Non-contrast CT Images Research Articles

Inter- and Intrarater Agreement of Spot Sign and Noncontrast CT Markers for Early Intracerebral Hemorrhage Expansion.

Background: The aim of this study was to assess the inter- and intrarater reliability of noncontrast CT (NCCT) markers [Black Hole Sign (BH), Blend Sign (BS), Island Sign (IS), and Hypodensities (HD)] and Spot Sign (SS) on CTA in patients with spontaneous intracerebral hemorrhage (ICH). Methods: Patients with spontaneous ICH at three German tertiary stroke centers were retrospectively included. Each CT scan was rated for four NCCT markers and SS on CTA by two radiology residents. Raters were blind to all demographic and outcome data. Inter- and intrarater agreement was determined by Cohen’s kappa (κ) coefficient and percentage of agreement. Results: Interrater agreement was excellent in 473 included patients, ranging from 96% to 99%. Interrater κ ranged from 0.85 (95% CI [0.78–0.91]) to 0.97 (95% CI [0.94–0.99]) for NCCT markers and 0.93 (95% CI [0.88–0.98]) for SS, all p-values < 0.001. Intrarrater agreement ranged from 96% to 100%, with κ ranging from 0.85 (95% CI [0.78–0.91]) to 1.00 (95% CI [0.10–0.85]) for NCCT markers and 0.96 (95% CI [0.92–1.00]) for SS, all p-values < 0.001. Conclusions: NCCT imaging findings and SS on CTA have good-to-excellent inter- and intrarater reliabilities, with the highest agreement for BH and SS.

CT perfusion core and ASPECT score prediction of outcomes in DEFUSE 3.

The role of Alberta Stroke Program Early CT Score (ASPECTS) for thrombectomy patient selection and prognostication in late time windows is unknown. We compared baseline ASPECTS and core infarction determined by CT perfusion (CTP) as predictors of clinical outcome in the Endovascular Therapy Following Imaging Evaluation for Ischemic Stroke 3 (DEFUSE) 3 trial. We included all DEFUSE 3 patients with baseline non-contrast CT and CTP imaging. ASPECTS and core infarction were determined by the DEFUSE 3 core laboratory. Primary outcome was functional independence (modified Rankin Scale (mRS) ≤2). Secondary outcomes included ordinal mRS shift at 90 days and final core infarction volume. Of the 142 patients, 85 patients (60%) had ASPECTS 8-10 and 57 (40%) had ASPECTS 5-7. Thirty-one patients (36%) with ASPECTS 8-10 and 11 patients (19%) with ASPECTS 5-7 were functionally independent at 90 days (p = 0.03). In the primary and secondary logistic regression analysis, there was no difference in ordinal mRS shift (p = 0.98) or functional independence (mRS ≤ 2; p = 0.36) at 90 days between ASPECTS 8-10 and ASPECTS 5-7 patients. Similarly, primary and secondary logistic regression analyses found no difference in ordinal mRS shift (p = 1.0) or functional independence (mRS ≤ 2; p = 0.87) at 90 days between patients with baseline small core ( < 50 ml) versus medium core (50-70 ml). Higher ASPECTS (8-10) correlated with functional independence at 90 days in the DEFUSE trial. ASPECTS and core infarction volume did not modify the thrombectomy treatment effect, which indicates that patients with a target mismatch profile on perfusion imaging should undergo thrombectomy regardless of ASPECTS or core infarction volume in late time windows.

Automated Hierarchy Evaluation System of Large Vessel Occlusion in Acute Ischemia Stroke.

