Computed Tomography Segmentation Research Articles

Topological structure and global features enhanced graph reasoning model for non-small cell lung cancer segmentation from CT

Objective. Accurate and automated segmentation of lung tumors from computed tomography (CT) images is critical yet challenging. Lung tumors are of various sizes and locations and have indistinct boundaries adjacent to other normal tissues. Approach. We propose a new segmentation model that can integrate the topological structure and global features of image region nodes to address the challenges. Firstly, we construct a weighted graph with image region nodes. The graph topology reflects the complex spatial relationships among these nodes, and each node has its specific attributes. Secondly, we propose a node-wise topological feature learning module based on a new graph convolutional autoencoder (GCA). Meanwhile, a node information supplementation (GNIS) module is established by integrating specific features of each node extracted by a convolutional neural network (CNN) into each encoding layer of GCA. Afterwards, we construct a global feature extraction model based on multi-layer perceptron (MLP) to encode the features learnt from all the image region nodes which are crucial complementary information for tumor segmentation. Main results. Ablation study results over the public lung tumor segmentation dataset demonstrate the contributions of our major technical innovations. Compared with other segmentation methods, our new model improves the segmentation performance and has generalization ability on different 3D image segmentation backbones. Our model achieved Dice of 0.7827, IoU of 0.6981, and HD of 32.1743 mm on the public dataset 2018 Medical Segmentation Decathlon challenge, and Dice of 0.7004, IoU of 0.5704 and HD of 64.4661 mm on lung tumor dataset from Shandong Cancer Hospital. Significance. The novel model improves automated lung tumor segmentation performance especially the challenging and complex cases using topological structure and global features of image region nodes. It is of great potential to apply the model to other CT segmentation tasks.

Liver Tumors Segmentation Using 3D SegNet Deep Learning Approach

An ultrasonic filter detects signs of malignant tumors by analysing the image’s pixel quality fluctuations caused by a liver ailment. Signs of malignant growth proximity are identified in an ultrasound filter through image pixel quality variations from a liver’s condition. Those changes are more common in alcoholic liver conditions than in other etiologies of cirrhosis, suggesting that the cause may be alcohol instead of liver disease. Existing Two-Dimensional (2D) ultrasound data sets contain an accuracy rate of 85.9% and a 2D Computed Tomography (CT) data set of 91.02%. The most recent work on designing a Three-Dimensional (3D) ultrasound imaging system in or close to real-time is examined. In this article, a Deep Learning (DL) model is implemented and modified to fit liver CT segmentation, and a semantic pixel classification of road scenes is recommended. The architecture is called semantic pixel-wise segmentation and comprises a hierarchical link of encoder-decoder layers. A standard data set was used to test the proposed model for liver CT scans and the tumor accuracy in the training phase. For the normal class, we obtained 100% precision for chronic cirrhosis hepatitis (73%), offset cirrhosis (59.26%), and offensive cirrhosis (91.67%) for chronic hepatitis or cirrhosis (73,0%). The aim is to develop a Computer-Aided Detection (CAD) screening tool to detect steatosis. The results proved 98.33% exactness, 94.59% sensitivity, and 92.11% case with Convolutional Neural Networks (CNN) classification. Although the classifier’s performance did not differentiate so clearly at this level, it was recommended that CNN generally perform better due to the good relationship between Area under the Receiver Operating Characteristics Curve (AUC) and accuracy.

