Tensor Gradient L₀-Norm Minimization-Based Low-Dose CT and Its Application to COVID-19

Abstract

Methods to recover high-quality computed tomography (CT) images in low-dose cases will be of great benefit. To reach this goal, sparse-data subsampling is one of the common strategies to reduce radiation dose, which is attracting interest among the researchers in the CT community. Since analytic image reconstruction algorithms may lead to severe image artifacts, the iterative algorithms have been developed for reconstructing images from sparsely sampled projection data. In this study, we first develop a tensor gradient L₀-norm minimization (TGLM) for low-dose CT imaging. Then, the TGLM model is optimized by using the split-Bregman method. The Coronavirus Disease 2019 (COVID-19) has been sweeping the globe, and CT imaging has been deployed for detection and assessing the severity of the disease. Finally, we first apply our proposed TGLM method for COVID-19 to achieve low-dose scan by incorporating the 3-D spatial information. Two COVID-19 patients (64 years old female and 56 years old man) were scanned by the [Formula: see text]CT 528 system, and the acquired projections were retrieved to validate and evaluate the performance of the TGLM.

Keywords: Chest CT; Coronavirus Disease 2019 (COVID-19); low-dose computed tomography (CT); tensor gradient L₀-norm.

Fig. 1.

Imaging system. (a) Photograph of a CT 528 system. (b) Illustration of helical scanning geometry.

Fig. 2.

(a) and (b) Two representative slices from healthy CT human (c)–(f) Representative image features from four different COVID-19 patients within appropriate display widows. Note that these COVID-19 patients are obtained from the People’s Hospital of China Three Gorges University (Yichang), Hubei, China.

Fig. 3.

Recovering results of three representative CT slices (40th, 100th, and 140th slices) of patient #1 using TGLM method. (a)–(c) and (a1)–(c1) represent the SIRT and TGLM results from an undersampling factor 12.

Fig. 4.

Reconstructed results for three representative axial slices. First–third columns represent 70th, 105th, and 135th slice, and first–fourth row represent the ground truth, SIRT, TVM, and TGLM results and the window is [−1000, 200] HU.

Fig. 5.

Magnified ROI A in Fig. 4. First–fourth columns represent ground truth, SIRT, TVM, and TGLM and their windows are [−900, 200] HU.

Fig. 6.

Similar to Fig. 5 but from Magnified ROI B of Fig. 4. The windows are [−900, 200] HU.

Fig. 7.

First and second rows represent the magnified ROIs C and D. First–fourth columns represent reference, SIRT, TVM, and TGLM and the window is [−1000, 0] HU.

Fig. 8.

Similar to Fig. 5 but from magnified ROI E of Fig. 4. The windows is [−1000, 0] HU.

Fig. 9.

Reconstructed results from the 370th coronal slice. The first–fourth rows represent the ground truth, SIRT, TVM, and TGLM results. The first column represents reconstructed results and the second–third columns are three ROIs, i.e., ROIs “F” and “H” and their display windows are [−1000, 0], [−200, 200], [−1000, 0], and [−200, 200] HU.

Fig. 10.

Reconstructed results from 170th sagittal slice. The first–fourth rows represent the ground truth, SIRT, TVM, and TGLM results. The second column is magnified ROI “I. ” The windows of the first–second columns are [−900, 100] HU.

Fig. 11.

Quantitative results of all axial slices using different reconstruction methods in terms of RMSE, SSIM, and FSIM.

Fig. 12.

Parameters comparison in terms of RMSE and SSIM. (a) and (b), (c) and (d), (e) and (f) RMSEs and SSIMs with different sets of the parameter , , and , respectively.

Fig. 13.

Convergence curves in terms of RMSEs versus iterations.

Fig. 14.

Reconstructed results for two representative axial slices of case 2. The first–second columns represent 100th and 115th slices, and the first–fourth rows represent the ground truth, SIRT, TVM, and TGLM results and the window is [−1000, 200] HU.

Fig. 15.

Reconstructed results for two representative axial slices of case 2. The first–third columns represent 80th, 90, and 135th slices, and the first–fourth rows represent the ground truth, SIRT, TVM, and TGLM results and the window is [−1000, 200] HU.

Fig. 16.

First and second columns represent reconstructed results from 360th coronal and 360th sagittal slice and their windows are [−1000, 200] and [−1000, 0] HU. The first–fourth rows represent the ground truth, SIRT, TVM, and TGLM results.