BackgroundThe detection of large vessel occlusion (LVO) plays a critical role in the diagnosis and treatment of acute ischemic stroke (AIS). Identifying LVO in the pre-hospital setting or early stage of hospitalization would increase the patients’ chance of receiving appropriate reperfusion therapy and thereby improve neurological recovery.MethodsTo enable rapid identification of LVO, we established an automated evaluation system based on all recorded AIS patients in Hong Kong Hospital Authority’s hospitals in 2016. The 300 study samples were randomly selected based on a disproportionate sampling plan within the integrated electronic health record system, and then separated into a group of 200 patients for model training, and another group of 100 patients for model performance evaluation. The evaluation system contained three hierarchical models based on patients’ demographic data, clinical data and non-contrast CT (NCCT) scans. The first two levels of modeling utilized structured demographic and clinical data, while the third level involved additional NCCT imaging features obtained from deep learning model. All three levels’ modeling adopted multiple machine learning techniques, including logistic regression, random forest, support vector machine (SVM), and eXtreme Gradient Boosting (XGboost). The optimal cut-off for the likelihood of LVO was determined by the maximal Youden index based on 10-fold cross-validation. Comparisons of performance on the testing group were made between these techniques.ResultsAmong the 300 patients, there were 160 women and 140 men aged from 27 to 104 years (mean 76.0 with standard deviation 13.4). LVO was present in 130 (43.3%) patients. Together with clinical and imaging features, the XGBoost model at the third level of evaluation achieved the best model performance on testing group. The Youden index, accuracy, sensitivity, specificity, F1 score, and area under the curve (AUC) were 0.638, 0.800, 0.953, 0.684, 0.804, and 0.847, respectively.ConclusionTo the best of our knowledge, this is the first study combining both structured clinical data with non-structured NCCT imaging data for the diagnosis of LVO in the acute setting, with superior performance compared to previously reported approaches. Our system is capable of automatically providing preliminary evaluations at different pre-hospital stages for potential AIS patients.

Nomogram Model for Predicting Hematoma Expansion in Spontaneous Intracerebral Hemorrhage: Multicenter Retrospective Study

To establish a new nomogram model and provide a new theoretical basis for the diagnosis and treatment of spontaneous intracerebral hemorrhage. The clinical data and noncontrast computed tomography images of patients with spontaneous intracerebral hemorrhage in 3 tertiary medical centers were collected continuously. Univariate and binary logistic regression analysis were performed to screen out the independent predictors that were significantly associated with hematoma expansion. The nomogram model was drawn by R programming language. According to the related risk factors of nomogram, decision curve analysis and clinical impact curve were established. The numbers of the 3 cooperative units were 554, 582, and 202, respectively. Island sign, blend sign, swirl sign, intraventricular hemorrhage, history of diabetes, time to baseline computed tomography scan, and baseline hematoma volume were independent predictors of hematoma expansion. Baseline hematoma volume >20 mL (odds ratio, 4.088; 95% confidence interval, 2.802-5.964; P < 0.0001) was the most dangerous factor for predicting hematoma expansion, followed by the time to baseline computed tomography scan ≤1 hour (odds ratio, 4.188; 95% confidence interval, 2.598-6.750; P < 0.0001). Decision curve analysis showed that the net benefit of patients was the highest when nomogram score existed. When the threshold probability was >40%, the prediction probability of hematoma expansion was close to the actual probability. This nomogram model could accurately predict hematoma expansion of spontaneous intracerebral hemorrhage, which provided a theoretical basis for clinicians to intervene in the early stage.

Abstract WP405: Automated Detection of Hemorrhagic Stroke From Non-Contrast Computed Tomography: A Machine Learning Approach

Introduction: Automated neuroimaging analysis is being used increasingly in the acute ischemic stroke (AIS) evaluation. However, current algorithms do not factor in an assessment of intracranial hemorrhage (ICH) in the workflow. In this study we present a machine learning (ML) algorithm that uses brain symmetry information to detect ICH. Methods: We prospectively collected non-contrast CT (NCCT) images on patients that presented to the Emergency Department for AIS evaluation between 2017 and 2019. Patients were included if they underwent technically adequate NCCT imaging. Diagnoses of ICH, AIS and non-stroke were confirmed by experienced neuroradiologists as well as review of the clinical record. A ML algorithm which integrates symmetry features as well as standard features for the whole brain was trained on 80% of the sample and validated on the remaining images. Training was performed without any prior segmentation. Evaluation of the model performance was conducted using receiver-operator curve and area under the curve (AUC) analysis. Results are given as median [IQR] and [AUC 95% CI]. Results: Among the 568 patients that met inclusion criteria, median age was 65 [55-76], 47% were female and 34% were white. 128 (23%) patients were determined to have ICH and 440 as non-ICH (70% AIS and 30% non-stroke). Among ICH patients, 108 (84%) had a supratentorial ICH. When analyzing the regions of the CT images that most strongly contributed to the algorithm’s diagnostic decisions, they corresponded with the regions of ICH (Fig. 1A). On the external validation data set, the algorithm successfully detected ICH (Fig. 1B) with high accuracy (AUC 0.99 [0.97-1.00]). Conclusion: We have developed a symmetry-sensitive ML method that can with very high fidelity identify ICH in an automated fashion. Without prior training, the algorithm autonomously was able to learn ICH location. These results may help contribute to an automated imaging workflow for all stroke evaluations, not just AIS.