Pore scale investigation of unsaturated granular soil behaviour by means ofin situCT experiments

With continuing evolution of imaging techniques from medical applications and materials science, non-destructive imaging experiments have also become an important method to investigate soil specimens. Amongst other methods, computed tomography (CT) has developed to a tool to visualise and better understand the microstructure of different soils based on 3D image data. Furthermore, the acquisition of a temporal series of CT images allows to study processes in soils on the microscale, e. g., during mechanical loading. In order to study the hydraulic and mechanical behaviour of unsaturated granular soils, we combine different custom-built miniaturised experimental set-ups with geomechanics background with computed tomography, yielding so-calledin situCT experiments. By means of image reconstruction and further image analysis based on segmented CT images acquired during different hydraulic and mechanical experiments, we study the drainage and imbibition process as well as the shear process of unsaturated sand and glass bead specimens on the pore or grain scale. The analysis of data on the microscopic level, including the phase distribution, interfacial areas, contact lines as well as radii of curvature of capillary menisci, allows to obtain insights into the macroscopic water retention behaviour and shear behaviour of granular soils.

Analysis of Deep Learning Techniques for Brain Tumour Classification from CT & MRI Images

Brain tumour detection in an initialpoint is a critical step to saving human life. Computed Tomography (CT) and Magnetic Resonance Image (MRI) provide very detailed information about brain tumour tissues. So the segmentation of tumour region is possible from pancreatic CT and brain MRI. CT and MRI is a non-invasive technique and it does not produce any harmful radiation to the patient. The patient suspected of tumour undergoes radiological evaluation such that the area, location and grade of the tumour can be predicted from the CT and MRI analysis. This critical information helps the doctors to decide about further treatment like chemotherapy, surgery, or radiation. The diagnosis requires an accurate and very fast segmentation and classification of CT and MRI images. But nowadays radiologists are doing this task manually and it is a tedious and time-consuming procedure. Also, there is a chance of variation in the result from one expert to another. Here comes the significance of automatic segmentation and classification of tumour types with the help of computers.The proposed work aims to develop an efficient system that can detect pancreatic and brain tumour and can classify the pancreatic CT and brain MRI into normal, benign or malignant. This work can be categorized into two approaches. Thus the dataset prepared for this research work contains CT and MRI images.The first approach proposes traditional machine learning technologies to achieve the goal. Image pre-processing, feature extraction, segmentation and classification are the various steps of the traditional machine learning method. A detailed investigation is performed through various feature extraction techniques and classification techniques for pancreatic (CT) and brain MRI. Discrete Wavelet Transform (DWT) feature, Grey Level Co-occurrence Matrix (GLCM) feature, Gabor feature, Tamura features and Edge Orientation Histogram (EOH) features and their combinations are used for the extraction of CT and MRI features. Benign tumours are non-cancerous, but malignant tumours are cancerous. In the first approach, the Support Vector Machine (SVM) is the main classifier used for pancreatic CT and brain MRI classification as normal, benign or malignant.In this technology, a huge amount of data and machines with high computational capabilities like Graphic Processing Unit (GPU) are available. Thus the second approach of this paper is to exploit all these available resources to produce accurate results. In this part, deep learning, the latest fast growing technology introduced in 2015 is used for the classification of brain MRI. A Deep Convolutional Neural Networks (DCNN) model is proposed to perform the classification task efficiently. The CNN results are compared with the results of a simple neural network classifier. This method provides accurate and it shows that deep learning based classification outperforms traditional machine learning techniques which produce an accurate result only. This research work again concentrates on the Transfer Learning (TL) methods to classify pancreatic CT and brain MRI.

Diagnostic Accuracy of AI for Opportunistic Screening of Abdominal Aortic Aneurysm in CT: A Systematic Review and Narrative Synthesis.

In this review, we focused on the applicability of artificial intelligence (AI) for opportunistic abdominal aortic aneurysm (AAA) detection in computed tomography (CT). We used the academic search system PubMed as the primary source for the literature search and Google Scholar as a supplementary source of evidence. We searched through 2 February 2022. All studies on automated AAA detection or segmentation in noncontrast abdominal CT were included. For bias assessment, we developed and used an adapted version of the QUADAS-2 checklist. We included eight studies with 355 cases, of which 273 (77%) contained AAA. The highest risk of bias and level of applicability concerns were observed for the "patient selection" domain, due to the 100% pathology rate in the majority (75%) of the studies. The mean sensitivity value was 95% (95% CI 100-87%), the mean specificity value was 96.6% (95% CI 100-75.7%), and the mean accuracy value was 95.2% (95% CI 100-54.5%). Half of the included studies performed diagnostic accuracy estimation, with only one study having data on all diagnostic accuracy metrics. Therefore, we conducted a narrative synthesis. Our findings indicate high study heterogeneity, requiring further research with balanced noncontrast CT datasets and adherence to reporting standards in order to validate the high sensitivity value obtained.

