An evaluation of clusters in the glaucomatous visual field

Area of Science:

• Ophthalmology
• Medical Statistics
• Computational Biology

Background:

• Glaucoma is a leading cause of irreversible blindness.
• Early detection and monitoring of glaucomatous visual field defects are crucial for timely intervention.
• Traditional analysis of visual field data can be confounded by noise and long-term fluctuation.

Purpose of the Study:

• To identify natural groupings (clusters) of test locations within visual fields exhibiting typical glaucomatous defects.
• To assess the diagnostic performance of these clusters in distinguishing early glaucoma from normal visual fields.
• To evaluate the effectiveness of cluster analysis in reducing long-term fluctuation in visual field data.

Main Methods:

• Statistical cluster analysis was applied to 76 visual fields with glaucomatous defects to define 11 clusters.
• Local mean defects within clusters and global mean defect were calculated for 70 early glaucomatous and 70 age-matched normal visual fields.
• Long-term fluctuation was analyzed in clustered versus individual test locations using data from 93 stable glaucoma patients and 105 normal subjects.

Main Results:

• The 11 clustered mean defects demonstrated high diagnostic accuracy with 90% sensitivity and 93% specificity.
• Global mean defect showed lower diagnostic performance with 81% sensitivity and 91% specificity.
• Mean fluctuation was significantly reduced in clustered test locations (3.5 dB2 in glaucoma, 0.6 dB2 in normal) compared to individual locations (7.0 dB2 in glaucoma, 1.8 dB2 in normal).

Conclusions:

• Cluster analysis is effective in identifying localized visual field loss characteristic of glaucoma.
• Grouping test locations into clusters enhances the signal-to-noise ratio, improving the discrimination between normal and glaucomatous visual fields.
• This approach shows promise for more robust monitoring of glaucoma progression and stability.