Vascular tree tracking and bifurcation points detection in retinal images using a hierarchical probabilistic model

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Retinal vascular tree extraction is vital for computer-aided diagnosis and surgical planning.
• Junction point detection and classification offer insights into vascular network structure for retinal disease analysis.

Purpose of the Study:

• To develop a novel machine learning algorithm for the joint classification and tracking of retinal blood vessels.
• To improve the accuracy of junction point detection and classification in retinal images.

Main Methods:

• A hierarchical probabilistic framework was employed for joint classification and tracking.
• Local intensity cross-sections were classified using Gaussian basis functions and Gamma distributions.
• A directed Probabilistic Graphical Model (PGM) was utilized with Maximum Likelihood (ML) estimation.

Main Results:

• The proposed method achieved 88.67% precision and 88.67% recall on the REVIEW database.
• Experiments demonstrated promising results in bifurcation point detection and classification.

Conclusions:

• The developed technique provides a classifier with high precision and recall.
• This method shows superior performance compared to existing approaches like Xu's method.