Poster Session 1
Category: Clinical Obstetrics
Poster Session 1
Un Yung Choi, MD
Seoul National University Hospital
Seoul, Seoul-t'ukpyolsi, Republic of Korea
Young Min Kim
Department of Transdisciplinary Medicine, Seoul National University Hospital, Seoul, Republic of Korea
Seoul, Seoul-t'ukpyolsi, Republic of Korea
Haeryoung Kim
Department of Pathology, Seoul National University Hospital, Seoul National University College of Medicine, Seoul, Republic of Korea
Seoul, Seoul-t'ukpyolsi, Republic of Korea
Ji Hoi Kim, MD, PhD
Seoul National University Hospital
Seoul, Seoul-t'ukpyolsi, Republic of Korea
Eun Na Kim
Department of Pathology, Seoul National University Hospital, Seoul National University College of Medicine, Seoul, Republic of Korea
Seoul, Seoul-t'ukpyolsi, Republic of Korea
Youngbin Ahn
Department of Transdisciplinary Medicine, Seoul National University Hospital, Seoul, Republic of Korea
Seoul, Seoul-t'ukpyolsi, Republic of Korea
Tae Kwan Lee
Department of Obstetrics and Gynecology, Seoul National University Hospital, Seoul National University College of Medicine, Seoul, Republic of Korea
Seoul, Seoul-t'ukpyolsi, Republic of Korea
Sun Min Kim, MD, PhD
Associate professor
Seoul National University College of Medicine
Seoul, Seoul-t'ukpyolsi, Republic of Korea
Byoung Jae Kim, MD, MS
Seoul National University College of Medicine
Seoul, Seoul-t'ukpyolsi, Republic of Korea
Jaehee Jeong
Department of Obstetrics and Gynecology, Seoul National University Hospital, Seoul National University College of Medicine, Seoul, Republic of Korea
Seoul, Seoul-t'ukpyolsi, Republic of Korea
So Hee Kim, N/A
Seoul National University Hospital
Seoul, Seoul-t'ukpyolsi, Republic of Korea
Chan-Wook Park, PhD
Department of Obstetrics and Gynecology, Seoul National University Hospital, Seoul National University College of Medicine, Seoul, Republic of Korea
Seoul, Seoul-t'ukpyolsi, Republic of Korea
Joong Shin Park, PhD
Department of Obstetrics and Gynecology, Seoul National University Hospital, Seoul National University College of Medicine, Seoul, Republic of Korea
Seoul, Seoul-t'ukpyolsi, Republic of Korea
Young-Gon Kim
Department of Transdisciplinary Medicine, Seoul National University Hospital, Seoul, Republic of Korea
Seoul, Seoul-t'ukpyolsi, Republic of Korea
Seung Mi Lee, MD, PhD
Department of Obstetrics and Gynecology, Seoul National University College of Medicine
Seoul, Seoul-t'ukpyolsi, Republic of Korea
Artificial intelligence (AI)-based imaging can enhance histopathologic diagnosis, but its clinical significance has not been well determined. The purpose of the current study is to compare the performance of an AI-generated inflammation index of umbilical cord with traditional pathologic criteria (funisitis) for identifying fetal inflammatory response syndrome (FIRS).
Study Design:
We developed an attention-based multiple instance learning (MIL) model to quantify umbilical cord inflammation from whole-slide images. The model was trained with patches from umbilical cord slides to localize inflammatory regions and output patch-wise inflammation scores. The proportion of inflamed areas was calculated by binarizing patch scores using a defined threshold. The performance of AI-generated inflammation index and the grade of funisitis were compared for identification of FIRS, which was defined as high neonatal blood CRP.
Results:
We analyzed a cohort comprising 386 cases who were delivered before 34 weeks of gestation. The FIRS was defined as neonatal CRP >0.4385, which was defined as the optimal cutoff for early neonatal sepsis in the current study population. The AUROC of AI-based inflammatory index was higher than that of funisitis grade for identification of FIRS (0.71 vs 0.67). Cases with high AI-based inflammatory index ( >75percentile) had lower gestational age at delivery and high frequency of histologic chorioamnionitis. Among the study population, 103 women underwent amniocentesis within 1 month of delivery. Women with AI-based inflammatory index were more likely to have intraamniotic infection/inflammation (27.6% vs 88.2%, p< 0.001), and high AF WBC counts (median, 2.00±10.625 vs 1700±2006.25, p< 0.001)
Conclusion:
The AI-derived inflammatory index from umbilical cord demonstrated improved discriminative ability over standard pathologic evaluation (funisitis grade) in identifying FIRS and were related to intraamniotic infection/inflammation, suggesting its potential as a more objective and quantitative tool for perinatal inflammation assessment.