Deepnoid and Uijeongbu St. Mary's Hospital Joint Research Published in Two International Journals

Deepnoid, a company specializing in medical artificial intelligence (AI), announced on May 22, 2025, that a paper co-authored with Uijeongbu St. Mary's Hospital of the Catholic University of Korea has been published in an international journal indexed in the Science Citation Index (SCI).

The international journals include AJP (The American Journal of Pathology) and Lung Cancer, both of which are Q1-level journals ranked in the top 25% of their respective academic fields.

The paper explored the potential of AI diagnostic models based on whole slide images. The research team demonstrated that a model with a fine-tuned architecture based on whole slide images delivers higher diagnostic performance compared to patch-based models.

The study analyzed multiple cytology samples (ovarian cancer) and pleural effusion samples (lung cancer) from different cancer types. Using high-quality quality control unit data collected from over 200 pathology departments nationwide through the Korean Society for Cytopathology, the team validated the clinical applicability and generalizability of the model through training and verification.

In the ovarian cancer study published in AJP, the official journal of the American Society for Investigative Pathology (ASIP), the whole slide image-based algorithm model achieved an AUC of 0.87 in external validation, showing approximately 9% better results than the patch-based model. The model can be trained without separate patch annotations, making it highly practical and suitable for complementary use alongside medical professionals' interpretations.

For the lung cancer diagnostic study published in the specialized clinical oncology journal 'Lung Cancer,' the research team set the whole slide image as the unit of analysis using nationwide multicenter data. By applying this algorithm, they confirmed improved performance and analysis efficiency compared to traditional image patch-based models. In the AI lung cancer diagnostic study, the model achieved an accuracy of 97% and an AUC of 0.97, demonstrating approximately 13% improved performance over patch-based models when applying the whole slide image-based algorithm.

Professor Joseph Jung of Uijeongbu St. Mary's Hospital, Catholic University of Korea, a member of the joint research team, stated, "Through this study, we have objectively confirmed the effectiveness of AI-based diagnosis with data. We will continue our research to support medical professionals on the front lines and improve diagnostic accuracy and efficiency by applying this approach to a wider variety of diseases beyond ovarian and lung cancer."

Yoon Hongjun, team leader at Deepnoid, said, "Through training and validation based on nationwide quality control data, we have simultaneously demonstrated both the clinical applicability and generalizability of the whole slide image-based model. Building on the objective validation of our research results through publication in leading international journals, we will continue to pursue in-depth research."