Biomarker Localization From Deep Learning Regression Networks.

Biomarker estimation methods from medical images have traditionally followed a segment-and-measure strategy. Deep-learning regression networks have changed such a paradigm, enabling the direct estimation of biomarkers in databases where segmentation masks are not present. While such methods achieve high performance, they operate as a black-box. In this work, we present a novel deep learning network structure that, when trained with only the value of the biomarker, can perform biomarker regression and the generation of an accurate localization mask simultaneously, thus enabling a qualitative assessment of the image locus that relates to the quantitative result. We showcase the proposed method with three different network structures and compare their performance against direct regression networks in four different problems: pectoralis muscle area (PMA), subcutaneous fat area (SFA), liver mass area in single slice computed tomography (CT), and Agatston score estimated from non-contrast thoracic CT images (CAC). Our results show that the proposed method improves the performance with respect to direct biomarker regression methods (correlation coefficient of 0.978, 0.998, and 0.950 for the proposed method in comparison to 0.971, 0.982, and 0.936 for the reference regression methods on PMA, SFA and CAC respectively) while achieving good localization (DICE coefficients of 0.875, 0.914 for PMA and SFA respectively, p < 0.05 for all pairs). We observe the same improvement in regression results comparing the proposed method with those obtained by quantify the outputs using an U-Net segmentation network (0.989 and 0.951 respectively). We, therefore, conclude that it is possible to obtain simultaneously good biomarker regression and localization when training biomarker regression networks using only the biomarker value.

Relationship between coronary arterial 18F-sodium fluoride uptake and epicardial adipose tissue analyzed using computed tomography.

18F-Sodium fluoride (18F-NaF) positron emission tomography (PET) has the potential to detect high-risk coronary plaques. Epicardial adipose tissue (EAT) reportedly correlates with coronary atherosclerosis progression. We evaluated the relationship between coronary arterial 18F-NaF uptake and EAT findings using computed tomography (CT). We studied 40 patients with ≥ 1 coronary plaque detected on cardiac CT who underwent 18F-NaF PET/CT. EAT volume was measured using CT and indexed to body surface area in each patient. Each plaque was evaluated for CT-based luminal stenosis and high-risk features. The mean EAT density surrounding each plaque was calculated as perilesional EAT density (PLED) using non-contrast CT images. Focal 18F-NaF uptake in each plaque was quantified using the maximum tissue-to-background ratio (TBRmax). EAT volume index was similar between patients with TBRmax ≥ 1.28 (previously reported optimal cutoff to predict coronary events) and those with lower TBRmax, but patients with TBRmax ≥ 1.28 showed higher maximum PLED per patient (- 86 ± 12 Hounsfield units (HU) versus - 98 ± 11 HU, P =0.0044). In the lesion-based analysis (n = 92), PLED was positively correlated with TBRmax, and the optimal PLED cutoff to identify TBRmax ≥ 1.28 was - 97 HU. On multivariate analysis adjusted for lesion location, obstructive stenosis, and high-risk plaque on CT, PLED ≥ - 97 HU remained a significant predictor of TBRmax ≥ 1.28. Increased PLED was associated with significant coronary arterial 18F-NaF uptake. Step-by-step analyses of EAT density on CT and coronary arterial 18F-NaF uptake on PET may offer novel strategies for risk prediction in coronary artery disease.

Radiofrequency and microwave ablation in a porcine liver model: non-contrast CT and ultrasound radiologic-pathologic correlation

Purpose The goal of this study was to compare intra-procedural radiofrequency (RF) and microwave ablation appearance on non-contrast CT (NCCT) and ultrasound to the zone of pathologic necrosis. Materials and methods Twenty-one 5-min ablations were performed in vivo in swine liver with (1) microwave at 140 W, (2) microwave at 70 W, or (3) RF at 200 W (n = 7 each). CT and US images were obtained simultaneously at 1, 3, and 5 min during ablation and 2, 5, and 10 min post-ablation. Each ablation was sectioned in the plane of the ultrasound image and underwent vital staining to delineate cellular necrosis. CT was reformatted to the same plane as the ultrasound transducer and transverse diameters of gas and hypoechoic/hypoattenuating zones at each time point were measured. CT, ultrasound and gross pathologic diameter measurements were compared using Student’s t-tests and linear regression. Results Visible gas and the hypoechoic zone on US images were more predictive of the pathologic ablation zone than on NCCT images (p < 0.05). The zone of necrosis was larger than the zone of visible gas on US (mean 3.2 mm for microwave, 6.4 mm for RF) and NCCT (7.6 mm microwave, 13.9 mm RF) images (p < 0.05). The zone of visible gas and hypoechoic zone on US are more predictive of pathology with microwave ablations when compared with RF ablations (p < 0.05). Conclusion When evaluating images during energy delivery, US is more accurate than CT and microwave- more predictable than RF-ablation based on correlation with in-plane pathology.

