Dual-energy CT of the brain: Comparison between DECT angiography-derived virtual unenhanced images and true unenhanced images in the detection of intracranial haemorrhage

Dual-Energy CT in Acute Stroke: Could Non-Contrast CT Be Replaced by Virtual Non-Contrast CT? A Feasibility Study

Purpose: We aimed to evaluate whether virtual non-contrast cerebral computed tomography (VNCCT) reconstructed from intravenous contrast-enhanced dual-energy CT (iv-DECT) could replace non-contrast CT (NCCT) in patients with suspected acute cerebral ischemia. Method: This retrospective study included all consecutive patients in whom NCCT followed by iv-DECT were performed for suspected acute ischemia in our emergency department over a 1-month period. The Alberta Stroke Program Early CT Score (ASPECTS) was used to determine signs of acute ischemia in the anterior and posterior circulation, the presence of hemorrhage, and alternative findings, which were randomly evaluated via the consensus reading of NCCT and VNCCT by two readers blinded to the final diagnosis. An intraclass correlation between VNCCT and NCCT was calculated for the ASPECTS values. Both techniques were evaluated for their ability to detect ischemic lesions (ASPECTS <10) when compared with the final discharge diagnosis (reference standard). Results: Overall, 148 patients (80 men, mean age 64 years) were included, of whom 46 (30%) presented with acute ischemia, 6 (4%) presented with intracerebral hemorrhage, 11 (7%) had an alternative diagnosis, and 85 (59%) had no pathological findings. The intraclass correlation coefficients of the two modalities were 0.97 (0.96-0.98) for the anterior circulation and 0.77 (0.69-0.83) for the posterior circulation. The VNCCT's sensitivity for detecting acute ischemia was higher (41%, 19/46) than that of NCCT (33%, 15/46). Specificity was similar between the two techniques, at 94% (97/103) and 98% (101/103), respectively. Conclusions: Our results show that VNCCT achieved a similar diagnostic performance as NCCT and could, thus, replace NCCT in assessing patients with suspected acute cerebral ischemia.

Gemstone Spectral CT Virtual Noncontrast Images and Iodine Maps for the Characterization of Thyroid Lesions and Distinguishing Thyroid Papillary Carcinoma from Nodular Goiter

BACKGROUND

Gemstone spectral contrast-enhanced CT with virtual noncontrast (VNC) images and iodine maps can potentially reduce the number of required CT scans for thyroid lesions. However, data regarding the clinical utility of VNC images and iodine maps in characterizing thyroid lesions and distinguishing thyroid papillary carcinoma from nodular goiter are still limited.

PURPOSE

To determine whether VNC images and iodine density could reliably aid in characterizing thyroid lesions and distinguishing thyroid papillary carcinoma from nodular goiter compared with true noncontrast (TNC) images.

METHODS

This retrospective study included patients with thyroid papillary carcinoma or nodular goiter who underwent TNC and contrast-enhanced gemstone spectral CT scans. The consistency of qualitative parameters, including intralesional calcification, necrosis, lesion boundary, thyroid edge interruption, and lymph node metastasis, between TNC and VNC images, was analyzed using the kappa statistic. TNC attenuation, VNC attenuation, absolute attenuation between TNC and VNC, and iodine density were compared between thyroid papillary carcinoma and nodular goiter by using Student's t-test. The diagnostic performance for distinguishing papillary carcinoma from nodular goiter was evaluated by using the area under the receiver operating characteristic curve (AUC) value, sensitivity, and specificity.

RESULTS

VNC and TNC imaging showed comparable performance in delineating calcification, necrosis, lesion boundary, thyroid edge interruption, and lymph node metastasis (all k > 0.75). Papillary carcinoma showed significantly lower absolute attenuation between VNC and TNC than nodular goiter (7.86 ± 6.74 vs. 13.43 ± 10.53, P=0.026), which was similarly observed for iodine density (31.45 ± 8.51 vs. 37.27 ± 10.34, P=0.016). The iodine density showed higher diagnostic performance (AUC = 0.727), accuracy (0.773 vs. 0.667), sensitivity (0.750 vs. 0.708), and specificity (0.786 vs. 0.643) than the absolute attenuation between TNC and VNC images (AUC = 0.683).

CONCLUSIONS

VNC imaging, a promising substitute for TNC imaging, has comparable diagnostic efficacy for reliably characterizing thyroid lesions. Iodine density could be valuable for distinguishing thyroid papillary carcinoma from nodular goiter.