U-NetCTS: U-Net deep neural network for fully automatic segmentation of 3D CT DICOM volume

The accurate segmentation of computed tomography (CT) scan volume is an essential step in radiomic analysis as well as in developing advanced surgical planning techniques with numerous medical applications. When this process is performed manually by a clinician, it is laborious, time consuming, prone to error, and its success depends to a large extent on the level of experience. In this work, we propose an automated deep learning (DL) segmentation framework for CT images called U-Net CT Segmentation (U-NetCTS) to combine the DL U-Net and CT images in the domain of automatic segmentation. Experimental results show that U-NetCTS framework can segment different CT DICOM image regions of interest in a range of random CT volumes. A statistical and qualitative comparison of the CT slices automatically segmented by U-NetCTS framework and ground-truth images indicates that U-NetCTS framework achieves a high level of accuracy, where the Tanimoto coefficient, dice similarity coefficient, and peak signal-to-noise ratio values are 99.06%, 99.52%, and 53.29 dB, respectively. The DC value is also higher than that of state-of-the-art DL techniques for automatic segmentation of CT images of various human organs. Furthermore, a total amount of 3595 CT slices is employed in this study with various CT region of interest to validate the results.

Neuroanatomy of dementia in Down Syndrome revealed using a voxel‐based morphometry approach to computed tomography

AbstractBackgroundPeople with Down syndrome (DS) are more likely to develop Alzheimer’s disease than the general population, but comorbidities and variable intellectual disability complicate diagnosis and disease monitoring. Although neuroimaging may help elucidate the pathogenesis of dementia, studies using magnetic resonance imaging (MRI) in DS subjects with dementia are limited due to greater motion artefact in those with severe cognitive impairment. Computed tomography (CT) is faster, thus less impacted by motion, but poor tissue contrast makes analysis of cortical and subcortical changes difficult. In this study, we use CTseg, an algorithm able to robustly spatially normalise and segment CT scans, to analyse grey matter (GM) and white matter (WM) differences in 98 DS subjects with and without cognitive impairment due to dementia.MethodsThe CT scans of 120 DS patients from the Adult Down Syndrome unit at the Hospital Universitario La Princesa were selected. Dementia level was classified by retrospective review of medical records as 0 (cognitively stable), 1 (mild dementia), 2 (moderate dementia), or 3 (advanced dementia) according to the DSM‐V operational criteria. Premorbid functionality was assessed using Part 1 (“Level of ‘best’ ability”) of the Dementia Screening Questionnaire for Individuals with Intellectual Disabilities (DSQIID). Those with artefacts, faulty scans, or incidental brain findings were excluded, leaving a final sample of 98 subjects (mean age: 48±8.5 years; 45% female; 62 cognitively stable). CT scans were segmented into GM, WM, and cerebrospinal fluid using CTseg (https://github.com/WCHN/CTseg). Voxel‐based morphometry (VBM) analyses were carried out separately on smoothed and modulated GM and WM images, correlating with dementia level whilst controlling for age, sex, DSQIID, and total intracranial volume.ResultsGM volume was significantly negatively correlated with dementia level in the bilateral parahippocampal gyri, extending into hippocampus, and amygdalae. By contrast, significant WM reductions associated with dementia level were primarily parietal and occipitotemporal and right‐lateralised in the superior and inferior longitudinal fasciculi (p<0.001, FWE‐corrected).ConclusionsOur study uses a novel algorithm that allows the analysis of CT scans using VBM methods, enabling the inclusion of a sizeable sample of DS subjects with dementia, a group difficult to study with neuroimaging due to practical constraints.