Semi-automatic measurement of intracranial hemorrhage growth on non-contrast CT.

Manual segmentations of intracranial hemorrhage on non-contrast CT images are the gold-standard in measuring hematoma growth but are prone to rater variability. We demonstrate that a convex optimization-based interactive segmentation approach can accurately and reliably measure intracranial hemorrhage growth. Baseline and 16-h follow-up head non-contrast CT images of 46 subjects presenting with intracranial hemorrhage were selected randomly from the ANNEXA-4 trial imaging database. Three users semi-automatically segmented intracranial hemorrhage to measure hematoma volume for each timepoint using our proposed method. Segmentation accuracy was quantitatively evaluated compared to manual segmentations by using Dice similarity coefficient, Pearson correlation, and Bland-Altman analysis. Intra- and inter-rater reliability of the Dice similarity coefficient and intracranial hemorrhage volumes and volume change were assessed by the intraclass correlation coefficient and minimum detectable change. Among the three users, the mean Dice similarity coefficient, Pearson correlation, and mean difference ranged from 76.79% to 79.76%, 0.970 to 0.980 (p < 0.001), and -1.5 to -0.4 ml, respectively, for all intracranial hemorrhage segmentations. Inter-rater intraclass correlation coefficients between the three users for Dice similarity coefficient and intracranial hemorrhage volume were 0.846 and 0.962, respectively, and the corresponding minimum detectable change was 2.51 ml. Inter-rater intraclass correlation coefficient for intracranial hemorrhage volume change ranged from 0.915 to 0.958 for each user compared to manual measurements, resulting in an minimum detectable change range of 2.14 to 4.26 ml. We spatially and volumetrically validate a novel interactive segmentation method for delineating intracranial hemorrhage on head non-contrast CT images. Good spatial overlap, excellent volume correlation, and good repeatability suggest its usefulness for measuring intracranial hemorrhage volume and volume change on non-contrast CT images.

Predictive value of abdominal aortic calcification index for mid-term cardiovascular events in patients with acute coronary syndrome.

The utility of abdominal aortic calcification (AAC) for prediction of cardiovascular events (CVEs) in patients with acute coronary syndrome (ACS) remains to be determined. The aim of this prospective study was to determine the predictive value of the abdominal aortic calcification index (ACI), a semi-quantitative measure of AAC, for CVEs in patients with ACS. We evaluated 314 patients with ACS. All patients underwent successful percutaneous coronary intervention to the culprit coronary vessel without in-hospital adverse events. ACI was calculated on non-contrast computed tomography images. CVEs were defined as a composite of cardiovascular death, ACS recurrence, and stroke. During a median follow-up period of 19.1months, CVEs occurred in 29 patients (9.2%). Multivariable regression analysis after adjustment for age and gender showed a significantly higher baseline ACI in patients with CVEs than in those without [median (interquartile ranges), 42.1 (25.9-60.2) vs. 20.8 (8.8-38.6) %; P = 0.021]. The cutoff value of ACI for prediction of CVEs, estimated by receiver-operating characteristic analysis, was 29.2%, with sensitivity of 76% and specificity of 64% (area under the curve, 0.69). After adjustment for conventional cardiovascular risk factors, Cox analysis showed high ACI (≥29.2%) to be significantly associated with increased risk of CVEs (P = 0.011; hazard ratio, 1.82). Multivariate analysis identified high ACI as an independent predictor of CVEs (P = 0.012; hazard ratio, 1.80). Stepwise forward selection procedure also showed that high ACI was a significant independent determinant of CVEs (P = 0.004; R2, 0.089). Both net reclassification improvement (0.64; P = 0.001) and integrated discrimination improvement (0.04; P < 0.001) improved significantly after the addition of high ACI to conventional risk factors. Evaluation of ACI using CT seems to provide valuable clinical information for proper assessment of mid-term CVEs in patients with ACS after percutaneous coronary intervention.

