

Contact Author

Yiseul Park
Deep Bio Inc.
+82 70-7703-4419
Email >
Visit website