Analysis Of CT Images Research Articles

Deep learning-based computed tomography applied to the diagnosis of rib fractures

ObjectiveTo explore the feasibility of using deep learning method to improve the efficiency of rib fracture defect diagnosis in CT images. MethodsThis study retrospective analysis of chest CT images of 2622 patients who were admitted to the outpatient and emergency departments due to chest trauma. The CT image is fed into HourglassNet for primary feature extraction, then into Inception for multi-scale feature extraction, and finally the different scale features are recombined, and then the deep convolutional neural networks (DCNN) model is imported. The model is trained by dividing fracture defects into 5 common categories, and after entering the pre-processed images, the DCNN network structure outputs the defect locations. ResultsA total of 997 rib fractures were found in 350 test set chest CT images, with 24 false-positive cases and 64 false-negative cases in the DCNN model. The accuracy of the diagnosis of rib fractures by low-senior physicians (93.2%) was lower than that of the DCNN model (95.6%) With the assistance of the DCNN model, the accuracy of the diagnosis of low-senior physicians increased (94.9%), and there was no significant difference (94.9%) between the DCNN model and the accuracy of the diagnosis of the low-senior physicians assisted by the DCNN model (95.5%). The recall rate (83.8%) for low-senior physicians to diagnose rib fractures was lower than that in the DCNN model (91.1%), and the recall rate for physician diagnosis was significantly higher (93.8%) with the assistance of the DCNN model. The average diagnostic time for low-senior physicians was (156.0 ± 31.6)s, while the diagnosis of rib fractures in the DCNN model was only (4.9 ± 1.5) s, and the diagnostic time for physicians with the assistance of the DCNN model could be shortened to (41.3 ± 7.2) s. ConclusionAfter the CT image is extracted by HourglassNet and Inception features, it is fed into the DCNN model. The DCNN model can accurately locate and diagnose rib fractures on chest CT images, significantly shortening the diagnostic time and reducing the rate of missed diagnosis and misdiagnosis. Deep learning makes it feasible to improve the efficiency of diagnosing rib fracture defects in chest CT images.

Radiomics-based analysis of CT imaging for the preoperative prediction of invasiveness in pure ground-glass nodule lung adenocarcinomas

ObjectiveThe purpose of the study is to investigate the performance of radiomics-based analysis in prediction of pure ground-glass nodule (pGGN) lung adenocarcinomas invasiveness using thin-section computed tomography images.MethodsA total of 382 patients surgically resected single pGGN and pathologically confirmed were enrolled in the retrospective study. The pGGN cases were divided into two groups: the noninvasive group and the invasive adenocarcinoma (IAC) group. 330 patients were randomly assigned to the training and testing cohorts with a ratio of 7:3 (245 noninvasive lesions, 85 IAC lesions), while 52 patients (30 noninvasive lesions, 22 IAC lesions) were assigned to the external validation cohort. A model, radiomics model, and combined clinical-radiographic-radiomic model were built using the LASSO and multivariate backward stepwise regression analysis on the basis of the selected and radiomics features. The area under the curve (AUC) and decision curve analysis (DCA) were used to evaluate and compare the model performance for invasiveness discrimination among the three cohorts.ResultsThree clinical-radiographic features (including age, gender and the mean CT value) and three radiomics features were selected for model building. The combined model and radiomics model performed better than the clinical-radiographic model. The AUCs of the combined model in the training, testing, and validation cohorts were 0.856, 0.859, and 0.765, respectively. The DCA demonstrated the radiomics signatures incorporating clinical-radiographic feature was clinically useful in predicting pGGN invasiveness.ConclusionsThe proposed radiomics-based analysis incorporating the clinical-radiographic feature could accurately predict pGGN invasiveness, providing a noninvasive biomarker for the individualized and precise medical treatment of patients.

