Head CT Images Research Articles

Predicting Intracerebral Hemorrhage Outcomes Using Deep Learning Models to Extract Head CT Imaging Features

Predicting Intracerebral Hemorrhage Outcomes Using Deep Learning Models to Extract Head CT Imaging Features

DSIFNet: Implicit feature network for nasal cavity and vestibule segmentation from 3D head CT

This study is dedicated to accurately segment the nasal cavity and its intricate internal anatomy from head CT images, which is critical for understanding nasal physiology, diagnosing diseases, and planning surgeries. Nasal cavity and it’s anatomical structures such as the sinuses, and vestibule exhibit significant scale differences, with complex shapes and variable microstructures. These features require the segmentation method to have strong cross-scale feature extraction capabilities. To effectively address this challenge, we propose an image segmentation network named the Deeply Supervised Implicit Feature Network (DSIFNet). This network uniquely incorporates an Implicit Feature Function Module Guided by Local and Global Positional Information (LGPI-IFF), enabling effective fusion of features across scales and enhancing the network's ability to recognize details and overall structures. Additionally, we introduce a deep supervision mechanism based on implicit feature functions in the network's decoding phase, optimizing the utilization of multi-scale feature information, thus improving segmentation precision and detail representation. Furthermore, we constructed a dataset comprising 7116 CT volumes (including 1,292,508 slices) and implemented PixPro-based self-supervised pretraining to utilize unlabeled data for enhanced feature extraction. Our tests on nasal cavity and vestibule segmentation, conducted on a dataset comprising 128 head CT volumes (including 34,006 slices), demonstrate the robustness and superior performance of proposed method, achieving leading results across multiple segmentation metrics.

Incidental Detection of Previously Unknown Strokes on Head CT Examinations: An Untapped Opportunity for Secondary Prevention

BackgroundThe incidental discovery of previously unknown strokes on neuroimaging is an opportunity to implement secondary prevention, reducing the risk of recurrent strokes by up to 80%. ObjectiveTo evaluate the prevalence of previously unknown strokes on emergency department (ED) head CT imaging and identify associated patient and imaging factors. MethodsRetrospective study of adult patients receiving head CT (age ≥ 18 years) at 3 EDs between July and December 2023. Old strokes on CT imaging were categorized as previously known or unknown. Patient and imaging factors associated with unknown strokes were assessed via univariable regression, multivariable regression, and decision tree analysis. ResultsIn 21,985 ED encounters with head CT, 869 (4.0%) examinations demonstrated an old stroke (mean age, 74.4 years ± 14.0 [SD]), of which 372 (43%; 1.7% of all CTs) were unknown. Univariable analysis showed that unknown strokes were associated with greater age (OR, 1.03; 95% CI: 1.02-1.04; P < 0.001), a single site of old stroke (OR, 2.7; 95% CI: 2.06-3.58; P < 0.001), smaller strokes (OR, 1.8, 95% CI: 1.6-2.0, P <0.001) as well as gangliocapsular (OR, 2.8; 95% CI: 1.9-4.0; P < 0.001) and cerebellar location (OR, 2.1; 95% CI: 1.4-3.2; P < 0.001). Results of further analyses corroborated those of the univariable regression. ConclusionMore than a third of old strokes on head CT imaging are unknown to patients and clinicians. Capturing this opportunity for secondary prevention could benefit 100,000-200,000 patients per year in the US, based on trends in ED care.

Intra-arterial contrast administration: a confounding appearance on stroke CT imaging.

A female patient in her 70s presented with mild slurred speech. Clinical assessment revealed only minor dysarthria. A stroke protocol CT study was performed comprising a non-contrast CT head, CT angiogram (CTA) and CT perfusion (CTP) imaging. Despite a normal non-contrast CT head, CTP imaging demonstrated a large left anterior circulation territory abnormality compatible with core infarct. The CTA displayed absent left middle cerebral artery opacification and an abnormal pattern of thalamic and posterior fossa structure enhancement. Following a neuroradiology review, it was identified that the contrast bolus had been administered into the right brachial artery causing direct arterial opacification of the right vertebral/carotid arteries and their distal branches. No contrast was present in the left carotids or distal left intracranial anterior circulation giving the false impression of a large vessel occlusion. The patient was subsequently discharged without receiving inappropriate reperfusion therapy. This case highlights the role of technical factors influencing CT interpretation, particularly when performed by artificial intelligence/automated analysis.

