Browse by Year

An ANN-based method to optimize the parameters and dramatically reduce diagnostic errors from cutting-edge MRI-based experimental medical diagnosis techniques for the brain
by Edna Zheng

March 28, 2024

Abstract

The spin-lattice relaxation time in a rotating frame(T1rho) is a tool that is currently under investigation. T1rho can map macromolecular properties within a tissue, giving experts new diagnostic opportunities. Current mapping techniques are error-prone due to signal evolution during image data acquisition. The neural network (NN) overcomes these errors, improves robustness, and accomplishes greater accuracy.

AI with NN algorithms was applied in the acceleration of T1rho, running on the testing and training data to find the optimal parameter for T1rho; it was evaluated and measured in terms of Normalized Absolute Deviation (NAD).  

The parameters differ for each trial; the NN ran multiple times to ensure accuracy. As the number of iterations increases, accuracy increases. Minimizing the number of iterations is pivotal because a large number of iterations can lead to overfitting, and increased run time. The parameters of the testing data between 10-12 iterations were similar and closely aligned with that of the training data. Hence, having ten iterations satisfies the demands to find the best parameter for accuracy ensuring minimal computational cost and exposure time of MRI radiation. 

The data has shown that the parameter at the tenth iteration accelerates T1rho mapping without sacrificing accuracy. This project demonstrates the application of the NN on the brain, which may be possible in other parts of the body.

References

  1. Burke, J. F., Skolarus, L. E., Callaghan, B. C., & Kerber, K. A. (2013). Choosing Wisely: highest-cost tests in outpatient neurology. Annals of neurology, 73(5), 679–683. https://doi.org/10.1002/ana.23865 
  2. Chen W. (2015). Errors in quantitative T1rho imaging and the correction methods. Quantitative imaging in medicine and surgery, 5(4), 583–591. https://doi.org/10.3978/j.issn.2223-4292.2015.08.05
  3. Wang, Y. X., Zhao, F., Yuan, J., Mok, G. S., Ahuja, A. T., & Griffith, J. F. (2013). Accelerated T1rho relaxation quantification in intervertebral disc using limited spin-lock times. Quantitative imaging in medicine and surgery, 3(1), 54–58. https://doi.org/10.3978/j.issn.2223-4292.2013.02.09.
  4. Zibetti, M. V. W., Sharafi, A., Otazo, R., & Regatte, R. R. (2018). Accelerating 3D-T1ρ mapping of cartilage using compressed sensing with different sparse and low rank models. Magnetic resonance in medicine, 80(4), 1475–1491. https://doi.org/10.1002/mrm.27138
  5. What causes young-onset dementia? Alzheimer's Society. (2022, October 10). Retrieved November 9, 2022, from https://www.alzheimers.org.uk/about-dementia/types-dementia/what-causes-young-onset-dementia 
  6. Menon, R. G., Zibetti, M. V. W., & Regatte, R. R. (2023). Data-driven optimization of sampling patterns for MR brain T1ρ mapping. Magnetic resonance in medicine, 89(1), 205–216. https://doi.org/10.1002/mrm.29445
  7. Zibetti, M. V. W., Sharafi, A., Otazo, R., & Regatte, R. R. (2020). Accelerated mono- and biexponential 3D-T1ρ relaxation mapping of knee cartilage using golden angle radial acquisitions and compressed sensing. Magnetic resonance in medicine, 83(4), 1291–1309. https://doi.org/10.1002/mrm.28019
  8. Menon, R. G., Zibetti, M. V. W., Jain, R., Ge, Y., & Regatte, R. R. (2021). Performance Comparison of Compressed Sensing Algorithms for Accelerating T1ρ Mapping of Human Brain. Journal of magnetic resonance imaging : JMRI, 53(4), 1130–1139. https://doi.org/10.1002/jmri.27421