Objective analysis of facial bone fracture CT images using curvature measurement in a surface mesh model

The human facial skeleton consists of multiple segments and causes difficulty during analytic processes. We developed image analysis software to quantify the amount of injury and validate the smooth curvature of the surface after facial bone reduction surgery. Three-dimensional computed tomography images of facial bone were obtained from 40 patients who had undergone open reduction surgery to treat unilateral zygomaticomaxillary fractures. Analytic software was developed based on the discrete curvature of a triangular mesh model. The discrete curvature values were compared before and after surgery using two regions of interest. For the inferior orbital rim, the weighted average of curvature changed from 0.543 ± 0.034 to 0.458 ± 0.042. For the anterior maxilla, the weighted average of curvature changed from 0.596 ± 0.02 to 0.481 ± 0.031, showing a significant decrement (P < 0.05). The curvature was further compared with the unaffected side using the Bray–Curtis similarity index (BCSI). The BCSI of the inferior orbital rim changed from 0.802 ± 0.041 to 0.904 ± 0.015, and that for the anterior maxilla changed from 0.797 ± 0.029 to 0.84 ± 0.025, demonstrating increased similarity (P < 0.05). In computational biology, adequate analytic software is crucial. The newly developed software demonstrated significant differentiation between pre- and postoperative curvature values. Modification of formulas and software will lead to further advancements.

Abstract WMP80: Multimodal CT Imaging Of Ischemic Stroke Thrombi Identifies Scale-Invariant Radiomic Features

Background: Paired analysis of ischemic stroke clot histology and pre-treatment CT imaging could identify digital indicators of treatment outcome or stroke etiology. Yet, differences in scale limit our ability to interpret the biology of the clot from pre-treatment images. Hypothesis: Radiomic analysis of resected clots on microCT has the potential to identify important, biologically-interpretable features that are correlated among pre-treatment imaging modalities (CTA, nCCT) and post-treatment histopathology. Methods: We performed multimodal CT imaging of 10 stroke clots retrieved by mechanical thrombectomy (example in top of Figure). Clots were manually segmented from co-registered, pre-treatment CTA and nCCT images. For the same cases, portions of resected clots were stained with iodine, imaged on a ScanCo microCT 100 (4.9 μm resolution), and segmented using adaptive intensity thresholding. Radiomic texture features ( n =93 for each modality, 279 total) were extracted from the images using PyRadiomics. Pearson correlation analysis was used to test associations between CTA (or nCCT) and microCT radiomics. Statistical significance was considered at R ≥0.7 and q &lt;0.05, after multiple testing correction. Results: 18 features were significantly correlated between CTA and microCT, and 5 features were significantly correlated between nCCT and microCT (bottom of Figure). Cluster Prominence ( R =0.96, q &lt;0.001) and Cluster Shade ( R =0.92, q &lt;0.001) on CTA and microCT exhibited near-perfect positive correlation. When comparing microCT and nCCT, FOS Energy ( R =0.90, q &lt;0.001) and Total Energy ( R =0.90, q &lt;0.001) had high positive correlation. Our findings suggest that RBC density (reflected in brighter pixels) and heterogeneity of clot constituents (reflected in pixel intensity) are scale-invariant clot attributes detectable on CT. Conclusion: Scale-invariant radiomic features can help bridge the gap between pre-treatment imaging and clot pathobiology.

Joint analysis of radiological reports and CT images for automatic validation of pathological brain conditions

We consider a problem of validation of radiological medical reports and computed tomography images for an automated analysis of brain structures. Two methods for solving the problem are proposed: a method based on the ruCLIP multimodal model, and a method based on the joint use of two separate classifiers – for a text report and for a brain CT image. We discuss methods evaluation and the obtained results. The proposed approaches make it possible to correctly classify 99.6% of radiological reports from a test sampling into 15 possible diagnoses.

Calculation and Characterization of Air Void in Mortar of the Hot Mix Asphalt (HMA) Based on CT Scanning and Image Analysis Methods

The air void content is one of the most important volumetric properties of asphalt matrixes, such as asphalt mixtures or mortars, because it can greatly affect the performance of the matrix. At the mixture level, there are standardized methods for measuring the air void content, which is an important design parameter of a mixture. However, at the mortar scale, no unified method has been proposed to determine the air void content corresponding to the asphalt mixture. Therefore, this research aims to, first, characterize the air void distribution within the mortar of an asphalt mixture and then develop an updated theoretical method for calculating the air void content of asphalt mortar. The internal structures of air voids from three commonly used asphalt mixtures with different gradations were captured using an industrial CT scanning technique and then morphologically characterized using the image analysis method. Three-dimensional models of the air voids were reconstructed, and the air void content calculated from the CT images was also verified by density tests. The scanning results show that the air void content and air void size fit the cumulative Weibull curve. The results show that the nominal maximum particle size (NMPS) of mortar and the mixture air void content were the main parameters that affected the mortar’s air voids. The mortar air void content had linear relationships with both the mixture air void content and the mixture asphalt aggregate ratio. The mortar air void content was more sensitive to the NMPS of the asphalt mixture compared with the air void content or asphalt aggregate ratio of the mixture. This research provides a quantitative method to calculate the air void content of asphalt mortar within a mixture, which may help in the matrix design of mortar.

