Encrypted | Login
The TrueCT challenge has concluded. This challenge provided 200 simulated sinogram CT images of virtual subject with three types of abnormality (i.e., COPD, lung lesion, liver lesion) utilizing the Center for Virtual Imaging Trials resources. The challenge provided an objective performance evaluation of CT reconstruction methods against the ground truth of the digital subjects.

The top five performing individuals and teams are:

  1. Team name: YouXia512
    Members: Xiao Wang, Isaac Lyngaas, Muralikrishnan Gopalakrishnan Meena, Amir Ziabari, Singanallur Venkatakrishnan, Balint Joo, Matthew Norman, Gregery Buzzard, Charles Bouman, Anuj Kapadia, Tom Beck,
    Institution: Oak Ridge National Laboratory, TN, China.
    Final score: ((normalized(SSIM) + (1-normalized(RMSE)))/2 = 0.665.

  2. Team name: WeiWang.
    Members: Wei Wang.
    Institution: Shenzhen University, Shenzhen, China.
    Final score = 0.564.

  3. Team name: MCR Toolkit.
    Members: Darin Clark, Christian Badea.
    Institution: Duke University, NC, USA
    Final score = 0.541.

  4. Team name: FashionImaging.
    Members: Siqi Ye, Liyue Shen, Lei Xing
    Institution: Stanford University, CA, USA
    Final score = 0.452.

  5. Team name: gljRecon.
    Members: Giavanna Jadick
    Institution: University of Chicago, IL, USA
    Final score = 0.429.

The received reconstructed images were evaluated based on task generic metrics that reflect their similarity to the ground truth of the virtual subjects. These metrics were root mean squared error and structural similarity index, both measured in the organ/abnormality of interest. For the COPD cases, which is a diffused disease, the metrics were measured in the entire lungs. For the lung nodule and liver lesion cases, the metrics were measured in a VOI around the lesions. Each metric was normalized to a [0 1] score across all the participants per abnormality. The final score was the average of the normalized scores.

Articles