Hippocampal MR imaging morphometry by means of general pattern matching

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Anatomy

Background:

• Accurate segmentation and measurement of brain structures like the hippocampus are crucial for neurological research.
• Manual segmentation of hippocampi on MR images can be time-consuming and prone to variability.

Purpose of the Study:

• To evaluate the repeatability and validity of a novel pattern-matching method for hippocampal segmentation and measurement on MR images.
• To compare the performance of pattern matching against manual segmentation techniques.

Main Methods:

• Two-dimensional MR images from 18 subjects (healthy controls and patients with schizophrenia) were segmented twice manually and twice using pattern matching with digital atlas transformation.
• The pattern-matching method involved two steps: global landmark-based matching followed by local viscous fluid model matching.

Main Results:

• The pattern-matching method demonstrated high repeatability with a mean difference of 1.33% between measurements, significantly lower than manual measurements (4.67%).
• Validity was confirmed by a high correlation (ICC = .96) with manual segmentation.
• While schizophrenic hippocampi tended to be smaller, no significant shape differences were observed between groups.

Conclusions:

• General pattern matching using a digital brain atlas is a mathematically sound, reproducible, and less variable method for measuring hippocampi on MR images.
• This automated approach offers advantages over traditional manual segmentation for neuroimaging studies.