Related Experiment Video
Updated: Jul 9, 2026
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
Deep Learning-Based Automated Segmentation and Multi-Parametric Quantitative Assessment of the Lacrimal Drainage
Peifang Xu1, Yishu Zhang1, Pengjie Chen1
1Eye Center, The Second Affiliated Hospital, School of Medicine, Zhejiang University, Zhejiang Provincial Key Laboratory of Ophthalmology, Zhejiang Provincial Clinical Research Center for Eye Diseases, Zhejiang Provincial Engineering Institute on Eye Diseases, Hangzhou, Zhejiang, People's Republic of China.
A deep learning framework automates lacrimal system analysis from CT dacryocystography (CT-DCG) for primary acquired nasolacrimal duct obstruction (PANDO). This offers objective, quantitative metrics for improved surgical planning and preoperative assessment.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Primary acquired nasolacrimal duct obstruction (PANDO) is a frequent cause of epiphora.
- Surgical success in PANDO treatment is influenced by obstruction site and anatomical variations.
Purpose of the Study:
- To present a deep learning-based framework for automated morphological and quantitative analysis of the lacrimal system using CT dacryocystography (CT-DCG).
- To provide objective anatomic metrics for preoperative assessment and surgical planning in PANDO.
Main Methods:
- A deep learning network (Attention U-Net) was developed to segment and reconstruct the lacrimal sac and bony nasolacrimal duct (BNLD) in 151 patients with unilateral PANDO.
- Automated quantification of lacrimal sac and BNLD dimensions was performed slice-by-slice.
- Segmentation accuracy was assessed using the Dice coefficient, and clinical agreement for lacrimal sac size classification was evaluated using Cohen's kappa coefficient.
Main Results:
- The deep learning model achieved a Dice coefficient of 0.79 for BNLD segmentation.
- The system enabled automatic, slice-by-slice morphological quantification, identifying the narrowest cross-sectional plane and obstruction site.
- Automated lacrimal sac size classification showed 84.7% agreement with expert assessment (Cohen's kappa = 0.76).
Conclusions:
- The deep learning framework is technically feasible for automated segmentation and morphological quantification of the lacrimal system on CT-DCG.
- Objective anatomic metrics from this system show potential for assisting in preoperative assessment and surgical planning for PANDO.
- Automated analysis offers an efficient, objective alternative to manual assessment for nasolacrimal duct obstruction.