Modelbased Automatic ACPC Detection on Threedimensional MRI Scans

Model-based Automatic AC/PC Detection on Three-dimensional MRI Scans Babak A. Ardekani, Ph. D. , Alvin H. Bachman, Ph. D. , Ali Tabesh, Ph. D. The Nathan S. Kline Institute for Psychiatric Research, Orangeburg, NY

The anterior and posterior commissures are bundles of transverse white matter fibers that connect the two cerebral hemispheres

AC PC The AC and PC landmarks are intersection points of these fibers with the mid-sagittal plane

Aim To develop an algorithm for automatic detection of the AC/PC on 3 D structural MRI scans

Applications • Orientation of the human brain for computerized image analysis • Definition of coordinate systems in brain atlases (Talairach-Tournoux; Schaltenbrand-Wahren) • Placement and orientation of the imaging FOV in MRI acquisition • Image registration

Example application FOV placement for MRI acquisition

Method • 3 D template matching • Normalized cross-correlation (NCC) similarity measure

Template definition AC/PC and MPJ landmarks are manually placed on example images by an expert

Algorithm MRI volume Detect the mid-sagittal plane (MSP) MPJ template AC/PC templates Detect midbrain-pons junction (MPJ) Detect AC/PC Update MSP Save AC/PD Save MSP

MSP detection Ax + By + Cz = 1 Ardekani et al. , IEEE Trans. Medical Imaging, 1997.

Algorithm MRI volume Detect the mid-sagittal plane (MSP) MPJ template AC/PC templates Detect midbrain-pons junction (MPJ) Detect AC/PC Update MSP Save AC/PD Save MSP

MPJ detection Candidate MPJ points are detection on a circular search region by template matching using NCC

Algorithm MRI volume Detect the mid-sagittal plane (MSP) MPJ template AC/PC templates Detect midbrain-pons junction (MPJ) Detect AC/PC Update MSP Save AC/PD Save MSP

AC/PC detection The AC/PC are detected based on each possible MPJ location

AC/PC detection The final decision is made by adding the NCC’s of all three landmarks

Algorithm MRI volume Detect the mid-sagittal plane (MSP) MPJ template AC/PC templates Detect midbrain-pons junction (MPJ) Detect AC/PC Update MSP Save AC/PD Save MSP

MRI data • 36 healthy volunteers, 17 patients with chronic schizophrenia (total: 53 scans) • Siemens Vision 1. 5 T • 3 D T 1 -weighted MPRAGE structural MRI scans • TR=11. 6 ms, TE=4. 9 ms, =8 , Matrix=256× 190, 1 mm 3 voxels

Template definition 3 patient (top row) and 3 control (bottom row) scans were used to construct templates for AC/PC and MPJ

Results Qualitatively correct AC/PC locations were detected on 52 of the 53 cases.

Results The 1 scan on which the algorithm failed

Results 5 of 53 scans had severe artifacts

Results Processing time: 6 s on Pentium 4, 3. 2 GHz (4. 5 s MSP detection + 1. 5 s AC/PC detection); 2 s on Quad Core Intel Xeon E 5430, 2. 66 GHz

Results: Manual vs. automatic detection AC PC Average error 0. 86 mm 0. 85 mm Maximum error 1. 60 mm 1. 74 mm Error < 1. 0 mm 34/42 33/42 3 D Euclidean distance (error) between manually and automatically detected AC/PC in 42 scans (53 -6 -5=42)

Summary • Fast, accurate, robust, and fully automatic AC/PC detection • Algorithm can be trained for contrasts other than T 1 (e. g. , T 2 -weighted FSE) • Algorithm shows robustness with respect to field strength, pulse sequence parameters, subject population • Available on www. nitrc. org