Single Centre Experience on Decision Making for Mechanical Thrombectomy Based on Single-Phase CT Angiography by Including NCCT and Maximum Intensity Projection Images - A Comparison with Magnetic Resonance Imaging after Non-Contrast CT.

ObjectiveThe purpose of this study was to suggest that computed tomography angiography (CTA) is valuable as the only preliminary examination for mechanical thrombectomy (MT). MT after single examination of CTA including non-contrast computed tomography (NCCT) and maximum intensity projection (MIP) improves door-to-puncture time as well as results in favorable outcomes. MethodsA total of 157 patients who underwent MT at Dong Kang Medical Center from April 2015 to March 2019 were divided into two groups based on the examination performed prior to MT : CTA group who underwent CTA with NCCT and MIP, and NCCT+magnetic resonance image (MRi) group who underwent MRI including perfusion images after NCCT. In the two groups, time to CTA imaging or NCCT+MRi imaging after symptom onset, and time to arterial puncture and reperfusion were characterized as time-related outcomes. The evaluation of vascular recanalization after MT was defined as a modified thrombolysis in cerebral infarction (mTICI) scale. National Institutes of Health Stroke Scale (NIHSS) was assessed at the time of the visit to the emergency room and modified Rankin Scale (mRS) was assessed after 90 days. ResultsTypically, there were 34 patients in the CTA group and 33 patients in the NCCT+MRi group. A significantly shorter delay for door-to-puncture time was observed (mean, 86±22.1 vs. 176±47.5 minutes; p<0.01). Also, a significantly shorter door-to-imege time in the CTA group was observed (mean, 13±6.8 vs. 93±30.8 minutes; p<0.01). Moreover, a significantly shorter onset-to-puncture time was observed (mean, 195±128.0 vs. 314±157.6 minutes; p<0.01). Reperfusion result of mTICI ≥2b was 100% (34/34) in the CTA group and 94% (31/33) in the NCCT+MRi group, and mTICI 3 in 74% (25/34) in the CTA group and 73% (24/33) in the NCCT+MRi group. Favorable functional outcomes (mRS score ≤2 at 90 days) were 68% (23/34) in the CTA group and 60% (20/33) in the NCCT+MRi group. ConclusionA single-phase CTA including NCCT and MIP images was performed as a single preliminary examination, which led to a reduction in the time of the procedure and resulted in good results of prognosis. Consequently, it is concluded that this method is of sufficient value as the only preliminary examination for decision making.

Hybrid AI based stroke characterization with explainable model

Stroke is the fourth leading cause of death in the United States. Treatment pathway depends on the type of stroke and it is imperative to provide a quick diagnosis of stroke type for the better outcome. The solution that is being provided in this work is that of hybrid approach of image processing and deep learning to rule out the presence of Intra Cranial Hemorrhage (ICH) accurately by analyzing the non-contrast CT images. The current solution is developed on a data of 379 scans (~180 subjects with ICH) and validated on 100 scans of different institution. A pre-trained Inception-ResNet V2 model was used for training with 10 epochs while freezing the convolutional layers and for 50 epochs unfreezing them. The accuracy achieved on the training dataset is 97% with a sensitivity of 95% and specificity of 98%. The model was tested on a blind dataset of different institution containing 100 subjects. It scored an accuracy of 91% with sensitivity of 90% and 100% specificity for ruling out the presence of ICH; we outperform the stroke characterization accuracy from state of art published work using hybrid AI approach. Further, the algorithm highlights affected region of ICH, maps to the anatomical part and provides the quantification of the bleed area resulting into automated computation of ICH score. In order to earn the radiologist’s confidence on the developed model, we built a visualization engine to understand the learnings of the model for the corresponding prediction as shown in (See Fig 1.)

Efficacy of Thrombus Density on Noninvasive Computed Tomography Neuroimaging for Predicting Thrombus Pathology and Patient Outcome after Mechanical Thrombectomy in Acute Ischemic Stroke.

