Computed Tomography Datasets Research Articles

An efficient quantification of COVID‐19 in chest CT images with improved semantic segmentation using U‐Net deep structure

AbstractThe worldwide spread of the coronavirus (COVID‐19) outbreak has proven devastating to public health. The severity of pneumonia relies on a rapid and accurate diagnosis of COVID‐19 in CT images. Accordingly, a computed tomography (CT) scan is an excellent screening tool for detecting COVID‐19. This paper proposes a deep learning‐based strategy for recognizing and segmenting a COVID‐19 lesion from chest CT images, which would introduce an accurate computer aided decision criteria for the physicians about the severity rate of the patients. Two main stages have been proposed for detecting COVID‐19; first, a convolutional neural network (CNN) deep structure recognizes and classifies COVID‐19 from CT images. Second, a U‐Net deep structure segments the COVID‐19 regions in a semantic manner. The proposed system is trained and evaluated on three different CT datasets for COVID‐19, two of which are used to illustrate the system's segmentation performance and the other is to demonstrate the system's classification ability. Experiment results reveal that the proposed CNN can achieve classification accuracy greater than 0.99, and the proposed U‐Net model outperforms the state‐of‐the‐art in segmentation with an IOU greater than 0.92.

Addressing the Global Expertise Gap in Radiation Oncology: The Radiation Planning Assistant.

Automation, including the use of artificial intelligence, has been identified as a possible opportunity to help reduce the gap in access and quality for radiotherapy and other aspects of cancer care. The Radiation Planning Assistant (RPA) project was conceived in 2015 (and funded in 2016) to use automated contouring and treatment planning algorithms to support the efforts of oncologists in low- and middle-income countries, allowing them to scale their efforts and treat more patients safely and efficiently (to increase access). In this review, we discuss the development of the RPA, with a particular focus on clinical acceptability and safety/risk across jurisdictions as these are important indicators for the successful future deployment of the RPA to increase radiotherapy availability and ameliorate global disparities in access to radiation oncology. RPA tools will be offered through a webpage, where users can upload computed tomography data sets and download automatically generated contours and treatment plans. All interfaces have been designed to maximize ease of use and minimize risk. The current version of the RPA includes automated contouring and planning for head and neck cancer, cervical cancer, breast cancer, and metastases to the brain. The RPA has been designed to bring high-quality treatment planning to more patients across the world, and it may encourage greater investment in treatment devices and other aspects of cancer treatment.

An Improved Combination of Faster R-CNN and U-Net Network for Accurate Multi-Modality Whole Heart Segmentation.

Detailed information of substructures of the whole heart is usually vital in the diagnosis of cardiovascular diseases and in 3D modeling of the heart. Deep convolutional neural networks have been demonstrated to achieve state-ofthe-art performance in 3D cardiac structures segmentation. However, when dealing with high-resolution 3D data, current methods employing tiling strategies usually degrade segmentation performances due to GPU memory constraints. This work develops a two-stage multi-modality whole heart segmentation strategy, which adopts an improved Combination of Faster R-CNN and 3D U-Net (CFUN+). More specifically, the bounding box of the heart is first detected by Faster R-CNN, and then the original Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) images of the heart aligned with the bounding box are input into 3D U-Net for segmentation. The proposed CFUN+ method redefines the bounding box loss function by replacing the previous Intersection over Union (IoU) loss with Complete Intersection over Union (CIoU) loss. Meanwhile, the integration of the edge loss makes the segmentation results more accurate, and also improves the convergence speed. The proposed method achieves an average Dice score of 91.1% on the Multi-Modality Whole Heart Segmentation (MM-WHS) 2017 challenge CT dataset, which is 5.2% higher than the baseline CFUN model, and achieves state-of-the-art segmentation results. In addition, the segmentation speed of a single heart has been dramatically improved from a few minutes to less than 6 seconds.

Forensic Age Estimation: A Multifactorial Approach in a Retrospective Population Study.