Dual Energy CT for Deep Learning-Based Segmentation and Volumetric Estimation of Early Ischemic Infarcts.

Ischemic changes are not visible on non-contrast head CT until several hours after infarction, though deep convolutional neural networks have shown promise in the detection of subtle imaging findings. This study aims to assess if dual-energy CT (DECT) acquisition can improve early infarct visibility for machine learning. The retrospective dataset consisted of 330 DECTs acquired up to 48h prior to confirmation of a DWI positive infarct on MRI between 2016 and 2022. Infarct segmentation maps were generated from the MRI and co-registered to the CT to serve as ground truth for segmentation. A self-configuring 3D nnU-Net was trained for segmentation on (1) standard 120kV mixed-images (2) 190keV virtual monochromatic images and (3) 120kV + 190keV images as dual channel inputs. Algorithm performance was assessed with Dice scores with paired t-tests on a test set. Global aggregate Dice scores were 0.616, 0.645, and 0.665 for standard 120kV images, 190keV, and combined channel inputs respectively. Differences in overall Dice scores were statistically significant with highest performance for combined channel inputs (p < 0.01). Small but statistically significant differences were observed for infarcts between 6 and 12h from last-known-well with higher performance for larger infarcts. Volumetric accuracy trended higher with combined inputs but differences were not statistically significant (p = 0.07). Supplementation of standard head CT images with dual-energy data provides earlier and more accurate segmentation of infarcts for machine learning particularly between 6 and 12h after last-known-well.

Deep Learning-Based Fine-Tuning Approach of Coarse Registration for Ear-Nose-Throat (ENT) Surgical Navigation Systems.

Accurate registration between medical images and patient anatomy is crucial for surgical navigation systems in minimally invasive surgeries. This study introduces a novel deep learning-based refinement step to enhance the accuracy of surface registration without disrupting established workflows. The proposed method integrates a machine learning model between conventional coarse registration and ICP fine registration. A deep-learning model was trained using simulated anatomical landmarks with introduced localization errors. The model architecture features global feature-based learning, an iterative prediction structure, and independent processing of rotational and translational components. Validation with silicon-masked head phantoms and CT imaging compared the proposed method to both conventional registration and a recent deep-learning approach. The results demonstrated significant improvements in target registration error (TRE) across different facial regions and depths. The average TRE for the proposed method (1.58 ± 0.52 mm) was significantly lower than that of the conventional (2.37 ± 1.14 mm) and previous deep-learning (2.29 ± 0.95 mm) approaches (p < 0.01). The method showed a consistent performance across various facial regions and enhanced registration accuracy for deeper areas. This advancement could significantly enhance precision and safety in minimally invasive surgical procedures.

CSF Susceptibility Variation in Patient With Intracranial Hemorrhage: Implications for Quantitative Susceptibility Mapping Reference Selection.

Quantitative susceptibility mapping (QSM) is an emerging MRI technique with multiple clinical applications. As tissue susceptibility cannot be directly measured using MRI, QSM imaging techniques must indirectly compute susceptibility values, requiring regularization methods. CSF is a popular choice for regularization due to its near water susceptibility in healthy controls. However, the impact of pus, elevated protein, or blood dissolved in CSF on QSM regularization is not well defined. This study aimed to investigate the effects of intracranial hemorrhage (ICH) on selecting CSF as reference for QSM imaging. A total of 87 subjects, 53 with ICH (5 intraventricular, 19 subarachnoid, 27 both, and 2 intraparenchymal only) and 37 without hemorrhage (27 with MS, 10 without MS), were included in this study. Imaging was performed using 3D multiecho gradient echo, FLAIR, and multiecho complex total field inversion (mcTFI) at 3 T. McTFI with and without CSF zero-referencing regularization was generated from the 3DMEGRE data and reviewed with FLAIR images. Regions of hemorrhagic (H+) and nonhemorrhagic (H-) CSF were manually selected in reference to head CT and FLAIR images by a PGY III diagnostic radiology resident and Certificate of Added Qualification-certified neuroradiologist with 10 years' experience. Paired Student t test and one-way ANOVA were used with post hoc multicomparisons. A P value <0.05 was considered statistically significant. Areas of H- CSF were noted to have higher regularized QSM values in subjects with ICH relative to subjects without. Unregularized H- QSM values were also noted to have a systematically higher value in ICH subjects relative to subjects without blood. Subjects with MS and without ICH did not show significant difference in H- CSF regularized or unregularized QSM values. QSM values of areas suggested to not have hemorrhage on other imaging showed significantly higher QSM values in ICH subjects relative to subjects without ICH. Additionally, areas of hemorrhage did not show significant QSM value difference between regularized and unregularized QSM images. These findings suggest that, in subjects with any area of ICH, QSM values for no-hemorrhagic areas may be significantly altered using CSF regularization relative to subjects without ICH, with implications for intra- and intersubject QSM value analysis.

