Automatic identification of landmarks for standard slice positioning in brain MRI

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Radiology

Background:

• Standardizing prospective MRI scans is crucial for accurate longitudinal studies and intersubject comparisons.
• Existing slice-positioning techniques can suffer from significant rotational and translational variances, leading to misalignments.
• Novel anatomical landmarks are needed to improve the robustness and automation of MRI slice positioning.

Purpose of the Study:

• To introduce and validate a new automatic slice-positioning technique using three novel anatomical landmarks.
• To standardize prospective MRI scans by minimizing rotational and translational variances.
• To assess the accuracy and robustness of automatic landmark capture for image segmentation.

Main Methods:

• Manual labeling of interpeduncular fossa corner and eyeball centers on 25 T1 MRI scans.
• Generation of new scans based on the Eyeball centers-Mesencephalon (EM) plane.
• Comparison of rotational angles (pitch and yaw) and translational variances between EM and original scans, including analysis of automatically captured landmarks.

Main Results:

• Significant reduction in intersubject variability for pitch (P < 0.005) and yaw (P < 0.001) angles, key causes of misalignment.
• Demonstrated accuracy of automatic segmentation through paired t-tests and significance tests.
• Validation of the EM technique's ability to lower variances.

Conclusions:

• The EM technique provides a new field of view and slice orientation, fixing follow-up scans by reducing rotational and translational variances.
• This method precisely matches intrasubject scans and produces more standardized intersubject scans.
• The identified landmarks possess sufficient distinguishing features for robust automatic capture, enabling precise and standardized neuroimaging.