A computer algorithm to quantitatively assess quality of digital optic disc images.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Science

Background:

• Optic disc photography is crucial for glaucoma diagnosis and monitoring.
• Accurate image quality assessment is essential for reliable glaucoma management.
• Current quality assessment relies on subjective human review.

Purpose of the Study:

• To develop and validate an objective, computerized algorithm for assessing optic disc photograph quality.
• To mimic the quality assessment procedures of human observers.
• To evaluate the algorithm's performance on both film-based and digital images.

Main Methods:

• A computerized algorithm was developed to objectively assess image quality.
• The algorithm was tested on film-based (mydriatic) and digital (non-mydriatic) optic disc images.
• Image sharpness was evaluated by masked human readers, and algorithm performance was analyzed using receiver operating characteristic (ROC) curves.

Main Results:

• The algorithm achieved an area under the ROC curve of 1.0 for identifying unreadable images in both digital and film-based datasets.
• For differentiating unreadable from mediocre images, the algorithm achieved AUCs of 0.91 (digital) and 1.0 (film-based).
• The algorithm successfully identified all unreadable images in the pilot study.

Conclusions:

• The developed algorithm demonstrates high accuracy in identifying unreadable optic disc images.
• This objective quality assessment tool shows potential for improving glaucoma management workflows.
• Further research is needed to assess the algorithm's generalizability across different imaging devices and fundus locations.