Development of the new Aquicnn algorithm for an augmentation of CT scans images for COVID-19 Patients

The world is restoring life balance after the global Covid-19 pandemic. This situation, giving birth to new problems, arose as an outcome of pre-and post pandemic scenarios. Healthcare system was under tremendous burden during this pandemic. Government bodies, scientists, drug discovery and drug registration were working in cooperation to fight the situation and to save lives. Out of all such activities, one healthcare domain is a key player, and that is the radiological department of hospitals. As discovery made those Covid-19 effects on lungs, the pressure on CT scan activities rose. To generate a CT scan quickly and to diagnose lung condition is the need of hour. Furthermore, that became a challenge for early detection of lung conditions. Hence, this paper presents the proposed research to develop iterative techniques using deep learning computation. Paper presents the proposed lung image acquisition and augmentation algorithm developed using a convolution neural network named “AquiCNN”. This proposed algorithm will be useful for quick and enhanced lung CT image analysis.

Pneumonic-type invasive mucinous adenocarcinoma and infectious pneumonia: clinical and CT imaging analysis from multiple centers

BackgroundPneumonic-type invasive mucinous adenocarcinoma (IMA) was often misdiagnosed as pneumonia in clinic. However, the treatment of these two diseases is different.MethodsA total of 341 patients with pneumonic-type IMA (n = 134) and infectious pneumonia (n = 207) were retrospectively enrolled from January 2017 to January 2022 at six centers. Detailed clinical and CT imaging characteristics of two groups were analyzed and the characteristics between the two groups were compared by χ2 test and Student’s t test. The multivariate logistic regression analysis was performed to identify independent predictors. Receiver operating characteristic curve analysis was used to determine the diagnostic performance of different variables.ResultsA significant difference was found in age, fever, no symptoms, elevation of white blood cell count and C-reactive protein level, family history of cancer, air bronchogram, interlobular fissure bulging, satellite lesions, and CT attenuation value (all p < 0.05). Age (odds ratio [OR], 1.034; 95% confidence interval [CI] 1.008–1.061, p = 0.010), elevation of C-reactive protein level (OR, 0.439; 95% CI 0.217–0.890, p = 0.022), fever (OR, 0.104; 95% CI 0.048–0.229, p < 0.001), family history of cancer (OR, 5.123; 95% CI 1.981–13.245, p = 0.001), air space (OR, 6.587; 95% CI 3.319–13.073, p < 0.001), and CT attenuation value (OR, 0.840; 95% CI 0.796–0.886, p < 0.001) were the independent predictors of pneumonic-type IMA, with an area under the curve of 0.893 (95% CI 0.856–0.924, p < 0.001).ConclusionDetailed evaluation of clinical and CT imaging characteristics is useful for differentiating pneumonic-type IMA and infectious pneumonia.

Novel computed tomographic analysis demonstrates differences in patterns of bone mineral content between fracture configurations in distal condylar fractures of the third metacarpal/metatarsal bones in 97 Thoroughbred racehorses.

To evaluate bone mineral content patterns between fracture configurations using novel CT image analysis. CT images from 97 Thoroughbred racehorses with third metacarpal/tarsal condyle fractures provide the case population for analysis. Fractures were grouped by radiographic appearance. Image analysis objectively measured area of highly attenuating pixels (aHAP), areal density of highly attenuating pixels (dHAP) utilizing novel convex hull analysis, and subjective assessment of apparent attenuation intensity ranking (AAIR) for each fracture. Differences between fracture configuration groups were evaluated. Analysis of dHAP identified lower-density regions of highly attenuating pixels in propagating fractures and higher-density regions of highly attenuating pixels in unicortical fractures (P = .028). Complete and incomplete configurations were almost indistinguishable in dHAP (P = 1.000). The ratio of dHAP between fractured and nonfractured condyles revealed higher density gradients between condyles in unicortical (P = .040) and incomplete (P = .031) fractures than propagating fractures. Differences in patterns of bone mineral content were identified between propagating, bicortical (incomplete and complete), and unicortical fractures of third metacarpal/tarsal bone condyles. Computer-assisted geometric measurement of dHAP identified on CT images could help to assess fracture risk in equine athletes. This application may have greater relevance as standing CT screening becomes more available.

