Units

PubMed

Doi

---

Citation

Type

Journal

Publication Date

Issn Print

Pages

Brief description/objective

PURPOSE
To compare automatic software-based co-registration of serial magnetic resonance imaging (MRI) scans with conventional visual comparison, by expert neuroradiologists.

MATERIAL AND METHODS
Sixty-four patients who were referred to our epilepsy MRI unit for cerebral imaging were identified as having potentially, non-, or slow-growing lesions or cerebral atrophy and followed with sequential scans over a period of up to 8 years, resulting in a total of 92 pairs of scans. Scans were categorized as showing either lesions or atrophy. Each pair of scans was reviewed twice for the presence of change, with and without co-registration, performed using automated software.

RESULTS
Co-registration and visual reporting without co-registration were discordant in the lesions group in nine out of 69 datasets (13%), and in 16 out of 23 pairs of scans in the atrophy group (69%). The most common cause of discordance was visual reporting not detecting changes apparent by co-registration. In three cases, changes detected visually were not detected following co-registration.

CONCLUSION
In the group of patients studied, co-registration was more sensitive for detecting changes than visual comparison, particularly with respect to atrophic changes of the brain. With the increasing availability of sophisticated independent consoles attached to MRI scanners that may be used for image co-registration, we propose that serial T1-weighted volumetric MRI brain co-registration should be considered for integration into routine clinical practice to assess patients with suspected progressive disease.