Background and Purpose:The aim of this study was to investigate the efficacy of thrombus density on noninvasive computed tomography (CT) neuroimaging for predicting thrombus pathology and patient outcome after mechanical thrombectomy in acute ischemic stroke.Materials and Methods:This retrospective chart and imaging review included patients that were treated by mechanical thrombectomy at Siriraj Hospital according to the American Heart Association/American Stroke Association guidelines for the early management of patients with acute ischemic stroke from March 2010 to February 2015 study period. Preintervention noncontrast CT (NCCT), CT angiography (CTA), and/or contrast-enhanced CT (CECT) images were interpreted using CT densitometry. Pathology results were classified as white, red, or mixed thrombi. The result of treatment was evaluated by the modified Rankin Scale at 90 days after treatment.Results:From 97 included patients – 97 NCCT images, 48 CTA images, 48 CECT images, and 54 pathologic results of cerebral thrombi were included in the final analysis. Mean clot Hounsfield unit values on NCCT, CTA, and CECT were significantly different between red and white thrombus (P = 0.001 on NCCT, P = 0.03 on CTA, and P = 0.001 on CECT), and between red and mixed thrombus (P = 0.043 on NCCT and P = 0.002 on CTA). However, no significant difference was observed between white thrombus and mixed thrombus (P = 0.09 on NCCT, P = 1.00 on CTA, and P = 0.054 on CECT). There was no significant correlation between type of cerebral thrombus or clot density and the result of treatment.Conclusion:Thrombus density on CT was found to be a significant predictor of thrombus pathology; however, no significant association was observed between thrombus type or clot density and patient outcome after mechanical thrombectomy.

Preoperative Evaluation of Liver Parenchyma of Potential Donors in Living Donor Liver Transplantation

BackgroundHepatic steatosis carries a risk of postoperative liver dysfunction in donors and graft nonfunction in recipients. This article discusses the evaluation of fatty infiltration in donor liver parenchyma on multidetector computed tomography. Materials and methodsThe methods of hepatic fat estimation include measurement of hepatic attenuation in HU and calculation of the liver attenuation index (LAI). Liver attenuation values reflect the degree of steatosis. Average attenuation of liver parenchyma is calculated by placing the circular region of interest of at least 1 cm2 area at multiple places in the liver on noncontrast CT images. Splenic attenuation is measured by placing the circular region of interest at its upper, middle, and lower poles. The LAI is the difference between mean hepatic attenuation and mean splenic attenuation. ResultsA total of 52 donors were evaluated as potential recipients (34 men, 18 women; mean age, 33.2 years; range, 23-55 years). In 34 donors liver attenuation index (LAI) values were from 2 HU to 22 HU. An LAI > 5 HU correctly predicted the absence of significant macrovesicular steatosis. These donors were acceptable for a liver transplant. The LAI values of −10 to 5 HU were suggestive of mild to moderate steatosis (6%-30%); 18 (34.6%) volunteers did not proceed to donation because of negative LAI < −5 HU. In 2 cases with LAI of −7 and LAI of −8 liver biopsy was performed, and 30% steatosis was confirmed in the pathohistologic examination. Unacceptable liver biopsy result was considered as contraindication for donation. The LAI values of < −10 HU were suggestive of moderate to severe hepatic steatosis of 30% or greater. In these cases liver biopsy is not needed, as such donors are not acceptable for liver transplant. ConclusionComputed tomography imaging provides a detailed evaluation of fatty infiltration in donor liver parenchyma.

Radiomics features on non-contrast computed tomography predict early enlargement of spontaneous intracerebral hemorrhage

ObjectiveTo explore the value of radiomics features on non-contrast computed tomography (NCCT) in predicting early enlargement of spontaneous intracerebral hemorrhage (SICH). Patients and Methods167 patients with SICH were divided into enlarged hematoma and non-enlarged hematoma groups based on the volume of hematoma on 24-h follow-up CT images > 30% and/or 6 ml of the baseline NCCT. The baseline NCCT images of all cases were imported into radiomics software to extract the radiomics features of the initial hematoma. For each case, the features with good predictability were retained after the feature-selected process; the remaining features were used to construct model with 23 algorithms one-by-one. A 5-fold method was used to cross-validate the model and repeated 5 times. The algorithm model with the highest accuracy was selected as predictive model for hematoma enlargement (HE) in SICH, its average parameters including AUC, accuracy, sensitivity, specificity, F1 score, positive predictive value (PPV), negative predictive value (NPV), false positive rate (FPR), false negative rate (FNR),and false discovery rate (FDR) were taken as evaluating indicators. ResultsA total of 1227 texture features of each cerebral hematoma were obtained. After the feature-selected process, 4 features (wavelet-LHL mean, wavelet-LLL _ Idm, wavelet-LLL _run length non-uniformity normalized, and wavelet-LLL _contrast) remained to construct the predictive models. Among 23 model algorithms, Linear Support Vector Classifier showed the highest accuracy (72.6%), and eventually was selected as the predictive model, its AUC, accuracy, sensitivity, specificity, F1 score, PPV, NPV, FPR, FNR, and FDR were 0.729, 0.726,0.717,0.736,0.714, 0.736, 0.741, 0.264, 0.283 and 0.264, respectively. ConclusionRadiomics features of cerebral hematoma on baseline NCCT images showed good performance in predicting HE of SICH.