Objectives: The objective of this study was to evaluate the accuracy of forensic age estimation in a German population by combining clavicle and wisdom teeth assessments based on cone beam computed tomography (CT) data. The study aimed to determine the reliability of this approach in predicting biological age. Material and Methods: A total of 161 CT data sets from 120 males and 41 females with known exact ages were evaluated by three raters. The clavicle was assessed according to stages 1-5 (including substages 2a-c and 3a-c), and the Demirjian stage's classification method was used for the wisdom teeth. Inter-class correlation (ICC) was calculated to assess the agreement among the three raters. Additionally, ordinary least square regressions were performed to predict chronological age using the clavicle or one of the four teeth. Finally, age prediction models using multiple indicators were developed. Results: The ICCs ranged from 0.82 for the clavicle to 0.86 and 0.88 for the wisdom teeth. Linear estimation tended to overestimate chronological age, especially in subjects over 18 years old. The clavicle showed the strongest overestimation. Combining age estimation from the clavicle with the upper and lower wisdom teeth improved the predictive power, resulting in a 14% and 15% increase in R² for the upper and lower wisdom teeth, respectively. Adding more than one tooth to the prediction did not improve the predictive power (all ΔR² < 1%). Conclusions: Age estimation using CT can be significantly improved by combining information from the analysis of wisdom teeth with age estimation based on the clavicle.

A novel denoising method for low-dose CT images based on transformer and CNN

Computed Tomography (CT) has become a mainstream imaging tool in medical diagnosis. However, the issue of increased cancer risk due to radiation exposure has raised public concern. Low-dose computed tomography (LDCT) technique is a CT scan with lower radiation dose than conventional scans. LDCT is used to make a diagnosis of lesions with the smallest dose of x-rays, and is currently mainly used for early lung cancer screening. However, LDCT has severe image noise, and these noises affect adversely the quality of medical images and thus the diagnosis of lesions. In this paper, we propose a novel LDCT image denoising method based on transformer combined with convolutional neural network (CNN). The encoder part of the network is based on CNN, which is mainly used to extract the image detail information. In the decoder part, we propose a dual-path transformer block (DPTB), which extracts the features of input of the skip connection and the features of input of the previous level through two paths respectively. DPTB can better restore the detail and structure information of the denoised image. In order to pay more attention to the key regions of the feature images extracted at the shallow level of the network, we also propose a multi-feature spatial attention block (MSAB) in the skip connection part. Experimental studies are conducted, and comparisons with the state-of-the-art networks are made, and the results demonstrate that the developed method can effectively remove the noise in CT images and improve the image quality in the evaluation metrics of peak signal to noise ratio (PSNR), structural similarity (SSIM), and root mean square error (RMSE) and is superior to the state-of-the-art models. Our method achieved 28.9720 of PSNR, 0.8595 of SSIM and 14.8657 of RMSE on the Mayo Clinic LDCT Grand Challenge dataset. For different noise level σ (15, 35, and 55) on the QIN_LUNG_CT dataset, our proposed also achieved better performances.

Joint liver and hepatic lesion segmentation in MRI using a hybrid CNN with transformer layers

Backgound and Objective: Deep learning-based segmentation of the liver and hepatic lesions therein steadily gains relevance in clinical practice due to the increasing incidence of liver cancer each year. Whereas various network variants with overall promising results in the field of medical image segmentation have been successfully developed over the last years, almost all of them struggle with the challenge of accurately segmenting hepatic lesions in magnetic resonance imaging (MRI). This led to the idea of combining elements of convolutional and transformer-based architectures to overcome the existing limitations. Methods: This work presents a hybrid network called SWTR-Unet, consisting of a pretrained ResNet, transformer blocks as well as a common Unet-style decoder path. This network was primarily applied to single-modality non-contrast-enhanced liver MRI and additionally to the publicly available computed tomography (CT) data of the liver tumor segmentation (LiTS) challenge to verify the applicability on other modalities. For a broader evaluation, multiple state-of-the-art networks were implemented and applied, ensuring direct comparability. Furthermore, correlation analysis and an ablation study were carried out, to investigate various influencing factors on the segmentation accuracy of the presented method. Results: With Dice similarity scores of averaged 98±2% for liver and 81±28% lesion segmentation on the MRI dataset and 97±2% and 79±25%, respectively on the CT dataset, the proposed SWTR-Unet proved to be a precise approach for liver and hepatic lesion segmentation with state-of-the-art results for MRI and competing accuracy in CT imaging. Conclusion: The achieved segmentation accuracy was found to be on par with manually performed expert segmentations as indicated by inter-observer variabilities for liver lesion segmentation. In conclusion, the presented method could save valuable time and resources in clinical practice.