Synthetic CT generation based on multi-sequence MR using CycleGAN for head and neck MRI-only planning.

The purpose of this study is to investigate the influence of different magnetic resonance (MR) sequences on the accuracy of generating computed tomography (sCT) images for nasopharyngeal carcinoma based on CycleGAN. In this study, 143 patients' head and neck MR sequence (T1, T2, T1C, and T1DIXONC) and CT imaging data were acquired. The generator and discriminator of CycleGAN are improved to achieve the purpose of balance confrontation, and a cyclic consistent structure control domain is proposed in terms of loss function. Four different single-sequence MR images and one multi-sequence MR image were used to evaluate the accuracy of sCT. During the model testing phase, five testing scenarios were employed to further assess the mean absolute error, peak signal-to-noise ratio, structural similarity index, and root mean square error between the actual CT images and the sCT images generated by different models. T1 sequence-based sCT achieved better results in single-sequence MR-based sCT. Multi-sequence MR-based sCT achieved better results with T1 sequence-based sCT in terms of evaluation metrics. For metrological evaluation, the global gamma passage rate of sCT based on sequence MR was greater than 95% at 3%/3mm, except for sCT based on T2 sequence MR. We developed a CycleGAN method to synthesize CT using different MR sequences, this method shows encouraging potential for dosimetric evaluation.

Investigating the Effects of Tube Current and Tube Voltage on Patient Dose in Computed Tomography Examinations with Principial Component Analysis and Cluster Analysis: Phantom Study.

The aim of this study was to investigate the effects of tube current and tube voltage choices on patient dose in adult and pediatric CT protocols by qualitative analysis using Principal Component Analysis (PCA), cluster analysis, and statistical analysis.Dose length product (DLP), Effective mAs (Eff. mAs), and volume-weighted CT dose index (CTDIvol) dose descriptors were obtained from 16 adult and pediatric head phantom CT examinations. Different tube voltage and tube current values were selected in both pediatric head and adult head CT imaging protocols, and PCA and cluster analysis were applied to the data obtained for qualitative analysis of the relationship between CTDIvol, Eff. mAs and Total DLP values. The two principial components (PC) with the highest values among those obtained as a result of the PCA method were used. PC1 was 70.97%, and PC2 was 28.03%. In the cluster analysis, it was observed that the values obtained from pediatric and adult phantom CT scans were classified into two different clusters. The correlation coefficient for adult patients was r = 0.998, and for pediatric patients, the correlation coefficient was r = 0.947. When the obtained clusters were examined, the degree of closeness or distance of the variables could be observed. In the study, as a result of the analysis of CTDIvol, Eff. mAs and Total DLP data based on manufacturer data at different kV and mA values with PCA and cluster analysis, it was shown that pediatric patients could be exposed to more radiation than the adult patients.

Listeria Rhombencephalitis: A Case of an Immunocompetent Adult

Listeriosis is one of the most important foodborne diseases, caused by a Gram-positive facultative anaerobic intracellular bacterium found in food and the environment, Listeria monocytogenes. It causes one of the most life-threatening bacterial infections of the central nervous system (CNS). Meningoencephalitis and rhomboencephalitis are the most common forms and immunosuppression represents the main risk factor. In this article, we report a case of severe, invasive Listeria monocytogenes rhombencephalitis in a healthy middle-aged man, admitted to our hospital with meningeal syndrome associated with left VI nerve palsy. No apparent abnormality was found in a head CT and brain magnetic resonance imaging. Following the preliminary LP results, a diagnosis of bacterial meningitis was suspected and empirical antibiotic and antiviral therapy was initiated. Because of the lack of any significant improvement in his clinical status, on day 2 of his admission, a genetic identification of CSF pathogens by high-throughput genome sequencing was performed and identified L. monocytogenes. The Clinical status improved with complete recovery after appropriate antibiotic treatment with sulfamethoxazole-trimethoprim. In conclusion, in an infectious context, pontine involvement should systematically evoke the diagnosis of listeriosis.

