Adaptive Tissue Cluster Tracking on Quantitative MRI for Fully Automatic Brain Segmentation on Young Children
Marcel Warntjes1,2, Suraj Serai3, James Leach3, and Blaise Jones3

1Center for Medical Imaging Science and Visualization, Linköping, Sweden, 2SyntheticMR AB, Linköping, Sweden, 3Department of radiology, Cincinnati, OH, United States

Synopsis

Brain tissue properties change rapidly during the first few years of life. This poses a problem for brain segmentation algorithms since adult tissue definitions for white matter and grey matter do not apply for young children. An automatic tissue cluster tracking algorithm was developed to determine WM and GM cluster positions in a 3-dimensional search-space of quantitative R1 relaxation rate, R2 relaxation rate and proton density. These positions are then used to segment the brain, independent of age.

Purpose

To create an algorithm that adaptively tracks the grey matter and white matter tissue properties in quantitative MRI data of longitudinal R1 relaxation rate, transverse R2 relaxation rate and proton density, in order to segment grey matter, white matter and cerebrospinal fluid volumes of the brain, independent of age.

Methods

A group of 23 quantified datasets at 3T of paediatric clinical cases in the range 0-20 years old was used to develop an algorithm to automatically track the mean R1, R2 and PD values of GM, myelinated WM and CSF. R1, R2 and PD values were simultaneously acquired using MAGiC on a GE 750 3T system. The positions of the tissue clusters were then used to define GM, myelinated WM and CSF partial volume. The largest contiguous volume of WM, GM and CSF was considered as the intracranial volume where the edge was refined to a PD = 50% threshold. The sum of all partial volumes in the intracranial volume resulted in an estimation of total GM, WM and CSF volumes of all subjects.

Results

The observed (R1/R2/PD) values times for GM changed from (0.5s-1/6.8s-1/86%) to (0.6s-1/12s-1/86%) in the first two years of life, whereas myelinated WM changed from (0.6s-1/7.2s-1/85%) to (1.1s-1/13s-1/72%). After two years R1, R2 and PD were relatively constant. CSF had R1/R2/PD = 0.24s-1/0.81s-1/100% for all ages. Application of adaptive tissue cluster tracking on GM and WM showed that myelinated WM volume, an average, increased from 0 to 252 mL, CSF decreased from 241 mL to 40 mL and total brain volume increased from 403 mL to 1225 mL in the first 4 years of life. In comparison to using fixed, adult cluster positions, the estimated WM volume was significantly lower and CSF volume was significantly higher when using adaptive cluster tracking.

Conclusions

Using adaptive tissue cluster tracking the differences in R1 and R2 relaxation rates and proton density between young children and adults can be corrected for, allowing fully automatic brain segmentation on all ages.

Acknowledgements

No acknowledgement found.

References

No reference found.

Figures

Fig. 1. The automatically found tissue cluster positions in the first 5 years of life.

Fig.2. Segmented WM and CSF of an 8-month old and a 23 year old subject using the same algorithm. The red line outlines the intracranial volume, WM is in blue, CSF in pink.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
1401