Ninety-degree angled collimator: a dosimetric study related to dynamic intensity-modulated radiotherapy in patients with endometrial carcinoma

BackgroundOur purpose was to ensure that the dose constraints of the organs at risk (OARs) were not exceeded while increasing the prescription dose to the planning target volume (PTV) from 45 to 50.4 Gray (Gy) with the dynamic intensity-modulated radiotherapy (IMRT) technique. While trying for this purpose, a new dynamic IMRT technique named 90° angled collimated dynamic IMRT (A-IMRT) planning was developed by us.MethodsThis study was based on the computed tomography data sets of 20 patients with postoperatively diagnosed International Federation of Gynecology and Obstetrics stage 2 endometrial carcinoma. For each patient, conventional dynamic IMRT (C-IMRT, collimator angle of 0° at all gantry angles), A-IMRT (collimator angle of 90° at gantry angles of 110°, 180°, 215°, and 285°), and volumetric modulated arc therapy (VMAT) were planned. Planning techniques were compared with parameters used to evaluate PTV and OARs via dose-volume-histogram analysis using the paired two-tailed Wilcoxon’s signed-rank test; p < 0.05 was considered indicative of statistical significance.ResultsAll plans achieved adequate dose coverage for PTV. Although the technique with the lowest mean conformality index was A-IMRT (0.76 ± 0.05) compared to both C-IMRT (0.79 ± 0.04, p = 0.000) and VMAT (0.83 ± 0.03, p = 0.000), it protected the OARs especially the bladder (V45 = 32.84 ± 2.03 vs. 44.21 ± 6.67, p = 0.000), rectum (V30 = 56.18 ± 2.05 vs. 73.80 ± 4.75, p = 0.000) and both femoral heads (V30 for right = 12.19 ± 1.34 vs. 21.42 ± 4.03, p = 0.000 and V30 for left = 12.58 ± 1.48 vs. 21.35 ± 4.16, p = 0.000) better than C-IMRT. While the dose constraints of the bladder, rectum and bilateral femoral heads were not exceeded in any patient with A-IMRT or VMAT, they were exceeded in 19 (95%), 20 (100%) and 20 (100%) patients with C-IMRT, respectively.ConclusionsOARs are better protected when external beam radiotherapy is applied to the pelvis at a dose of 50.4 Gy by turning the collimator angle to 90° at some gantry angles with the dynamic IMRT technique in the absence of VMAT.

TDS-U-Net: Automatic liver and tumor separate segmentation of CT volumes using attention gates1

Considering the heterogeneity, diffusive shape, and complex background of tumors, automatic segmentation of hepatic lesions in computed tomography (CT) images has been considered a challenging task. The performance of existing methods remains subject to segmentation uncertainties, especially in tumor boundary regions. The pixel information in these regions will be affected by both sides, thereby exposing the segmentation results to missing marks. To this end, a new network architecture named Two Direction Segmentation U-Net (TDS-U-Net) is hereby designed based on the classic Attention U-Net to tackle this problem. As the most important blocks of the Attention U-Net network, attention gates (AGs) focus on the target structures of different shapes and sizes. In the last layer of TDS-U-Net, two dichotomous convolutional networks are applied to obtain the segmentation maps of the liver and the tumor respectively. Superimposing two segmented maps to obtain the final image addresses the above problems. The entire structure has been verified on two widely accepted public CT datasets, LiTS17 and KiTS19. Compared with the state of the art, this method exhibits superior performance and excellent shape extractions with high detection sensitivity, perfectly demonstrating its effectiveness in medical image segmentation.