Detection of Early Ischemic Changes with Virtual Noncontrast Dual-Energy CT in Acute Ischemic Stroke: A Noninferiority Analysis

BACKGROUND AND PURPOSE

Dual-energy virtual NCCT has the potential to replace conventional NCCT to detect early ischemic changes in acute ischemic stroke. In this study, we evaluated whether virtual NCCT is noninferior compared with standard linearly blended NCCT, a surrogate of conventional NCCT, regarding the detection of early ischemic changes with ASPECTS.

MATERIALS AND METHODS

Adult patients who presented with suspected acute ischemic stroke and who underwent dual-energy NCCT and CTA and brain MR imaging within 48 hours were included. Standard linearly blended images were reconstructed to match a conventional NCCT. Virtual NCCT images were reconstructed from CTA. ASPECTS was evaluated on conventional NCCT, virtual NCCT, and DWI, which served as the reference standard. Agreement between CT assessments and the reference standard was evaluated with the Lin concordance correlation coefficient. Noninferiority was assessed with bootstrapped estimates of the differences in ASPECTS between conventional and virtual NCCT with 95% CIs.

RESULTS

Of the 193 included patients, 100 patients (52%) had ischemia on DWI. Compared with the reference standard, the ASPECTS concordance correlation coefficient for conventional and virtual NCCT was 0.23 (95% CI, 0.15-0.32) and 0.44 (95% CI, 0.33-0.53), respectively. The difference in the concordance correlation coefficient between virtual and conventional NCCT was 0.20 (95% CI, 0.01-0.39) and did not cross the prespecified noninferiority margin of -0.10.

CONCLUSIONS

Dual-energy virtual NCCT is noninferior compared with conventional NCCT for the detection of early ischemic changes with ASPECTS.

Simultaneous vessel segmentation and unenhanced prediction using self-supervised dual-task learning in 3D CTA (SVSUP)

BACKGROUND AND OBJECTIVE

The vessel segmentation in CT angiography (CTA) provides an important basis for automatic diagnosis and hemodynamics analysis. Virtual unenhanced (VU) CT images obtained by dual-energy CT can assist clinical diagnosis and reduce radiation dose by obviating true unenhanced imaging (UECT). However, accurate segmentation of all vessels in the head-neck CTA (HNCTA) remains a challenge, and VU images are currently not available from conventional single-energy CT imaging.

METHODS

In this paper, we proposed a self-supervised dual-task deep learning strategy to fully automatically segment all vessels and predict unenhanced CT images from single-energy HNCTA based on a developed iterative residual-sharing scheme. The underlying idea was to use the correlation between the two tasks to improve task performance while avoiding manual annotation for model training.

RESULTS

The feasibility of the strategy was verified using the data of 24 patients. For vessel segmentation task, the proposed model achieves a significantly higher average Dice coefficient (84.83%, P-values 10^-3 in paired t-test) than the state-of-the-art segmentation model, vanilla VNet (78.94%), and several popular 3D vessel segmentation models, including Hessian-matrix based filter (62.59%), optically-oriented flux (66.33%), spherical flux model (66.91%), and deep vessel net (66.47%). For the unenhanced prediction task, the average ROI-based error compared to the UECT in the artery tissue is 6.1±4.5 HU, similar to previously reported 6.4±5.1 HU for VU reconstruction.

CONCLUSIONS

Results show that the proposed dual-task framework can effectively improve the accuracy of vessel segmentation in HNCTA, and it is feasible to predict the unenhanced image from single-energy CTA, providing a potential new approach for radiation dose saving. Moreover, to our best knowledge, this is the first reported annotation-free deep learning-based full-image vessel segmentation for HNCTA.

ASPECTS estimation using dual-energy CTA-derived virtual non-contrast in large vessel occlusion acute ischemic stroke: a dose reduction opportunity for patients undergoing repeat CT?

PURPOSE

Recent studies have shown the feasibility of dual-energy CT (DECT) virtual non-contrast (VNC) for determining infarct extent. In this study, patients presenting with large-vessel occlusion (LVO) acute ischemic stroke (AIS), we assess whether ASPECTS on DECTA-VNC differs from non-contrast CT (NCCT).

