Automated algorithm for counting microbleeds in patients with familial cerebral cavernous malformations

Area of Science:

• Neuroimaging
• Medical Algorithms
• Genetics

Background:

• Familial cerebral cavernous malformation (CCM) is characterized by multiple, growing lesions.
• Susceptibility-weighted imaging (SWI) reliably detects these lesions.
• Manual microbleed counting is time-consuming and prone to variability.

Purpose of the Study:

• To develop and validate an automated algorithm for counting cerebral cavernous malformation (CCM) microbleeds on SWI images.
• To improve the accuracy and consistency of lesion quantification in familial CCM.

Main Methods:

• Modified an existing algorithm for microbleed detection on SWI.
• Trained the algorithm on manually delineated CCM microbleeds.
• Compared automated counts with manual counts using linear regression and ICC.
• Validated the algorithm's sensitivity against a consensus read.

Main Results:

• Strong agreement between automated and manual counts (ICC=0.95 baseline, ICC=0.88 longitudinal).
• No significant difference in average counts between manual and automated methods.
• Algorithm achieved 50% sensitivity in validation, with strong agreement (ICC=0.77).

Conclusions:

• The automated algorithm provides a consistent and reliable method for counting microbleeds in familial CCM.
• This tool can significantly facilitate lesion quantification and tracking in CCM patients.
• Automated counting aids in monitoring disease progression and treatment response.