The Inter-Observer Agreement Between Resident and Consultant Radiologists in Reporting Emergency Head CT Scans

Objectives: To determine the inter-observer agreement between resident and consultant radiologists in reporting emergency head CT scans. Materials and Methods: This descriptive cross-sectional retrospective study was performed in a tertiary care hospital Karachi from 1st October 2021 to 31st march 2022. Total of 111 number of patients of 18-70 years of age of either gender who underwent emergency Head CT scans were included. Patients who came for follow-up CT scans were excluded. CT findings were interpreted by the radiology resident on duty. Subsequently, these CT images were interpreted by a consultant radiologist, and the decision of both resident and faculty member CT was correlated for inter-observer agreement. Results: The mean age was 47.47 ± 14.20 years. The majority of the patients 46 (41.44%) were between 56 to 70 years of age. Out of these 111 patients, 72 (64.86%) were male and 39 (35.14%) were female with a ratio of 1.9:1. Discrepancy between resident and consultant radiologists in reporting emergency head CT scans is seen in 6 (5.41%) patients. Inter-observer agreement between resident and consultant radiologists in reporting emergency head CT scans was found to be 94.59% with a kappa “κ” value of 0.885 which showed a very strong agreement. Conclusion: This study concluded that little discrepancy was found in inter-observer agreement between the radiology residents and the faculty members for interpretation of CT images of Head.

Adaptive mask-based brain extraction method for head CT images.

Brain extraction is an important prerequisite for the automated diagnosis of intracranial lesions and determines, to a certain extent, the accuracy of subsequent lesion identification, localization, and segmentation. To address the problem that the current traditional image segmentation methods are fast in extraction but poor in robustness, while the Full Convolutional Neural Network (FCN) is robust and accurate but relatively slow in extraction, this paper proposes an adaptive mask-based brain extraction method, namely AMBBEM, to achieve brain extraction better. The method first uses threshold segmentation, median filtering, and closed operations for segmentation, generates a mask for the first time, then combines the ResNet50 model, region growing algorithm, and image properties analysis to further segment the mask, and finally complete brain extraction by multiplying the original image and the mask. The algorithm was tested on 22 test sets containing different lesions, and the results showed MPA = 0.9963, MIoU = 0.9924, and MBF = 0.9914, which were equivalent to the extraction effect of the Deeplabv3+ model. However, the method can complete brain extraction of approximately 6.16 head CT images in 1 second, much faster than Deeplabv3+, U-net, and SegNet models. In summary, this method can achieve accurate brain extraction from head CT images more quickly, creating good conditions for subsequent brain volume measurement and feature extraction of intracranial lesions.

Motion Artifact Detection Based on Regional–Temporal Graph Attention Network from Head Computed Tomography Images

Artifacts are the main cause of degradation in CT image quality and diagnostic accuracy. Because of the complex texture of CT images, it is a challenging task to automatically detect artifacts from limited image samples. Recently, graph convolutional networks (GCNs) have achieved great success and shown promising results in medical imaging due to their powerful learning ability. However, GCNs do not take the attention mechanism into consideration. To overcome their limitations, we propose a novel Regional–Temporal Graph Attention Network for motion artifact detection from computed tomography images (RT-GAT). In this paper, head CT images are viewed as a heterogeneous graph by taking regional and temporal information into consideration, and the graph attention network is utilized to extract the features of the constructed graph. Then, the feature vector is input into the classifier to detect the motion artifacts. The experimental results demonstrate that our proposed RT-GAT method outperforms the state-of-the-art methods on a real-world CT dataset.

