Sign in Subscribe
Machine learning

Deep Learning Techniques Show Promise in Medical Image Processing

Strategic Insights

— 1 min read

Researchers from Marmara University have developed a computer-based diagnostic software that uses deep learning techniques to assess the segmentation of the mandibular condyle in ultrasound images. This innovative approach has demonstrated high accuracy, with an F1 score of 0.93, sensitivity of 0.90, and precision of 0.96. The study's findings suggest that this technology has the potential to save time in the diagnostic process for surgeons, radiologists, and other specialists.

Key Takeaways:

  • A total of 668 retrospective ultrasound images of anonymous adult mandibular condyles were analyzed in the study.
  • The CranioCatch labeling program was used to annotate the mandibular condyle using a polygonal labeling method, which was reviewed and validated by experts in oral and maxillofacial radiology.
  • The YOLOv8 deep learning artificial intelligence (AI) model was used to detect and segment the mandibular condyle from ultrasound images.
  • The model's performance in image estimation was evaluated, achieving an F1 score of 0.93, sensitivity of 0.90, and precision of 0.96.
  • The study's findings suggest that this technology has the potential to save time in the diagnostic process for surgeons, radiologists, and other specialists.
  • The research has been peer-reviewed and published in the Journal of Imaging Informatics In Medicine.
  • The study's authors include Gaye Keser, Hakan Yulek, Ayse Gul Oner Talmac, Ibrahim Sevki Bayrakdar, Filiz Namdar Pekiner, and Ozer Celik.

Statistics:

  • 668 retrospective ultrasound images of anonymous adult mandibular condyles were analyzed in the study.
  • The YOLOv8 deep learning AI model achieved an F1 score of 0.93, sensitivity of 0.90, and precision of 0.96 in image estimation.
  • The CranioCatch labeling program was used to annotate the mandibular condyle using a polygonal labeling method.

Sources:

  • A Deep Learning Approach for Mandibular Condyle Segmentation on Ultrasonography. Journal of Imaging Informatics In Medicine, 2025.
  • NewsRx. Marmara University Reports Findings in Artificial Intelligence (A Deep Learning Approach for Mandibular Condyle Segmentation on Ultrasonography). Journal of Engineering. May 19, 2025; p 1586.

Read more