Comparison between single and serial computed tomography images in classification of acute appendicitis, acute right-sided diverticulitis, and normal appendix using EfficientNet.

This study aimed to develop a convolutional neural network (CNN) using the EfficientNet algorithm for the automated classification of acute appendicitis, acute diverticulitis, and normal appendix and to evaluate its diagnostic performance. We retrospectively enrolled 715 patients who underwent contrast-enhanced abdominopelvic computed tomography (CT). Of these, 246 patients had acute appendicitis, 254 had acute diverticulitis, and 215 had normal appendix. Training, validation, and test data were obtained from 4,078 CT images (1,959 acute appendicitis, 823 acute diverticulitis, and 1,296 normal appendix cases) using both single and serial (RGB [red, green, blue]) image methods. We augmented the training dataset to avoid training disturbances caused by unbalanced CT datasets. For classification of the normal appendix, the RGB serial image method showed a slightly higher sensitivity (89.66 vs. 87.89%; p = 0.244), accuracy (93.62% vs. 92.35%), and specificity (95.47% vs. 94.43%) than did the single image method. For the classification of acute diverticulitis, the RGB serial image method also yielded a slightly higher sensitivity (83.35 vs. 80.44%; p = 0.019), accuracy (93.48% vs. 92.15%), and specificity (96.04% vs. 95.12%) than the single image method. Moreover, the mean areas under the receiver operating characteristic curve (AUCs) were significantly higher for acute appendicitis (0.951 vs. 0.937; p < 0.0001), acute diverticulitis (0.972 vs. 0.963; p = 0.0025), and normal appendix (0.979 vs. 0.972; p = 0.0101) with the RGB serial image method than those obtained by the single method for each condition. Thus, acute appendicitis, acute diverticulitis, and normal appendix could be accurately distinguished on CT images by our model, particularly when using the RGB serial image method.

Targeted self-supervised attention network for coronavirus disease 2019 detection

The worldwide spread of the coronavirus disease 2019 (COVID-19) has had a significant impact on healthcare systems globally. Timely and accurate diagnosis is crucial for controlling and preventing outbreaks. Computed tomography (CT) has been demonstrated as an effective diagnostic tool for COVID-19. Deep learning techniques have been applied in related fields to solve relevant problems. Normally, they achieve good performance on large datasets directly. However, due to the need to protect patient privacy and the sudden outbreak of the epidemic, the amount of publicly available CT data is insufficient for utilizing these methods. In addition, some natural images, which are used for model pre-training, are quite different from chest CT images, causing some inevitable errors. To address these issues, a targeted self-supervised attention network based on an improved inception ResNet called SIRANet is proposed. SIRANet incorporates a self-supervised learning method during the pre-training process and introduces a attention module to improve feature selection and focus on key informative features. Moreover, a transfer learning-based strategy for COVID-19 detection is presented and experiments are conducted on three widely used datasets, including ImageNet, an unlabeled chest CT dataset, and a COVID-19 CT dataset. Experimental results show that SIRANet outperforms other state-of-the-art techniques in terms of accuracy, precision, recall, and f1-score, demonstrating its effectiveness in capturing key informative features from unlabeled and labeled images. The implementation of SIRANet is available at https://github.com/Xiiin6/COVID-19.