Neuroanatomy of dementia in Down Syndrome revealed using a voxel‐based morphometry approach to computed tomography

AbstractBackgroundPeople with Down syndrome (DS) are more likely to develop Alzheimer’s disease than the general population, but comorbidities and variable intellectual disability complicate diagnosis and disease monitoring. Although neuroimaging may help elucidate the pathogenesis of dementia, studies using magnetic resonance imaging (MRI) in DS subjects with dementia are limited due to greater motion artefact in those with severe cognitive impairment. Computed tomography (CT) is faster, thus less impacted by motion, but poor tissue contrast makes analysis of cortical and subcortical changes difficult. In this study, we use CTseg, an algorithm able to robustly spatially normalise and segment CT scans, to analyse grey matter (GM) and white matter (WM) differences in 98 DS subjects with and without cognitive impairment due to dementia.MethodsThe CT scans of 120 DS patients from the Adult Down Syndrome unit at the Hospital Universitario La Princesa were selected. Dementia level was classified by retrospective review of medical records as 0 (cognitively stable), 1 (mild dementia), 2 (moderate dementia), or 3 (advanced dementia) according to the DSM‐V operational criteria. Premorbid functionality was assessed using Part 1 (“Level of ‘best’ ability”) of the Dementia Screening Questionnaire for Individuals with Intellectual Disabilities (DSQIID). Those with artefacts, faulty scans, or incidental brain findings were excluded, leaving a final sample of 98 subjects (mean age: 48±8.5 years; 45% female; 62 cognitively stable). CT scans were segmented into GM, WM, and cerebrospinal fluid using CTseg (https://github.com/WCHN/CTseg). Voxel‐based morphometry (VBM) analyses were carried out separately on smoothed and modulated GM and WM images, correlating with dementia level whilst controlling for age, sex, DSQIID, and total intracranial volume.ResultsGM volume was significantly negatively correlated with dementia level in the bilateral parahippocampal gyri, extending into hippocampus, and amygdalae. By contrast, significant WM reductions associated with dementia level were primarily parietal and occipitotemporal and right‐lateralised in the superior and inferior longitudinal fasciculi (p<0.001, FWE‐corrected).ConclusionsOur study uses a novel algorithm that allows the analysis of CT scans using VBM methods, enabling the inclusion of a sizeable sample of DS subjects with dementia, a group difficult to study with neuroimaging due to practical constraints.

Novel CT-based parameters assessing relative cross-sectional area to guide surgical management and predict clinical outcomes in patients with acute subdural hematoma.

Acute subdural hematoma (aSDH) is one of the most devastating entities secondary to traumatic brain injury (TBI). Even though radiological computed tomography (CT) findings, such as hematoma thickness (HT), midline shift (MLS), and MLS/HT ratio, have an important prognostic role, they suffer from important drawbacks. We hypothesized that relative cross-sectional area (rCSA) of specific brain regions would provide valuable information about brain compression and swelling, thus being a key determining factor governing the clinical course. We performed an 8-year retrospective analysis of patients with moderate to severe TBI with surgically evacuated, isolated, unilateral aSDH. We investigated the influence of aSDH rCSA and ipsilateral hemisphere rCSA along the supratentorial region on the subsequent operative technique employed for aSDH evacuation and patient's clinical outcomes (early death and Glasgow Outcome Scale [GOS] at discharge and after 1-year follow-up). Different conventional radiological variables were also assessed. The study included 39 patients. Lower HT, MLS, hematoma volume, and aSDH rCSA showed a significant association with decompressive craniectomy (DC) procedure. Conversely, higher ipsilateral hemisphere rCSA along the dorso-ventral axis and, specifically, ipsilateral hemisphere rCSA at the high convexity level were predictors for DC. CT segmentation analysis exhibited a modest relationship with early death, which was limited to the basal supratentorial subregion, but could not predict long-term outcome. rCSA is an objectifiable and reliable radiologic parameter available on admission CT that might provide valuable information to optimize surgical treatment.