METHODS

After IRB approval, LVO-AIS patients undergoing NCCT and DECTA between October 2016 and September 2018 were retrospectively reviewed. DECTA-VNC images were derived using Syngo.via (Siemens, Erlangen, Germany). ASPECTS was scored by two blinded neuroradiologists. Square-weighted kappa statistic, diagnostic performance, Wilcoxon signed-rank tests between groups, and CT doses were calculated.

RESULTS

Fifty-one patients met inclusion criteria, with median age of 76 (IQR 67-82); 26/51 (51%) were female. Median time between last-known-well and CT was 120 min (IQR 60-252). DECTA-VNC ASPECTS score differed by ≤ 1 from consensus NCCT in 49/51 (96%) patients for reader 1 and in 46/51 (90%) for reader 2. ASPECTS on DECTA-SI and consensus NCCT differed by ≤ 1 in 45/51 (88%) for both readers. On a per ASPECTS-region basis, DECTA-VNC had 87% sensitivity, 95% specificity, 0.82% PPV, and 0.96% NPV. ASPECTS inter-rater agreement was highest for DECTA-VNC (κ = 0.71), DECTA-SI (κ = 0.48), and NCCT (κ = 0.40). NCCT median CTDIvol was 63.7 mGy (IQR 60.7-67.2); DLP was 1060.0 mGy·cm (IQR 981.0-1151.5). DECTA-VNC dose was lower: median CTDIvol was 20.9 mGy (IQR 19.8-22.2); DLP was 804.1 (IQR 691.6-869.4), p < 0.0001.

CONCLUSION

DECTA-derived VNC yielded similar ASPECTS scores as NCCT and is therefore non-inferior in early ischemia-related low attenuation edema/infarct detection in acute LVO-AIS patients. Further evaluation of the role of DECTA-VNC in AIS imaging is warranted.

Prediction of Intracerebral Hemorrhage After Endovascular Treatment of Acute Ischemic Stroke: Combining Quantitative Parameters on Dual-Energy CT with Clinical Related Factors

OBJECTIVES

To evaluate the predictive value of dual-energy CT (DECT) quantitative parameters and clinical influence factors for intracerebral hemorrhage (ICH) complications after endovascular treatment in patients with acute ischemic stroke (AIS).

METHODS

Seventy-two consecutive patients who underwent brain DECT immediately after endovascular treatment for AIS from November 2017 to October 2019 were included. Retrospectively, the volume of brain parenchymal hyperdensity area (HDA), the maximum iodine concentration, and maximum CT value on DECT images was evaluated and measured by two radiologists blinded to any clinical information independently. Follow-up CT imaging (24-72 h) were used to assess the development of ICH complications. DECT parameters and clinical influence factors were analyzed by Chi-square test or Fisher's exact test and Mann-Whitney U test. Receiver operating characteristic curves were generated for continuous variables.

RESULTS

Follow-up CT images confirmed that forty of 72 patients (55.6%) developed ICH. The volume of HDA, median maximum iodine concentration and maximum CT value between ICH group and non-ICH group were significantly different (P < 0.001). Combining the DECT quantitative parameters with clinical predictors, receiver operating characteristic analysis revealed an area under the curve of 0.985, for identifying patients developing ICH with sensitivity, specificity, positive predictive value and negative predictive value were 90%, 100%, 100% and 88.9%, respectively.

CONCLUSIONS

Three quantitative parameters of DECT and clinical predictors showed great predictive performance in identifing ICH complications in patients with brain parenchyma HDA after endovascular therapy, which may contribute to better clinical decision-making.

Precise diagnosis of intracranial hemorrhage and subtypes using a three-dimensional joint convolutional and recurrent neural network

OBJECTIVES

To evaluate the performance of a novel three-dimensional (3D) joint convolutional and recurrent neural network (CNN-RNN) for the detection of intracranial hemorrhage (ICH) and its five subtypes (cerebral parenchymal, intraventricular, subdural, epidural, and subarachnoid) in non-contrast head CT.

METHODS

A total of 2836 subjects (ICH/normal, 1836/1000) from three institutions were included in this ethically approved retrospective study, with a total of 76,621 slices from non-contrast head CT scans. ICH and its five subtypes were annotated by three independent experienced radiologists, with majority voting as reference standard for both the subject level and the slice level. Ninety percent of data was used for training and validation, and the rest 10% for final evaluation. A joint CNN-RNN classification framework was proposed, with the flexibility to train when subject-level or slice-level labels are available. The predictions were compared with the interpretations from three junior radiology trainees and an additional senior radiologist.

