Title: Cesarean Scar Defect Segmentation in Transvaginal Ultrasound Images: a Dataset and Benchmark

Abstract

Cesarean Scar Defect (CSD) stands as one of the most common complications arising after a cesarean delivery. While transvaginal ultrasonography is a standard tool for initial CSD screening, accurately defining the defect's boundaries and measurements remains essential for effective treatment. Nevertheless, in resource-limited environments, sonographers often miss these defects due to a combination of factors: the lesions' small size and irregular shapes, subpar image quality, and a general lack of clinical awareness.

Although artificial intelligence has made significant strides in medical imaging, there has been a notable absence of public datasets dedicated to segmenting CSDs in transvaginal ultrasound. To bridge this critical gap, we introduce a robust CSD dataset containing 1,111 images and 16 videos. This collection includes 501 confirmed positive cases, each accompanied by precise, pixel-level manual annotations. These annotations were developed in strict adherence to standardized clinical guidelines, resulting from a collaborative effort between seasoned sonographers and trained PhD students.

This study offers high-quality benchmark resources designed to drive advancements in medical image segmentation algorithms and foster clinical innovation. By facilitating better CSD diagnosis and refining subsequent treatment strategies, this work aims to improve the quality of life for women of reproductive age, thereby delivering substantial value to both medical research and clinical practice.

Source: arXiv Generated at: 2026-06-03 00:00:00 UTC