Abstract 15312: Automatic Left Atrial Volume and Sphericity Index Calculation From Cardiac CT Using Computer Graphics Imaging and Deep Learning

Introduction: Left atrial (LA) volume and sphericity index are important for atrial fibrillation (AF) recurrence prediction after ablation and AF-associated atrial remodeling analysis. However, calculating these indexes from computed tomography (CT) requires manual outlining, which is subjective and difficult to scale. Hypothesis: We hypothesized LA body can be "virtually" dissected from electroanatomic shells using computer graphics imaging (CGI). We further hypothesized that the results of virtual dissection can be used to train a deep learning (DL) model for automatic LA volume and sphericity index calculation. Methods: First, by encoding the anatomic knowledge with geometric modeling, we developed a mathematical method (termed "virtual dissection") based on CGI techniques to isolate the LA body from electroanatomic shells (Fig A). Second, we used virtual dissection to train a DL model to segment CTs in N=20 patient atria (Fig B-C). Finally, we used this pipeline to segment raw CTs in a test cohort of N=100 patients (30.0% women, 66.0±10.3Y; Fig D). Results: In the independent test cohort (N=100), the LA volume and sphericity index were automatically calculated from DL-based segmentation, with the agreement within 4.04 ml (95% CI: -11.48 to 19.55; r=0.98; p < 0.0001; Fig E-F) and 2.75% (95% CI: -5.71 to 11.22; r=0.88; p < 0.0001; Fig G-H), compared to expert annotation. Conclusions: Encoded anatomical knowledge combined with deep learning in small datasets can accurately extract 3D cardiac models from raw CT scans in large populations, yielding automatic calculation of indexes such as volume and sphericity. This computational pipeline may simplify the outcome analysis and prediction using geometric indexes as predictors.

Automatic left atrial segmentation from cardiac CT using computer graphics imaging and deep learning

Abstract Introduction Segmenting left atrial (LA) substructures, including the LA body, appendage (LAA), and pulmonary veins (PVs), from computed tomography (CT) is central to electroanatomic mapping for ablation and functional studies in patients with atrial fibrillation (AF). However, this process requires manual outlining which needs special training, is subjective, and is difficult to scale. Computer graphics imaging (CGI) has been applied in media, film, and computer-aided design to reliably segment complex structures using their basic geometric representations. Purpose We hypothesized that LA substructures can be “virtually” dissected using CGI to separate geometric contours of the “convex ellipsoid” LA, “tubular” PVs, and “conical” LAA. We further hypothesized that the results of virtual dissection can be used to train a deep learning (DL) model to segment raw CT scans. Methods First, a mathematical method based on CGI techniques – erosion and dilation – was developed to “virtually dissect” the convex LA body from the original concave shell in publicly available digital atria with diverse simulated morphologies (Fig. 1A). The PVs and LAA were then automatically revealed and labeled by a 3D subtraction approach. Second, we refined precise LA/PV/LAA boundaries by tuning hyper-parameters from N=5 patient shells (Fig. 1B). Third, we used virtual dissection to train a DL model to segment CTs in N=20 patient atria (Fig. 1C). Finally, we applied this pipeline to segment raw CTs in a validation cohort of N=105 patients (23.8% women, 63.8±10.3Y; Fig. 1D). Results Virtual dissection accurately identified LA/PV/LAA boundaries in the training set (Dice coefficients 89–98%). In the independent test cohort (N=105), this automated pipeline accurately segmented raw CTs with Dice 81–95% (Fig. 1D) compared to a panel of experts (p<0.001). Conclusion CGI of basic cardiac geometry combined with deep learning in small datasets can accurately segment raw CT scans in large populations. This computational pipeline may automate and simplify cardiac image processing and ablation procedures, and could be applied to the ventricle or other organ systems for diverse therapeutic strategies or to train machine learning. Funding Acknowledgement Type of funding sources: Public grant(s) – National budget only. Main funding source(s): National Institutes of Health

DBF-Net: a semi-supervised dual-task balanced fusion network for segmenting infected regions from lung CT images.