Abstract WP66: Increases in Emergency Department Utilization of CT Angiography Brain and CT Perfusion in the Transferred Patient Population Across the Endovascular Therapy Era

Background: With the advent of endovascular therapy as a treatment modality for LVO stroke, the importance of stroke networks has become increasingly clear. We sought to understand the way advanced neuroimaging is utilized in transferred emergency department (ED) encounters using a large national dataset. Methods: Retrospective cohort study of adult ED encounters, comparing ED transfers (to other acute care hospitals) to discharged patients. Data from the 2013-2018 National Emergency Department Sample (NEDS), an AHRQ dataset containing a weighted 20% sample of US ED encounters were analyzed. CTA Head, CTP, MRI Brain, and Non-contrast CT Head imaging use was determined from Common Procedural Terminology (CPT) codes. ICD10 Codes I60-I69 were used to indicate cerebrovascular disease. Statistical comparisons used 95% CI for rate per 100,000 ED visits. Results: ED utilization of advanced neuroimaging was higher in transferred patients than discharged patients in all years studied (Fig). Between 2013 and 2018, utilization of CTA and CTP increased 8-fold and 20-fold, respectively in the transferred population. Noncontrast head CT and MRI were relatively stable but more frequent in transferred encounters. Between 2013 and 2018, encounters associated with a transfer came disproportionately from smaller hospitals (45% from hospitals with &lt;=20,000 ED visits per year, and 24% from ED visits at hospitals with between 20,000 and 40,000 ED visits per year). In 2018 transferred encounters with a CTA Head or CTP, 75% (72-77) and 76%(70-83), respectively had a diagnosis code associated with cerebrovascular disease. Conclusions: Utilization of CTA and CTP increased substantially between 2013-2018 in the ED transfer population, and most of these encounters are associated with a cerebrovascular disease diagnosis. Detailed diagnosis and hospital demographic data to be presented. These results have implications for the design and optimization of stroke systems of care.

Abstract WP160: Mechanisms of Brain Hemorrhage in Chronic Kidney Disease

Introduction: Chronic kidney disease (CKD) is a risk factor for intracerebral hemorrhage (ICH) and for worse outcomes following ICH. In this study, we examined the relationship between ICH and CKD, as well as intracranial arterial calcification (IAC) as an intermediate biomarker. We hypothesized that CKD is an independent risk factor for IAC burden, and that high IAC burden is an independent risk factor for ICH. Methods: A retrospective cohort of spontaneous ICH and non-ICH patients with non-contrast head CT imaging was identified. For each patient, eGFR (estimated glomerular filtration rate) along with age, sex, race, ethnicity, hypertension (HTN), diabetes (DM), heart disease, and hyperlipidemia were abstracted. A convolutional neural network was developed to automate quantification of IAC burden within the anterior and posterior circulation, and all results were visually verified for accuracy by an expert neuroradiologist. The relationship between eGFR and IAC was assessed with Pearson correlation and one-way ANOVA. Univariate analysis of ICH risk factors was evaluated using t-test and Chi-square, while multivariate models were analyzed using logistic regression. Results: Among the 717 patients in this study, 98 were ICH patients. In all patients, mean IAC burden increased significantly with CKD stage (p &lt; 0.01). Univariate analysis for IAC risk factors yielded statistically significant associations for eGFR (p &lt; 0.01) as well as age (p &lt; 0.01), HTN (p = 0.01), and African-American race (p = 0.02). In multivariate analysis, eGFR remained a significant independent predictor of IAC (p &lt; 0.01) along with HTN (p &lt; 0.01), DM (p &lt; 0.01), and race (p &lt; 0.01). Furthermore, IAC burden was significantly higher in ICH patients (15.4 mL vs. 8.7 mL, p = 0.02). In multivariate analysis, IAC remained a significant independent predictor of ICH (p &lt; 0.01) along with HTN (p &lt; 0.01), DM (p = 0.03), sex (p &lt; 0.01), and race (p &lt; 0.01). Conclusion: This is the first demonstration that eGFR predicts strictly quantified IAC, which then predicts ICH. These findings offer novel insights into determinants of IAC and mechanisms underlying the CKD-ICH relationship.

Essential parameters needed for a U-Net-based segmentation of individual bones on planning CT images in the head and neck region using limited datasets for radiotherapy application.