An improved 3D KiU-Net for segmentation of liver tumor

It is a challenging task to accurately segment liver tumors from Computed Tomography (CT) images. The widely used U-Net and its variants generally suffer from the issue to accurately segment the detailed edges of small tumors, because the progressive down sampling operations in the encoder module will gradually increase the receptive fields. These enlarged receptive filed have limited ability to learn the information about tiny structures. KiU-Net is a newly proposed dual-branch model that can effectively perform image segmentation for small targets. However, the 3D version of KiU-Net has high computational complexity, which limits its application. In this work, an improved 3D KiU-Net (named TKiU-NeXt) is proposed for liver tumor segmentation from CT images. In TKiU-NeXt, a Transformer-based Kite-Net (TK-Net) branch is proposed to build the over-complete architecture to learn more detailed features for small structures, and an extended 3D version of UNeXt is developed to replace the original U-Net branch, which can effectively reduce computational complexity but still with superior segmentation performance. Moreover, a Mutual Guided Fusion Block (MGFB) is designed to effectively learn more features from two branches and then fuse the complementary features for image segmentation. The experimental results on two public CT datasets and a private dataset demonstrate that the proposed TKiU-NeXt outperforms all the compared algorithms, and it also has less computational complexity. It suggests the effectiveness and efficiency of TKiU-NeXt.

Artificial Intelligence in CT and MR Imaging for Oncological Applications.

Cancer care increasingly relies on imaging for patient management. The two most common cross-sectional imaging modalities in oncology are computed tomography (CT) and magnetic resonance imaging (MRI), which provide high-resolution anatomic and physiological imaging. Herewith is a summary of recent applications of rapidly advancing artificial intelligence (AI) in CT and MRI oncological imaging that addresses the benefits and challenges of the resultant opportunities with examples. Major challenges remain, such as how best to integrate AI developments into clinical radiology practice, the vigorous assessment of quantitative CT and MR imaging data accuracy, and reliability for clinical utility and research integrity in oncology. Such challenges necessitate an evaluation of the robustness of imaging biomarkers to be included in AI developments, a culture of data sharing, and the cooperation of knowledgeable academics with vendor scientists and companies operating in radiology and oncology fields. Herein, we will illustrate a few challenges and solutions of these efforts using novel methods for synthesizing different contrast modality images, auto-segmentation, and image reconstruction with examples from lung CT as well as abdome, pelvis, and head and neck MRI. The imaging community must embrace the need for quantitative CT and MRI metrics beyond lesion size measurement. AI methods for the extraction and longitudinal tracking of imaging metrics from registered lesions and understanding the tumor environment will be invaluable for interpreting disease status and treatment efficacy. This is an exciting time to work together to move the imaging field forward with narrow AI-specific tasks. New AI developments using CT and MRI datasets will be used to improve the personalized management of cancer patients.

Laser Target System in Combination with an Aiming Device for Percutaneous CT-Guided Interventions – An Accuracy Study

To evaluate the targeting accuracy of laser-guided punctures in combination with an aiming device for computed tomography (CT) interventions during in vitro experiments. MATERIALS ANDMETHODS: A total of 600 CT-guided punctures were performed using a laser target system, half of them with the additional help of an aiming device. Conically shaped targets in a plexiglass phantom were punctured. The planning CT data sets were acquired with 1.25, 2.5 and 5mm slice thickness. Needle placement accuracy, as well as procedural time, wasassessed. The Euclidean (ED) and normal distances (ND) were calculated at the target point. Using the aiming device, the accomplished mean ND at the target for the 1.25, 2.5 and 5mm slice thickness was 1.76mm (SD±0.92), 2.09mm (SD±1.06) and 1.93mm (SD±1.38), respectively. Without aiming device, the corresponding results were 2.55mm (SD±1.42), 2.7mm (SD±1.43) and 2.31mm (SD±1.64). At a slice thickness of 1.25mm and 2.5mm, punctures with the aiming device were significantly more accurate for both the ED and ND as compared to the punctures without aiming device (p<0.001).The mean time required to complete the procedure, including image acquisition, trajectory planning, the placement of 10 needles, and the control-CT scan was 24.8min without and 29.8min with the aiming device. The additional use of the aiming device in combination with the commercially available laser guidance system significantly increased the level of accuracy during this in vitro experiment compared to freehand passes.

Computed tomographic angular measurements using a bone-centered three-dimensional coordinate system are accurate in a femoral torsional deformity model and precise in clinical canine patients.