Accurate segmentation of infected regions in lung computed tomography (CT) images is essential to improve the timeliness and effectiveness of treatment for coronavirus disease 2019 (COVID-19). However, the main difficulties in developing of lung lesion segmentation in COVID-19 are still the fuzzy boundary of the lung-infected region, the low contrast between the infected region and the normal trend region, and the difficulty in obtaining labeled data. To this end, we propose a novel dual-task consistent network framework that uses multiple inputs to continuously learn and extract lung infection region features, which is used to generate reliable label images (pseudo-labels) and expand the dataset. Specifically, we periodically feed multiple sets of raw and data-enhanced images into two trunk branches of the network; the characteristics of the lung infection region are extracted by a lightweight double convolution (LDC) module and fusiform equilibrium fusion pyramid (FEFP) convolution in the backbone. According to the learned features, the infected regions are segmented, and pseudo-labels are made based on the semi-supervised learning strategy, which effectively alleviates the semi-supervised problem of unlabeled data. Our proposed semi-supervised dual-task balanced fusion network (DBF-Net) creates pseudo-labels on the COVID-SemiSeg dataset and the COVID-19 CT segmentation dataset. Furthermore, we perform lung infection segmentation on the DBF-Net model, with a segmentation sensitivity of 70.6% and specificity of 92.8%. The results of the investigation indicate that the proposed network greatly enhances the segmentation ability of COVID-19 infection.

Body composition from single versus multi-slice abdominal computed tomography: Concordance and associations with colorectal cancer survival.

Computed tomography (CT) scans are routinely obtained in oncology and provide measures of muscle and adipose tissue predictive of morbidity and mortality. Automated segmentation of CT has advanced past single slices to multi-slice measurements, but the concordance of these approaches and their associations with mortality after cancer diagnosis have not been compared. A total of 2871 patients with colorectal cancer diagnosed during 2012-2017 at Kaiser Permanente Northern California underwent abdominal CT scans as part of routine clinical care from which mid-L3 cross-sectional areas and multi-slice T12-L5 volumes of skeletal muscle (SKM), subcutaneous adipose (SAT), visceral adipose (VAT) and intermuscular adipose (IMAT) tissues were assessed using Data Analysis Facilitation Suite, an automated multi-slice segmentation platform. To facilitate comparison between single-slice and multi-slice measurements, sex-specific z-scores were calculated. Pearson correlation coefficients and Bland-Altman analysis were used to quantify agreement. Cox models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for death adjusting for age, sex, race/ethnicity, height, and tumour site and stage. Single-slice area and multi-slice abdominal volumes were highly correlated for all tissues (SKM R=0.92, P<0.001; SAT R=0.97, P<0.001; VAT R=0.98, P<0.001; IMAT R=0.89, P<0.001). Bland-Altman plots had a bias of 0 (SE: 0.00), indicating high average agreement between measures. The limits of agreement were narrowest for VAT ( 0.42 SD) and SAT ( 0.44 SD), and widest for SKM ( 0.78 SD) and IMAT ( 0.92 SD). The HRs had overlapping CIs, and similar magnitudes and direction of effects; for example, a 1-SD increase in SKM area was associated with an 18% decreased risk of death (HR=0.82; 95% CI: 0.72-0.92), versus 15% for volume from T12 to L5 (HR=0.85; 95% CI: 0.75-0.96). Single-slice L3 areas and multi-slice T12-L5 abdominal volumes of SKM, VAT, SAT and IMAT are highly correlated. Associations between area and volume measures with all-cause mortality were similar, suggesting that they are equivalent tools for population studies if body composition is assessed at a single timepoint. Future research should examine longitudinal changes in multi-slice tissues to improve individual risk prediction.