Differentiation between Pancreatic Ductal Adenocarcinoma and Normal Pancreatic Tissue for Treatment Response Assessment using Multi-Scale Texture Analysis of CT Images.

Pancreatic ductal adenocarcinoma (PDAC) is the most prevalent type of pancreas cancer with a high mortality rate and its staging is highly dependent on the extent of involvement between the tumor and surrounding vessels, facilitating treatment response assessment in PDAC. This study aims at detecting and visualizing the tumor region and the surrounding vessels in PDAC CT scan since, despite the tumors in other abdominal organs, clear detection of PDAC is highly difficult. This retrospective study consists of three stages: 1) a patch-based algorithm for differentiation between tumor region and healthy tissue using multi-scale texture analysis along with L1-SVM (Support Vector Machine) classifier, 2) a voting-based approach, developed on a standard logistic function, to mitigate false detections, and 3) 3D visualization of the tumor and the surrounding vessels using ITK-SNAP software. The results demonstrate that multi-scale texture analysis strikes a balance between recall and precision in tumor and healthy tissue differentiation with an overall accuracy of 0.78±0.12 and a sensitivity of 0.90±0.09 in PDAC. Multi-scale texture analysis using statistical and wavelet-based features along with L1-SVM can be employed to differentiate between healthy and pancreatic tissues. Besides, 3D visualization of the tumor region and surrounding vessels can facilitate the assessment of treatment response in PDAC. However, the 3D visualization software must be further developed for integrating with clinical applications.

Value of CT spectral imaging in the differential diagnosis of sarcoidosis and Hodgkin's lymphoma based on mediastinal enlarged lymph node: A STARD compliant article.

To investigate the imaging characteristics of sarcoidosis and Hodgkin's lymphoma based on mediastinal enlarged lymph node using spectral CT and evaluate whether the quantitative information can improve the differential diagnosis of these diseases. This retrospective study was approved by the institutional review board, and written informed consent was obtained from all patients. Overall, 21 patients with sarcoidosis and 39 patients with Hodgkin's lymphoma were examined with CT spectral imaging during the arterial phase (AP) and venous phase (VP). The CT values on 40 to 140 keV monochromatic images and iodine (water) concentrations of enlarged lymph nodes were obtained in AP and VP. Iodine concentrations (ICs) were normalized to the iodine concentration in the aorta. The differences in normalized iodine concentrations (NICs) and hounsfield units (HU) curve slop (λHU) were calculated. Anatomical distribution of mediastinal lymph nodes and morphologic features were also compared. Receiver operating characteristic curves were generated to help establish threshold values for the parameters required for the significant differentiation of sarcoidosis from lymphomas. The CT values on 40 to 100 keV monochromatic images in AP and 40 to 50 keV in VP were higher in sarcoidosis than those in Hodgkin's lymphoma, the differences were statistically significant (P < .05); NICs during the AP and λHU during the AP (VP) in patients with sarcoidosis differed significantly from those in patients with Hodgkin's lymphoma. Receiver operating characteristic curves analysis showed that the monochromatic CT value on 40 keV in AP had the highest sensitivity (71.4%) and specificity (100%) in differentiating sarcoidosis from Hodgkin's lymphoma. The anatomic distribution, coalescence, calcification, compression, enhancement pattern and enhancement degree of the mediastinal enlarged lymph node differed significantly between the groups (P < .05). The combination of monochromatic CT value, NICs and λHU had higher sensitivity and specificity than did those of conventional qualitative CT image analysis during the combined phases. CT spectral imaging has promising potential for the diagnostic differentiation of Hodgkin's lymphomas and sarcoidosis. The monochromatic CT value, iodine content and λHU could be valuable parameters for differentiating Hodgkin's lymphomas and sarcoidosis based on mediastinal enlarged lymph node.