In small animal orthopedics, angular measurements in the canine femur are often applied in clinical patients with bone deformities and especially in complex and severe cases. Computed tomography (CT) has been shown to be more precise and accurate than two-dimensional radiography, and several methods are described. Measurement techniques evaluated in normal bones must prove accuracy in deformed bones in clinical settings. The goals of our study were to evaluate the accuracy of canine femoral torsion angle measurements in a femoral torsional deformity model and to test repeatability and reproducibility of canine femoral neck inclination, torsion, and varus angle measurements in CT datasets of dogs applying a CT-based technique using a three-dimensional (3D) bone-centered coordinate system. For precision testing, femoral torsion, femoral neck inclination, and femoral varus angles were measured in CT data of 68 canine hind limbs by two operators, and their results were compared. For accuracy testing, a femoral torsional deformity model was preset from 0° to +/-90° with a goniometer and scanned. Torsion angles were measured in the CT data and compared to the preset value. In the femoral torsion model, the Bland-Altman plots demonstrated a mean difference of 2.11°, and the Passing-Bablok analysis demonstrated a correlation between goniometer and CT-based measurements. In the clinical CT scans, intra- and interobserver agreement resulted in coefficients of variation for repeated measurements (%) between 1.99 and 8.26 for the femoral torsion, between 0.59 and 4.47 for the femoral neck inclination, and between 1.06 and 5.15 for the femoral varus angles. Evaluation of femoral malformations with torsional deformities is the target area of this technique. Further studies are required to assess its value in different types, degrees, and combinations of osseous deformities and to establish normal reference values and guidelines for corrective osteotomies. Based on the results of this study, the accuracy of the torsion angle measurements and the precision of inclination, torsion, and the varus angle measurements were considered acceptable for clinical application.

Ensemble deep honey architecture for COVID-19 prediction using CT scan and chest X-ray images.

Recently, the infectious disease COVID-19 remains to have a catastrophic effect on the lives of human beings all over the world. To combat this deadliest disease, it is essential to screen the affected people quickly and least inexpensively. Radiological examination is considered the most feasible step toward attaining this objective; however, chest X-ray (CXR) and computed tomography (CT) are the most easily accessible and inexpensive options. This paper proposes a novel ensemble deep learning-based solution to predict the COVID-19-positive patients using CXR and CT images. The main aim of the proposed model is to provide an effective COVID-19 prediction model with a robust diagnosis and increase the prediction performance. Initially, pre-processing, like image resizing and noise removal, is employed using image scaling and median filtering techniques to enhance the input data for further processing. Various data augmentation styles, such as flipping and rotation, are applied to capable the model to learn the variations during training and attain better results on a small dataset. Finally, a new ensemble deep honey architecture (EDHA) model is introduced to effectively classify the COVID-19-positive and -negative cases. EDHA combines three pre-trained architectures like ShuffleNet, SqueezeNet, and DenseNet-201, to detect the class value. Moreover, a new optimization algorithm, the honey badger algorithm (HBA), is adapted in EDHA to determine the best values for the hyper-parameters of the proposed model. The proposed EDHA is implemented in the Python platform and evaluates the performance in terms of accuracy, sensitivity, specificity, precision, f1-score, AUC, and MCC. The proposed model has utilized the publicly available CXR and CT datasets to test the solution's efficiency. As a result, the simulated outcomes showed that the proposed EDHA had achieved better performance than the existing techniques in terms of Accuracy, Sensitivity, Specificity, Precision, F1-Score, MCC, AUC, and Computation time are 99.1%, 99%, 98.6%, 99.6%, 98.9%, 99.2%, 0.98, and 820s using the CXR dataset.

Split X-Jaw techniques of volumetric modulated arc radiotherapy in nasopharyngeal cancer: A dosimetric comparison.