Semi-automated infarct segmentation from follow-up noncontrast CT scans in patients with acute ischemic stroke.

Cerebral infarct volume observed in follow-up noncontrast computed tomography (NCCT) scans of acute ischemic stroke (AIS) patients is as an important radiologic outcome measure of the effectiveness of endovascular therapy (EVT). In this paper, our aim is to propose a semiautomated segmentation approach that can accurately measure ischemic infarct volume from NCCT images of AIS patients. A novel cascaded random forest (RF) learning is first employed to classify each voxel into normal or ischemic voxel, leading to an infarct probability map. Four types of features: intensity, statistical information in local region, symmetric difference compared to the contralateral side, and spatial probability of infarct occurrence generated by the STAPLE method, are extracted. These features are input into RF to train a first-stage classifier. The coarse segmentation results generated by the first-stage classifier are then used to train a fine second-stage classifier with fivefold cross validation. The RF estimated infarct probability map obtained in the second-stage testing as well as user input high-level knowledge are then incorporated into a convex optimization function to obtain final segmentation. One hundred AIS patients were collected in this study, of which 70 patient images were used for evaluation while the remaining 30 patient images were used for RF training. Quantitative results show that the proposed approach is capable of yielding a dice coefficient (DC) of 79.42%, significantly outperforming some state-of-the-art automated segmentation methods, such as the RF-based methods and convolutional neural network (CNN)-based segmentation method, U-net. The infarct volume obtained by the proposed method is strongly correlated with the manually segmented volume. In addition, interobserver variability analysis initialized by two observers suggests low user dependency. Our proposed semiautomated segmentation method can accurately segment infarct from NCCT of AIS patients.

Noncontrast computer tomography–based radiomics model for predicting intracerebral hemorrhage expansion: preliminary findings and comparison with conventional radiological model

To develop a radiomics model for predicting hematoma expansion in patients with intracerebral hemorrhage (ICH) and to compare its predictive performance with a conventional radiological feature-based model. We retrospectively analyzed 251 consecutive patients with acute ICH. Two radiologists independently assessed baseline noncontrast computed tomography (NCCT) images. For each radiologist, a radiological model was constructed from radiological variables; a radiomics score model was constructed from high-dimensional quantitative features extracted from NCCT images; and a combined model was constructed using both radiological variables and radiomics score. Development of models was constructed in a primary cohort (n = 177). We then validated the results in an independent validation cohort (n = 74). The primary outcome was hematoma expansion. We compared the three models for predicting hematoma expansion. Predictive performance was assessed with the receiver operating characteristic (ROC) curve analysis. In the primary cohort, combined model and radiomics model showed greater AUCs than radiological model for both readers (all p < .05). In the validation cohort, combined model and radiomics model showed greater AUCs, sensitivities, and accuracies than radiological model for reader 2 (all p < .05). Combined model showed greater AUC than radiomics model for reader 1 only in the primary cohort (p = .03). Performance of three models was comparable between reader 1 and reader 2 in both cohorts (all p > .05). NCCT-based radiomics model showed high predictive performance and outperformed radiological model in the prediction of early hematoma expansion in ICH patients. • Radiomics model showed better performance for prediction of hematoma expansion in patients with intracerebral hemorrhage than radiological feature-based model. • Hematomas which expanded in follow-up NCCT tended to be larger in baseline volume, more irregular in shape, more heterogeneous in composition, and coarser in texture. • A radiomics model provides a convenient and objective tool for prediction of hematoma expansion that helps to define subsets of patients who would benefit from anti-expansion therapy.