Development of a CT Image Analysis Model for Cast Iron Products Based on Artificial Intelligence Methods.

This paper presents an assessment of the possibility of using digital image classifiers for tomographic images concerning ductile iron castings. The results of this work can help the development of an efficient system suggestion allowing for decision making regarding the qualitative assessment of the casting process parameters. Special attention should be focused on the fact that automatic classification in the case of ductile iron castings is difficult to perform. The biggest problem in this aspect is the high similarity of the void image, which may be a sign of a defect, and the nodular graphite image. Depending on the parameters, the tests on different photos may look similar. Presented in this article are test scenarios of the module analyzing two-dimensional tomographic images focused on the comprehensive assessment by convolutional neural network models, which are designed to classify the provided image. For the purposes of the tests, three such models were created, different from each other in terms of architecture and the number of hyperparameters and trainable parameters. The described study is a part of the decision-making system, supporting the process of qualitative analysis of the obtained cast iron castings.

Abstract 15482: Self-Adaptive Deep Learning-Based Segmentation for Uniform and Functional Preclinical and Clinical CT Image Analysis

Currently, effective treatment options for heart failure are lacking. Methods to monitor cardiac function longitudinally can accelerate preclinical and clinical research into disease progression and novel drugs. However, manual image analysis of cardiac substructures is resource-intensive and error-prone. While automated methods exist for clinical CT images, translating these to preclinical μCT data is challenging. We aimed at employing deep-learning to automatically extract quantitative data on cardiac functional parameters, developing a uniform method for both μCT and CT images. For this, we collected 110 contrast-enhanced μCT images of wild-type and accelerated aging (Ercc1 Δ/- ) mice, as well as a public dataset of 60 segmented cardiac CT images. Substructures, i.e., the left ventricle, left myocardium and right ventricle, were manually segmented in 24 μCT images to create a training set. After template-based heart detection, two separate segmentation neural network architectures were trained using the nnU-Net framework. Automatic and manual segmentations of the μCT training set were nearly identical (mean Dice score 0.983). The estimated median Dice score (0.940) of the test set results was comparable to existing methods for μCT image segmentation. Also, the automatic volume metrics were similar to manual observations by an expert. In Ercc1 Δ/- mice, ejection fractions had significantly decreased, and myocardial mass increased by age 24 weeks. The mean Dice score of the CT segmentation results (0.925 ± 0.019, n = 40) was superior to those achieved by state-of-the-art algorithms. While further optimization and validation of the models is recommended, automated data extraction expands the use application of (μ)CT imaging, while reducing subjectivity and workload. In conclusion, the proposed method offers a pipeline to efficiently monitor disease progression in animals and humans, allowing uniform translation from preclinical to clinical studies.

ODP011 Single Nuclei RNA-Sequencing of Frozen Human Adipose Tissue and PET/CT Imaging Coupled With Deep-Learning Reveals Single-Cell Landscape and Transcriptional Dynamics Related to Adipocyte Hypertrophy and in Vivo Dysmetabolism