The current study aims to compare the split x-jaw planning technique of volumetric modulated arc radiotherapy (VMAT) with the traditional open and limited jaw techniques of VAMT in nasopharyngeal carcinoma treatment. The multi-leaf collimators on the varian linear accelerator move on a carriage with a maximum leaf span of 15 cm. Therefore, treatment of larger planning target volumes, such as in nasopharyngeal cancer with traditional open and limited jaw technique, yields compromised dose distribution. Computed tomography data sets of 10 nasopharynx cancer patients were enrolled for the study. For each case, three separate treatment plans were generated viz. open, limited, and split x-jaw planning techniques with similar planning objectives. Only PTVs requiring a field size larger than 18 cm in the x-jaw position were considered. Comparable results were obtained regarding organs at risk (OAR) sparing in all the techniques. The target dose coverage with split x-jaw VMAT was superior to both open and limited jaw planning techniques, with a statistically significant difference in the intermediate dose planning target volumes (PTVs) (PTV59.4), P < 0.05. However, the split technique's dose to the spinal cord and larynx was significantly lower (P < 0.05). The split x-jaw planning technique of VMAT can be adapted for larger PTVs requiring an x-jaw of more than 15 cm. The only concern with this technique is the increased MU.

Anatomical Variants Identified on Computed Tomography of Oropharyngeal Carcinoma Patients.

Background and Objectives: Anatomical variations in the head, neck and chest are common, and are observed as occasional findings on computed tomography (CT). Although anatomical variations are mostly asymptomatic and do not cause any negative influence on the body function, they may jeopardize diagnosis and may be confused with pathological conditions. The presence of variations may also limit surgical access during tumor removal. The aim of this study was to investigate the prevalence of six anatomical variations-os acromiale, episternal ossicles, cervical rib, Stafne bone cavity, azygos lobe and tracheal bronchus-in an open-access computed tomography dataset obtained from oropharyngeal cancer patients. Materials and Methods: A total of 606 upper-chest and neck computed-tomography scans (79.4% male and 20.6% female) were retrospectively investigated. Sex difference was evaluated using the z-test for two proportions. Results: Os acromiale, episternal ossicles, cervical rib, Stafne bone cavity, azygos lobe, and tracheal bronchus were present in 3.1%, 2.2%, 0.2%, 0%, 0.3% and 0.5%, respectively, of all patients. Os acromiale was identified as meso-acromion in 86.6%, and as pre-acromion in 17.4%, of all acromia. Episternal ossicles were present unilaterally in 58.3%, and bilaterally in 41.7%, of all sterna. Only the cervical rib showed a sex difference in prevalence. Conclusions: awareness of these variations is important for radiologists interpreting head, neck and chest CTs; for example, those of oropharyngeal cancer patients. This study also illustrates the applicability of publicly available datasets in prevalence-based anatomical research. While most of the variations investigated in the present study are well-known, the episternal ossicles are not well explored, and need further investigation.

Unsupervised Visual Representation Learning Based on Segmentation of Geometric Pseudo-Shapes for Transformer-Based Medical Tasks.

Recently, transformer-based architectures have been shown to outperform classic convolutional architectures and have rapidly been established as state-of-the-art models for many medical vision tasks. Their superior performance can be explained by their ability to capture long-range dependencies of their multi-head self-attention mechanism. However, they tend to overfit on small- or even medium-sized datasets because of their weak inductive bias. As a result, they require massive, labeled datasets, which are expensive to obtain, especially in the medical domain. This motivated us to explore unsupervised semantic feature learning without any form of annotation. In this work, we aimed to learn semantic features in a self-supervised manner by training transformer-based models to segment the numerical signals of geometric shapes inserted on original computed tomography (CT) images. Moreover, we developed a Convolutional Pyramid vision Transformer (CPT) that leverages multi-kernel convolutional patch embedding and local spatial reduction in each of its layer to generate multi-scale features, capture local information, and reduce computational cost. Using these approaches, we were able to noticeably outperformed state-of-the-art deep learning-based segmentation or classification models of liver cancer CT datasets of 5,237 patients, the pancreatic cancer CT datasets of 6,063 patients, and breast cancer MRI dataset of 127 patients.