Non-contrast CT image characteristics on admission predict the 3-month outcome of cerebral venous sinus thrombosis: an observational study in a single institution

BackgroundVarious computed tomography (CT) appearances of cerebral venous sinus thrombosis (CVST) were associated with different prognosis and the patients with large intracranial hematoma will have adverse outcomes, but no in-depth study of non-contrast CT image appearances was carried out. We aimed to test the hypothesis that non-contrast CT image characteristics on admission are associated with and predict the outcome of CVST at 3 months.MethodsThree hundred and six patients with CVST between 2008 and 2017 were collected. Age, sex, onset of CVST(acute, subacute, or chronic), etiology, clinical manifestations, midline shift, occluded venous sinus, location of infarction, non-contrast CT image characteristics, and the 3-month outcome were recorded. In addition, we established a non-contrast CT image-based classification and grading system to test the hypothesis; the CVST patients were classified into four grades (namely non-contrast CT image-based classification): grade I, no obvious abnormality; grade II, simple vein infarction without hemorrhage or with subarachnoid hemorrhage; grade III, cerebral venous infarction with subarachnoid hemorrhage; and grade IV, cerebral vein infarction with hematoma. All enrolled patients had received subcutaneous injections of low molecular weight heparin subcutaneous injection for 14 days. Thereafter, oral anticoagulant therapy with warfarin was continued. Patients with epilepsy were given antiepileptic drugs, and patients with cerebral herniation received decompressive craniotomy.ResultsOur observational findings revealed that midline shift (> 5 mm), location of lesion (frontal lobe and temporal lobe), and cerebral venous infarction with subarachnoid or hematoma (grade III and IV) were associated with 3-month poor outcome (p < 0.05); the respective increased risks were 12.730 [risk ratio (RR) 12.730, 95% confidence interval (CI) 1.680–96.490, p = 0.014], 46.538 (RR 146.538, 95% CI 6.222–348.079, p = 0.000), 32.549 (RR 32.549, 95% CI 2.180–486.104, p = 0.012), 37.725 (RR 37.725, 95% CI 2.051–693.778, p = 0.015), and 93.164-fold (RR 93.164, 95% CI 11.137–779.328, p = 0.000). However, seizure, hemiplegia, location of occluded venous sinus (super sagittal sinus and deep venous systems), location of infarction (parietal lobe), and non-contrast CT image-based classification (I) were not correlated with the adverse outcome (p > 0.05).ConclusionsOur findings suggested that non-contrast CT image characteristics on admission were associated with and predict the 3-month outcome of CVST. However, the ultimate conclusions need to be confirmed by a large sample of CVST patients at multiple institutions.

Clot permeability and histopathology: is a clot’s perviousness on CT imaging correlated with its histologic composition?

BackgroundClot perviousness in large vessel occlusion has been shown to be associated with improved recanalization outcomes with mechanical thrombectomy and intravenous thrombolysis.ObjectiveTo evaluate the association between clot perviousness based on...

Improved detection of metastatic lymph nodes in oesophageal squamous cell carcinoma by combined interpretation of fluorine-18-fluorodeoxyglucose positron-emission tomography/computed tomography

BackgroundWe sought to evaluate the diagnostic performance of fluorine-18-fluorodeoxyglucose positron-emission tomography/computed tomography (18F-FDG PET/CT) in the detection of metastatic lymph nodes by combined interpretation of PET/CT images in patients with oesophageal squamous cell carcinoma.MethodsTwo hundred three patients with oesophageal squamous cell carcinoma underwent 18F-FDG PET/CT before oesophagectomy and lymph node dissection. Maximum standardized uptake value (SUVmax), mean Hounsfield unit (HU), short axis diameter (size), and visual CT attenuation (high, iso-, low) were evaluated on noncontrast CT and PET images following PET/CT scan. In this combined interpretation protocol, the high attenuated lymph nodes were considered benign, even if the SUVmax value was high. The diagnostic accuracy of each method was compared using the postoperative histologic result as a reference standard.ResultsA total of 1099 nodal stations were dissected and 949 nodal stations were proven to demonstrate metastasis. SUVmax and size of the malignant lymph nodes were higher than those of the benign nodes, and visual CT attenuation was significantly different among the two groups (P < 0.001). Using cutoff values of 2.6 for SUVmax and 10.2 mm for size, the combined interpretation of an SUVmax of more than 2.6 with iso- or low CT attenuation [area under the curve (AUC): 0.846, 95% confidence interval (CI): 0.824–0.867] showed significantly better diagnostic performance for detecting malignant lymph nodes than SUVmax only (AUC: 0.791, 95% CI: 0.766–0.815) and size (AUC: 0.693, 95% CI: 0.665–0.720) methods (P < 0.001) in a receiver operating characteristic curve analysis.ConclusionsThe diagnostic accuracy of PET/CT for nodal metastasis in oesophageal squamous cell carcinoma was improved by the combined interpretation of 18F-FDG uptake and visual CT attenuation pattern.