Objective.The field of radiotherapy is highly marked by the lack of datasets even with the availability of public datasets. Our study uses a very limited dataset to provide insights on essential parameters needed to automatically and accurately segment individual bones on planning CT images of head and neck cancer patients.Approach.The study was conducted using 30 planning CT images of real patients acquired from 5 different cohorts. 15 cases from 4 cohorts were randomly selected as training and validation datasets while the remaining were used as test datasets. Four experimental sets were formulated to explore parameters such as background patch reduction, class-dependent augmentation and incorporation of a weight map on the loss function.Main results.Our best experimental scenario resulted in a mean Dice score of 0.93 ± 0.06 for other bones (skull, mandible, scapulae, clavicles, humeri and hyoid), 0.93 ± 0.02 for ribs and 0.88 ± 0.03 for vertebrae on 7 test cases from the same cohorts as the training datasets. We compared our proposed solution approach to a retrained nnU-Net and obtained comparable results for vertebral bones while outperforming in the correct identification of the left and right instances of ribs, scapulae, humeri and clavicles. Furthermore, we evaluated the generalization capability of our proposed model on a new cohort and the mean Dice score yielded 0.96 ± 0.10 for other bones, 0.95 ± 0.07 for ribs and 0.81 ± 0.19 for vertebrae on 8 test cases.Significance.With these insights, we are challenging the utilization of an automatic and accurate bone segmentation tool into the clinical routine of radiotherapy despite the limited training datasets.

Multiple Kernel Synthesis of Head CT Using a Task-Based Loss Function

In CT imaging of the head, multiple image series are routinely reconstructed with different kernels and slice thicknesses. Reviewing the redundant information is an inefficient process for radiologists. We address this issue with a convolutional neural network (CNN)-based technique, synthesiZed Improved Resolution and Concurrent nOise reductioN (ZIRCON), that creates a single, thin, low-noise series that combines the favorable features from smooth and sharp head kernels. ZIRCON uses a CNN model with an autoencoder U-Net architecture that accepts two input channels (smooth- and sharp-kernel CT images) and combines their salient features to produce a single CT image. Image quality requirements are built into a task-based loss function with a smooth and sharp loss terms specific to anatomical regions. The model is trained using supervised learning with paired routine-dose clinical non-contrast head CT images as training targets and simulated low-dose (25%) images as training inputs. One hundred unique de-identified clinical exams were used for training, ten for validation, and ten for testing. Visual comparisons and contrast measurements of ZIRCON revealed that thinner slices and the smooth-kernel loss function improved gray-white matter contrast. Combined with lower noise, this increased visibility of small soft-tissue features that would be otherwise impaired by partial volume averaging or noise. Line profile analysis showed that ZIRCON images largely retained sharpness compared to the sharp-kernel input images. ZIRCON combined desirable image quality properties of both smooth and sharp input kernels into a single, thin, low-noise series suitable for both brain and skull imaging.

Detection and Segmentation of Subdural Hemorrhage on Head CT Images

Detection and Segmentation of Subdural Hemorrhage on Head CT Images

Two-stage approach to intracranial hemorrhage segmentation from head CT images

Two-stage approach to intracranial hemorrhage segmentation from head CT images

The use of low dose CT scouts for MR safety screening: A multi-reader evaluation

IntroductionPlain film radiographs are recommended to assist in MRI safety screening of patients with unknown medical histories, especially in an emergency setting where patients might be unable to answer a safety questionnaire. This study assesses the performance of CT scout images, which have low radiation dose and are faster and easier to acquire compared to plain film radiographs, in finding and naming a range of head and body implants. MethodsA retrospective analysis of 40 CT Head and Neck (HN) scout images and 40 CT Chest, Abdomen and Pelvis (CAP) scout images was undertaken. A subset of these were chosen to include a range of common internal implants not identifiable externally to the patient. The images were assessed by three readers with varying levels of clinical experience in MRI who were asked to find and name any implants seen. ResultsCollectively, all readers reached a sensitivity of 85 % in finding internal implants, regardless of their clinical experience or experience in reviewing CT images, and a minimum specificity of 95 %. Implants were correctly named in 74 % of the images presented. ConclusionCT scout images were able to reveal most of the implants included. However, clinical experience in reviewing the images enhances a reader's ability to identify the type of implant. Implications for practiceIn an emergency setting, imaging can be critical in the management of patients presenting with acute illnesses. In the unconscious or unresponsive patient, the use of CT scouts, where this is the only option available, could provide valuable MRI safety information prior to a scan, improving access to the MRI scan in a timely manner.