RESULTS

It took our algorithm less than 30 s on average to process a 3D CT scan. For the two-type classification task (predicting bleeding or not), our algorithm achieved excellent values (≥ 0.98) across all reporting metrics on the subject level. For the five-type classification task (predicting five subtypes), our algorithm achieved > 0.8 AUC across all subtypes. The performance of our algorithm was generally superior to the average performance of the junior radiology trainees for both two-type and five-type classification tasks.

CONCLUSIONS

The proposed method was able to accurately detect ICH and its subtypes with fast speed, suggesting its potential for assisting radiologists and physicians in their clinical diagnosis workflow.

KEY POINTS

• A 3D joint CNN-RNN deep learning framework was developed for ICH detection and subtype classification, which has the flexibility to train with either subject-level labels or slice-level labels. • This deep learning framework is fast and accurate at detecting ICH and its subtypes. • The performance of the automated algorithm was superior to the average performance of three junior radiology trainees in this work, suggesting its potential to reduce initial misinterpretations.

Unenhanced Dual-Energy Computed Tomography: Visualization of Brain Edema

PURPOSE

The aim of this study was to determine whether dual-energy computed tomography (DECT) imaging is superior to conventional noncontrast computed tomography (CT) imaging for the detection of acute ischemic stroke.

MATERIALS AND METHODS

This was a retrospective, single-center study of 40 patients who presented to the emergency department (ED) of a major, acute care, teaching center with signs and symptoms of acute stroke. Only those patients who presented to the ED within 4 hours of symptom onset were included in this study. All 40 patients received a noncontrast DECT of the head at the time of presentation. Each patient also received standard noncontrast CT of the head 24 hours after their initial presentation to the ED. "Brain edema" images were then reconstructed using 3-material decomposition with parameters adjusted to suppress gray/white matter contrast while preserving edema and increasing its conspicuity. The initial unenhanced, mixed images, brain edema, and 24-hour follow-up true noncontrast (TNC) images were reviewed and assigned Alberta Stroke Program Early CT scores. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated.

RESULTS

Of the 40 patients, 28 (70%) were diagnosed with an acute infarction. Brain edema reconstructions were better able to predict end infarction volume, with Alberta Stroke Program Early CT scores similar to the 24-hour follow-up TNC CT (7.75 vs 7.7; P > 0.05), whereas the mixed images routinely underestimated the extent of infarction (8.975 vs 7.7; P < 0.001). Initial TNC images had a sensitivity, specificity, PPV, and NPV of 80% (95% confidence interval [CI], 51.9%-95.7%), 72.7% (95% CI, 39%-94%), 80% (95% CI, 51.9%-95.7%), and 72.73% (95% CI, 51.91%-95.67%), respectively. The DECT brain edema images provided a sensitivity, specificity, PPV, and NPV of 93.33% (95% CI, 68.05%-99.83%), 100% (95% CI, 71.51%-100%), 100% (95% CI, 76.84%-100%), and 91.67% (95% CI, 61.52%-99.79%), respectively. There was very good interrater reliability across all 3 imaging techniques.

CONCLUSION

Brain edema reconstructions are able to more accurately detect edema and end-infarct volume as compared with initial TNC images. This provides a better assessment of the degree and extent of infarction and may serve to better guide therapy in the future.

Dual-layer spectral detector CT monoenergetic reconstruction improves image quality of non-contrast cerebral CT as compared with conventional single energy CT

PURPOSE

To investigate and compare image quality of monoenergetic reconstructions from spectral NCCT to conventional 120 kVp images acquired at a similar dose.

MATERIALS AND METHODS

Patients undergoing NCCT on a dual-layer spectral detector CT (n = 30) and a conventional CT (n = 30) were enrolled in the study. The spectral detector CT data was reconstructed at monoenergetic images from 40 to 140 keV in 5-keV increments and 65-70 keV in 1-keV increments (Group A1) and using single energy CT equivalent reconstruction (Group A2). The reference conventional 120kVp images (Group B) were acquired using a standard-of-care protocol with the same radiation dose. We evaluated the image quality of monoenergetic images and determined the optimal keV level using HU attenuation, noise, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), artifact evaluation in posterior fossa by placing region-of-interest (ROI) and subjective image score by 2 radiologists independently using a 4-point scale (1-excellent, 4-undiagnostic).