Abstract Background Adipocyte hypertrophy is associated with glucose and lipid dysmetabolism, ectopic fat accumulation, and metabolic complications independently of the level of obesity [1]. In this study, we aimed to explore potential mechanisms driving the obesity-independent, hypertrophy-associated dysmetabolism at a single-cell resolution and to assess the applicability of single nuclei RNA-sequencing of flash-frozen human subcutaneous adipose tissue (AT) to probe for these metabolic differences. Methods In 40 healthy participants, peri-umbilical subcutaneous adipose tissue biopsies were collected by needle aspiration to determine adipocyte size. Tissue uptake of dietary fat was quantified via oral 18-fluoro-thiaheptadecanoic acid administration with positron emission tomography (PET) [2]. Rates of fatty acid oxidation and esterification were measured using intravenous [11C]-palmitate with PET and multi-compartmental modeling. PET/CT segmentation using deep convolutional neural networks was done with DeepImageTranslator [3,4]. Single nuclei RNA-sequencing was performed in two individuals with AT hypertrophy vs. hyperplasia but matched for the same sex, ethnicity, glucose-tolerance status, BMI, total and percent body fat, and waist circumference. Nuclei were extracted from frozen AT and purified by flow cytometry following a protocol that we have developed and validated. Capture of single nuclei and preparation of RNA libraries were done with 10X Genomics technology. Results Unsupervised clustering identified a number of cell-type markers identical to those previously reported in murine epididymal AT and human AT. We found reduced quantities of fibroadipogenic cells per volume of AT with fat cell hypertrophy. Furthermore, the fraction of DPP4+ fibroblasts is increased with AT hypertrophy. Analyses of RNA velocity shows decreased velocity of adipocyte transformation from a transcriptionally active (TADA) phenotype to mature adipocytes in AT hypertrophy. Furthermore, in AT hypertrophy, adipocytes from the early TADA stage are destined to become insulin-resistant adipocytes, whereas those in AT hyperplasia are developing into an insulin-sensitive phenotype. Metabolic imaging results show that hypertrophied subcutaneous AT is associated with lower subcutaneous and higher visceral AT volume and dietary fatty acid storage capacity. Conclusion We have documented, for the first time, the applicability and consistency of single-nuclei RNA-sequencing of frozen human subcutaneous AT. We uncovered changes in cell clusters and subclusters of adipocytes, fibroblasts, pericytes, endothelial and immune cells as well as differences in transcriptional dynamics and developmental trajectory. [1] Ye RZ, et al. Endocrine reviews. 2022 Feb;43(1): 35-60. [2] Montastier É, Ye RZ, et al. American Journal of Physiology-Endocrinology and Metabolism. 2021 Jun 1;320(6): E1093-106. [3] R. Z. Ye, et al., DeepImageTranslator: A free, user-friendly graphical interface for image translation using deep-learning and its applications in 3D CT image analysis, SLAS Technology, https://doi.org/10.1016/j.slast.2021.10. 014, [4] Ye EZ, Ye EH, Ye RZ. DeepImageTranslator V2: analysis of multimodal medical images using semantic segmentation maps generated through deep learning. bioRxiv. 2021 Jan 1. Presentation: No date and time listed

Clinical Application of Individualized 3D-Printed Templates in the Treatment of Condylar Osteochondroma.

Osteochondroma (OC) is one of the most common benign tumors of the long bones, but it rarely occurs in the maxillofacial skeleton. However, mandibular condylar OC often leads to severe facial deformity in affected patients, including facial asymmetry, deviation of the chin, and malocclusion. This study aimed to explore the clinical application of individualized 3D-printed templates to accurately and effectively treat condylar OC. A total of 8 patients with mandibular condylar OC were treated from July 2015 to August 2021. The enrolled patients (5 women and 3 men) had a median age of 27 years (range: 21-32 years). All patients exhibited symptoms of facial asymmetry and occlusal disorders preoperatively. The digital software used to virtually design the process consisted of three-dimensional reconstruction, 3D-cephalometry analysis, virtual surgery, individualized templates, and postoperative facial soft-tissue prediction. A set of 3D-printed templates (DOS and DOT) were used in all cases to stabilize the occlusion and guide the osteotomy. Then, pre- and post-operative complications, mouth opening, clinical signs, and the accuracy of the CT imaging analysis were all evaluated. All the measurement data were presented as means ± SD; Bonferroni and Tamhane T2 multiple comparison tests were used to examine the differences between the groups. All patients healed uneventfully. None of the patients exhibited facial nerve injury at follow-up. In comparing the condylar segments with T0p and T1, the average deviation of the condylar segments was 0.5796 mm, indicating that the post-operative reconstructed condyles showed a high degree of similarity to the reconstruction results of the virtual surgical plan. Individualized 3D-printed templates simplified surgical procedures and improved surgical accuracy, proving to be an effective method for the treatment of patients with slight asymmetric deformities secondary to condylar OC.

Preoperative CT image analysis to improve risk stratification for clinically relevant pancreatic fistula after distal pancreatectomy.

Preoperative CT image analysis to improve risk stratification for clinically relevant pancreatic fistula after distal pancreatectomy.

