Task-based medical imaging and image restoration assessment and optimization

Our lab is actively engaged in objective optimization and assessment of medical imaging and image restoration operators based on its clinical tasks. In medical imaging, images are often acquired for specific purposes and the use of objective measures of image quality has been widely advocated for assessing imaging systems and image processing algorithms. However, the performance of these methods is typically evaluated in terms of traditional image quality metrics such as root mean square error (RMSE), structural similarity index metric (SSIM) or peak signal-to-noise ratio (PSNR). The objective evaluation of modern DNN-based medical image restoration methods remains largely lacking. Our lab has conducted a strategy to assess modern DNN-based denoising methods and super-resolution methods by use of objective IQ measures on their performance on signal detection task. I have developed objective CT simulation optimization model that maps a novel image quality index to organ segmentation accuracy. In addition, we have developed advanced techniques to create stochastic object model to provide the grand truth for the objective evaluation and assessment of these methods.

1. Learning stochastic object model for task-based image quality assessment in radiation therapy

2. Task-based image quality assessment in radiation therapy

3. Objective assessment of DNN-based image denoising method (collaborative project with Professor Anastasio. )

4.Task-based evaluation of deep image super-resolution in medical imaging (collaborative project with Professor Anastasio.)

 

Related publications

  1. Varun A. Kelkar, Xiaohui Zhang, Jason Granstedt, Hua Li, and Mark A. Anastasio “Task-based evaluation of deep image super-resolution in medical imaging“, Proc. SPIE 11599, Medical Imaging 2021: Image Perception, Observer Performance, and Technology Assessment, 115990X.
  2. Kaiyan Li, Weimin Zhou, Hua Li and M. A. Anastasio, “Assessing the Impact of Deep Neural Network-based Image Denoising on Binary Signal Detection Tasks,” in IEEE Transactions on Medical Imaging, doi: 10.1109/TMI.2021.3076810.
  3. Weimin Zhou, Hua Li, Mark A. Anastasio, “ Approximating the Ideal Observer for joint signal detection and localization tasks by use of supervised learning methods”, IEEE Transactions on Medical Imaging, 2020.
  4. Steven Dolly, Yang Lou, Mark Anastasio, Hua Li*,“Task-based image quality assessment in radiation therapy: initial characterization and demonstration with computer-simulation study”, Physics in Medicine & Biology (2019).
  5. Weimin Zhou, Hua Li, Mark Anastasio, “Approximating the Ideal Observer and Hotelling Observer for binary signal detection tasks by use of supervised learning methods”, IEEE Trans. on Medical Imaging, April, 2019, doi:10.1109/TMI.2019.2911211.
  6. Steven Dolly, Yang Lou, Mark Anastasio, Hua Li*, “Learning-based Stochastic Object Models for Use in Optimizing Imaging System“, Physics in Medicine and Biology, 2018, 63(6):065004. doi: 10.1088/1361-6560/aab000.
  7. Hua Li*, Steven Dolly, Hsin-Chen Chen, Mark A. Anastasio, Daniel A. Low, Harold H. Li, Jeff M. Michalski, Wade L. Thorstad, Hiram Gay, Sasa Mutic, “A Comparative Study Based on Image Quality and Clinical Task Performance for CT Reconstruction Algorithms in Radiotherapy“, Journal of Applied Clinical Medical Physics, 2016, Vol.17, No.4, pp: 377-390.
  8. Steven Dolly, Hsin-Chen Chen, Mark A. Anastasio, Sasa Mutic, Hua Li*, “Practical considerations for noise power spectra estimation for simulation CT in radiation therapy“, Journal of Applied Clinical Medical Physics, 2016, Vol.17, No.3, pp: 392-407.