Retrospective study of deep learning to reduce noise in non-contrast head CT images

PURPOSE

Presented herein is a novel CT denoising method uses a skip residual encoder-decoder framework with group convolutions and a novel loss function to improve the subjective and objective image quality for improved disease detection in patients with acute ischemic stroke (AIS).

MATERIALS AND METHODS

In this retrospective study, confirmed AIS patients with full-dose NCCT head scans were randomly selected from a stroke registry between 2016 and 2020. 325 patients (67 ± 15 years, 176 men) were included. 18 patients each with 4-7 NCCTs performed within 5-day timeframe (83 total scans) were used for model training; 307 patients each with 1-4 NCCTs performed within 5-day timeframe (380 total scans) were used for hold-out testing. In the training group, a mean CT was created from the patient's co-registered scans for each input CT to train a rotation-reflection equivariant U-Net with skip and residual connections, as well as a group convolutional neural network (SRED-GCNN) using a custom loss function to remove image noise. Denoising performance was compared to the standard Block-matching and 3D filtering (BM3D) method and RED-CNN quantitatively and visually. Signal-to-noise ratio (SNR) and contrast-to-noise (CNR) were measured in manually drawn regions-of-interest in grey matter (GM), white matter (WM) and deep grey matter (DG). Visual comparison and impact on spatial resolution were assessed through phantom images.

RESULTS

SRED-GCNN reduced the original CT image noise significantly better than BM3D, with SNR improvements in GM, WM, and DG by 2.47x, 2.83x, and 2.64x respectively and CNR improvements in DG/WM and GM/WM by 2.30x and 2.16x respectively. Compared to the proposed SRED-GCNN, RED-CNN reduces noise effectively though the results are visibly blurred. Scans denoised by the SRED-GCNN are shown to be visually clearer with preserved anatomy.

CONCLUSION

The proposed SRED-GCNN model significantly reduces image noise and improves signal-to-noise and contrast-to-noise ratios in 380 unseen head NCCT cases.

Utility of Hounsfield unit in the diagnosis of tandem occlusion in acute ischemic stroke

Background

Tandem occlusions can complicate medical and endovascular stroke treatment. To identify these occlusions, computed tomography angiography (CTA) represents the best imaging modality. However, CTA is still not initially performed in some patients not admitted directly to stroke centers. Early identification of an additional occlusion of the proximal extracranial internal carotid artery may improve the best suitable treatment strategy. The purpose of this study was to find a valuable threshold of thrombus attenuation in a non-contrast head CT (NCCT) scan to facilitate a safe diagnosis of tandem occlusions.

Materials and methods

Consecutive patients with acute middle cerebral artery (MCA) occlusions who underwent endovascular treatment were identified from our registry of neuroendovascular interventions. Thrombus attenuations of the affected MCA and contralateral vessel were measured by NCCT. To compare individual baseline blood attenuations, the difference between the thrombus attenuation and the contralateral MCA attenuation (referred to as ΔTM) was calculated.

Results

Three hundred and twenty-five patients were included. There was a highly significant difference between mean thrombus attenuation with isolated MCA occlusion and additional extracranial internal carotid artery (ICA) occlusion (49.9 ± 8 vs. 56.2 ± 10 Hounsfield units (HU); P < 0.001). The area under the receiver operating characteristic curve of ΔTM was 0.72. The optimal threshold value was 13.5 HU, with a sensitivity of 67.5% and a specificity of 68.6%.

Conclusion

Despite a significant difference in thrombus attenuation in MCA occlusions with an additional extracranial ICA occlusion compared with isolated MCA occlusions, a relevant threshold of thrombus attenuation was not found.