Clinical aspects and cytomorphological and functional features of the nasal mucosa in chronic pathology of the intranasal structures and their verification from CT data

Background. Approximately 30% of the general human population suffers from chronic pathology of intranasal structures, the main manifestations of which are impaired nasal breathing and sense of smell. The main instrumental diagnostic methods for this pathology are X-ray computed tomography (CT), which allows obtaining data on the architecture of the anatomical structures of the upper respiratory tract, and rhinomanometry, based on the results of which it is possible to assess the functional capacity of the nasal cavity during breathing. Also, a thorough study of the cytological material of the mucous membrane of the upper respiratory tract is an important component in determining the functional state of the nasal cavity, clarifying the diagnosis and choosing a treatment method. This allows the doctor to determine the composition and number of cellular elements in the material, assess their condition (destruction, proliferation, dystrophy, necrosis, etc.), ascertain the intensity of the body’s reactive abilities, monitor the dynamics of tissue recovery or the healing process in them, and encourages researchers to study in more detail and comparing the materials of clinical, radiological and cytological studies with the aim of developing a pathogenetically directed complex treatment of patients with nasal breathing disorders. Therefore it is necessary to know aspects of correlation between the results of rhinocytography and CT data in typical pathological conditions with nasal congestion are considered. Purpose – is to study the clinical aspects and cytomorphological and functional features of the nasal mucosa in patients with pathology of intranasal structures with respiratory and olfactory disorders and research their independent verification based on the CT data. Materials and Methods. Clinical examination of patients included the study of complaints, anamnesis of the disease, examination of the ENT organs, rhinomanometry, endoscopic examination of the nasal cavity and nasopharynx, The CT of the paranasal sinuses using 3D cone beam tomography on the Vatech PaX-i3D device, as well as cytological examination of the nasal mucosa. The criteria for participation in the study were the absence of chronic diseases of the cardiovascular, respiratory, digestive, urinary systems, as well as heredity burdened by these diseases. Results. Formation of a different nature of the course and severity of disorders is associated with inflammatory, dyscirculatory and trophic disorders in the nasal mucosa, which weaken both mucociliary clearance and local immunity. This applies mainly to the I group of observations. The consequence of a decrease in local immunity factors in the nasal mucosa is microbial contamination, which is associated with a long-term nasal breathing disorder in the I and II groups of observations, up to five years and six months, respectively. The results of rhinocytography mostly correspond with the aerodynamic models data of nasal сavity from the CT datasets. Conclusions. Despite the reliability of the examinations carried out by us, the cytological examination of the nasal mucosa is only an additional analysis, the interpretation of which should be based on the clinical picture of a particular patient. Proposed aerodynamic model from CT-datasets actually provides an independent verification of the aerodynamic characteristics of the nasal cavity, obtained from rhinomanometry data, and may indicate a violation of nasal breathing according to changes in the internal anatomical configuration of the nasal chanel.

REBOA Zone Estimation from the Body Surface Using Semantic Segmentation

Resuscitative endovascular balloon occlusion of the aorta (REBOA) is an endovascular procedure for hemorrhage control. In REBOA, the balloon must be placed in the precise place, but it may be performed without X-ray fluoroscopy. This study aimed to estimate the REBOA zones from the body surface using deep learning for safe balloon placement. A total of 198 abdominal computed tomography (CT) datasets containing the regions of the REBOA zones were collected from open data libraries. Then, depth images of the body surface generated from the CT datasets and the images corresponding to the zones were labeled for deep learning training and validation. DeepLabV3+, a deep learning semantic segmentation model, was employed to estimate the zones. We used 176 depth images as training data and 22 images as validation data. A nine-fold cross-validation was performed to generalize the performance of the network. The median Dice coefficients for Zones 1-3 were 0.94 (inter-quarter range: 0.90-0.96), 0.77 (0.60-0.86), and 0.83 (0.74-0.89), respectively. The median displacements of the zone boundaries were 11.34 mm (5.90-19.45), 11.40 mm (4.88-20.23), and 14.17 mm (6.89-23.70) for the boundary between Zones 1 and 2, between Zones 2 and 3, and between Zone 3 and out of zone, respectively. This study examined the feasibility of REBOA zone estimation from the body surface only using deep learning-based segmentation without aortography.