Exploratory Dijkstra forest based automatic vessel segmentation: applications in video indirect ophthalmoscopy (VIO)

Area of Science:

• Ophthalmology
• Medical Imaging
• Computational Biology

Background:

• Accurate segmentation of the retinal vascular network is crucial for diagnosing various eye conditions.
• Existing methods often require manual intervention or struggle with preserving vessel thickness and branching patterns.

Purpose of the Study:

• To develop and validate an automated methodology for retinal vascular network extraction.
• To improve the accuracy and efficiency of vessel segmentation in retinal images.

Main Methods:

• Utilized Dijkstra's shortest-path algorithm for vascular network extraction.
• Developed a novel approach to preserve vessel thickness and natural branching.
• Constructed a retinal video indirect ophthalmoscopy (VIO) image database from pediatric patients for testing.

Main Results:

• The proposed method demonstrated superior performance compared to state-of-the-art approaches in segmenting retinal vasculature.
• Achieved high accuracy in preserving vessel thickness and following natural branching patterns.
• Outperformed existing methods on both single VIO frames and enhanced large field-of-view mosaics.

Conclusions:

• The developed algorithm provides an efficient and accurate automated solution for retinal vascular network extraction.
• This method has the potential to enhance the diagnosis and monitoring of retinal diseases.
• The freely available dataset and source code will facilitate further research in the field.