A Novel Curvature-Based Algorithm for Automatic Grading of Retinal Blood Vessel Tortuosity.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computational Biology

Background:

• Retinal blood vessel tortuosity is a significant clinical sign.
• Diabetic retinopathy and retinopathy of prematurity are major causes of vision impairment linked to vessel tortuosity.

Purpose of the Study:

• To introduce an automated, image-based method for quantifying single vessel and vessel network tortuosity.
• To enhance existing curvature calculation methods for improved linearity and accuracy.

Main Methods:

• Developed an automatic image-based algorithm for retinal vessel tortuosity measurement.
• Modified the template disk method for curvature calculation to improve linearity.
• Validated the algorithm on public and private datasets.

Main Results:

• The proposed method demonstrates high correlation with expert subjective assessments (0.94 for vessel tortuosity, 0.95 for vessel network tortuosity in diabetic retinopathy, 0.7 for vessel network tortuosity in retinopathy of prematurity).
• The algorithm is computationally efficient and simpler to implement than current state-of-the-art techniques.
• Established a non-linear relationship between expert-perceived tortuosity and curvature measurements.

Conclusions:

• The novel automated method accurately measures retinal vessel tortuosity.
• The improved curvature calculation enhances the reliability of tortuosity assessment.
• This technique offers a valuable tool for diagnosing and monitoring conditions like diabetic retinopathy and retinopathy of prematurity.