RESULTS

The SNR and subjective image score were optimal at 66-70keV, while monoenergetic 68 keV images with a higher SNR (18.48 ± 1.94, 15.55 ± 1.56 and 14.33 ± 1.38 for Group 68keV, A2 and B respectively, p < 0.001), CNR (4.09 ± 0.65, 3.43 ± 0.56 and 3.52 ± 0.55 for Group 68keV, A2 and B respectively, p < 0.001) and a lower noise (1.80 ± 0.19, 2.11 ± 0.19 and 2.25 ± 0.25 for Group 68keV, A2 and B respectively, p < 0.001).

CONCLUSION

Spectral NCCT monoenergetic reconstructions at 68 keV improve image quality and reduce artifact compared to conventional single energy CT without radiation dose penalty.

Iodine Extravasation Quantification on Dual-Energy CT of the Brain Performed after Mechanical Thrombectomy for Acute Ischemic Stroke Can Predict Hemorrhagic Complications

BACKGROUND AND PURPOSE

Intracerebral hemorrhage represents a potentially severe complication of revascularization of acute ischemic stroke. The aim of our study was to assess the capability of iodine extravasation quantification on dual-energy CT performed immediately after mechanical thrombectomy to predict hemorrhagic complications.

MATERIALS AND METHODS

Because this was a retrospective study, the need for informed consent was waived. Eighty-five consecutive patients who underwent brain dual-energy CT immediately after mechanical thrombectomy for acute ischemic stroke between August 2013 and January 2017 were included. Two radiologists independently evaluated dual-energy CT images for the presence of parenchymal hyperdensity, iodine extravasation, and hemorrhage. Maximum iodine concentration was measured. Follow-up CT examinations performed until patient discharge were reviewed for intracerebral hemorrhage development. The correlation between dual-energy CT parameters and intracerebral hemorrhage development was analyzed by the Mann-Whitney U test and Fisher exact test. Receiver operating characteristic curves were generated for continuous variables.

RESULTS

Thirteen of 85 patients (15.3%) developed hemorrhage. On postoperative dual-energy CT, parenchymal hyperdensities and iodine extravasation were present in 100% of the patients who developed intracerebral hemorrhage and in 56.3% of the patients who did not (P = .002 for both). Signs of bleeding were present in 35.7% of the patients who developed intracerebral hemorrhage and in none of the patients who did not (P < .001). Median maximum iodine concentration was 2.63 mg/mL in the patients who developed intracerebral hemorrhage and 1.4 mg/mL in the patients who did not (P < .001). Maximum iodine concentration showed an area under the curve of 0.89 for identifying patients developing intracerebral hemorrhage.

CONCLUSIONS

The presence of parenchymal hyperdensity with a maximum iodine concentration of >1.35 mg/mL may identify patients developing intracerebral hemorrhage with 100% sensitivity and 67.6% specificity.

Neuroprotection Exerted by Netrin-1 and Kinesin Motor KIF1A in Secondary Brain Injury following Experimental Intracerebral Hemorrhage in Rats

Binding of extracellular netrin-1 to its receptors, deleted in colorectal cancer (DCC) and uncoordinated gene 5H2 (UNC5H2), inhibits apoptosis mediated by these receptors. A neuron-specific kinesin motor protein, KIF1A, has been shown to participate in netrin-1 secretion. This study aimed to identify the roles of netrin-1 and KIF1A in secondary brain injury after intracerebral hemorrhage (ICH) and the potential mechanisms. An autologous blood ICH model was established in adult male Sprague-Dawley rats, and cultured neurons were exposed to OxyHb to mimic ICH conditions in vitro. Mouse recombinant netrin-1, expression vectors encoding KIF1A, and KIF1A-specific siRNAs were administered intracerebroventricularly. After ICH, protein levels of netrin-1, DCC, and UNC5H2 increased, while protein levels of KIF1A decreased. Levels of UNC5H2 and DCC bound to netrin-1 increased after ICH but were significantly lower than the increase in total amount of protein. Administration of recombinant netrin-1 attenuated neuronal apoptosis and degeneration in ICH rats. Moreover, KIF1A overexpression increased concentrations of netrin-1 in cerebrospinal fluid and cell culture supernatant and exerted neuroprotective effects via netrin-1 and its receptor pathways. KIF1A plays a critical role in netrin-1 secretion by neurons. An increase in protein levels of netrin-1 may be a neuroprotective strategy after ICH. However, this process is almost completely abolished by ICH-induced loss of KIF1A. An exogenous increase of KIF1A may be a potential strategy for neuroprotection via the netrin-1 pathway.