Texture Analysis of CT Images in Head and Neck Tumors Differentiation

Objective: to determine the diagnostic significance of computed tomography texture analysis (CTTA) in differentiating head and neck tumors.Material and methods. The study included 118 patients aged from 4 to 80 years with a verified diagnosis of benign and malignant (37 and 81, respectively) head and neck tumors. CTTA was performed using the LIFEx program, version 6.30. Thirty eight (38) texture indices extracted from routine CT images were tested by regression analysis with creation of logistic texture models with associations of four indices as independent predictors.Results. The possibility of using derived models – probability textural indices for benign and malignant tumors differentiation was established: area under ROC-curve (AUC) 0.854 ± 0.035 (p &lt; 0.001); for differentiation of locally spread from locally limited tumors: AUC 0.840 ± 0.049 (p &lt; 0.001); for differentiation of moderately, poorly, and undifferentiated cancer (G2, G3, G4) from well-differentiated (G1) head and neck cancer: AUC 0.826 ± 0.085 (p &lt; 0.001).Conclusion. CT images texture analysis allows to make non-invasive prognosis of benign or malignant nature of a visualized head and neck tumor, as well as to determine the extent and degree of tumor malignancy.

Study on the Extraction of CT Images with Non-Uniform Illumination for the Microstructure of Asphalt Mixture.

An adaptive image-processing method for CT images of asphalt mixture is proposed in this paper. Different methods are compared according to the error analysis calculated between the real gradation and 3D reconstruction gradation. As revealed by the test results, the adaptive image-processing method was effective in carrying out different brightness homogenization processes for each image. The Wiener filter with 7 × 7 size filter was able to produce a better noise reduction effect without compromising image sharpness. Among the three methods, the adaptive image-processing method performed best in the accuracy of coarse aggregate recognition, followed by the ring division method and the global threshold segmentation method. The error of the gradation extracted by the adaptive image-processing method was found to be lowest compared with the real gradation. For a variety of engineering applications, the developed method helps to improve the analysis of CT images of asphalt mixtures.

Specific imaging features indicate the clinical features of patients with hepatic perivascular epithelioid cell tumor by comparative analysis of CT and ultrasound imaging.

The objective of the study was to explore the CT and ultrasound features and clinical significance of perivascular epithelioid cell tumor (PEComa) of the liver. Eleven hepatic PEComa patients treated in our hospital were retrospectively analyzed based on the characteristics of the imaging results of the patients, including conventional ultrasound, CDFI, contrast-enhanced ultrasound (CEUS), and contrast-enhanced CT (CECT). CT scans showed that all lesions were hypodense. Ultrasonography showed that lesions were either hyperechoic (4/11, 36.36%), hypoechoic (4/11, 36.36%), isoechoic (1/11, 9.09%), or heterogeneously echoic (2/11, 18.18%). CDFI showed that most of the lesions had an abundant blood supply (9/11, 81.82%). Whether on CT scan or ultrasonography, the margins of the lesions were dominated by clear margins. Ultrasonography revealed more features: hyperechoic patterns around lesions (3/11, 27.27%) and lateral shadow (5/11, 45.45%). The CDFI showed that large blood vessels were observed around the lesions (9/11, 81.82%). CECT shows two enhancement patterns: "fast in and fast out (FIFO)" (8/11, 72.72%) and "fast in and slow out (FISO)" (3/11, 27.27%). CEUS shows that all lesions had the enhancement pattern of "FISO," which was different from CECT. All lesions displayed rapid enhancement during HAP in CEUS during 7-20 s. Four patients (36.36%) washed out at 60-180 s, another four (36.36%) washed out at 180-300 s, and the remaining three patients (27.27%) showed no signs of washout even at 360 s. Some imaging features, such as clear margins, peripheral hyperechoic around the lesion, lateral shadow, the large blood vessels around lesions, and the "FISO" enhancement pattern, may indicate expansive growth of the tumor and be helpful in the diagnosis of PEComa. Ultrasound images may provide more details for clinical reference.

COVID-19 Diagnosis by Stationary Wavelet Entropy and Extreme Learning Machine

COVID-19 has swept the world and has had great impact on us. Rapid and accurate diagnosis of COVID-19 is essential. Analysis of chest CT images is an effective means. In this paper, an automatic diagnosis algorithm based on chest CT images is proposed. It extracts image features by stationary wavelet entropy (SWE), classifies and trains the input dataset by extreme learning machine (LEM), and finally determines the model through k-fold cross-validation (k-fold CV). By detecting 296 chest CT images of healthy individuals and COVID-19 patients, the algorithm outperforms state-of-the-art methods in sensitivity, specificity, precision, accuracy, F1, MCC, and FMI.