A lightweight neural network with multiscale feature enhancement for liver CT segmentation

Segmentation of abdominal Computed Tomography (CT) scan is essential for analyzing, diagnosing, and treating visceral organ diseases (e.g., hepatocellular carcinoma). This paper proposes a novel neural network (Res-PAC-UNet) that employs a fixed-width residual UNet backbone and Pyramid Atrous Convolutions, providing a low disk utilization method for precise liver CT segmentation. The proposed network is trained on medical segmentation decathlon dataset using a modified surface loss function. Additionally, we evaluate its quantitative and qualitative performance; the Res16-PAC-UNet achieves a Dice coefficient of 0.950 ± 0.019 with less than half a million parameters. Alternatively, the Res32-PAC-UNet obtains a Dice coefficient of 0.958 ± 0.015 with an acceptable parameter count of approximately 1.2 million.

GFNet: Automatic segmentation of COVID-19 lung infection regions using CT images based on boundary features

In early 2020, the global spread of the COVID-19 has presented the world with a serious health crisis. Due to the large number of infected patients, automatic segmentation of lung infections using computed tomography (CT) images has great potential to enhance traditional medical strategies. However, the segmentation of infected regions in CT slices still faces many challenges. Specially, the most core problem is the high variability of infection characteristics and the low contrast between the infected and the normal regions. This problem leads to fuzzy regions in lung CT segmentation. To address this problem, we have designed a novel global feature network(GFNet) for COVID-19 lung infections: VGG16 as backbone, we design a Edge-guidance module(Eg) that fuses the features of each layer. First, features are extracted by reverse attention module and Eg is combined with it. This series of steps enables each layer to fully extract boundary details that are difficult to be noticed by previous models, thus solving the fuzzy problem of infected regions. The multi-layer output features are fused into the final output to finally achieve automatic and accurate segmentation of infected areas. We compared the traditional medical segmentation networks, UNet, UNet++, the latest model Inf-Net, and methods of few shot learning field. Experiments show that our model is superior to the above models in Dice, Sensitivity, Specificity and other evaluation metrics, and our segmentation results are clear and accurate from the visual effect, which proves the effectiveness of GFNet. In addition, we verify the generalization ability of GFNet on another “never seen” dataset, and the results prove that our model still has better generalization ability than the above model. Our code has been shared at https://github.com/zengzhenhuan/GFNet.

Tumorous kidney segmentation in abdominal CT images using active contour and 3D-UNet.

The precise segmentation of the kidneys in computed tomography (CT) images is vital in urology for diagnosis, treatment, and surgical planning. Medical experts can get assistance through segmentation, as it provides information about kidney malformations in terms of shape and size. Manual segmentation is slow, tedious, and not reproducible. An automatic computer-aided system is a solution to this problem. This paper presents an automated kidney segmentation technique based on active contour and deep learning. In this work, 210 CTs from the KiTS 19 repository were used. The used dataset was divided into a train set (168 CTs), test set (21 CTs), and validation set (21 CTs). The suggested technique has broadly four phases: (1) extraction of kidney regions using active contours, (2) preprocessing, (3) kidney segmentation using 3D U-Net, and (4) reconstruction of the segmented CT images. The proposed segmentation method has received the Dice score of97.62%, Jaccard index of 95.74%, average sensitivity of 98.28%, specificity of 99.95%, and accuracy of 99.93% over the validation dataset. The proposed method can efficiently solve the problem of tumorous kidney segmentation in CT images by using active contour and deep learning. The active contour was used to select kidney regions and 3D-UNet was used for precisely segmenting the tumorous kidney.

Development of 3-stage hybrid computer aided design (3-HCAD) system for multi-modal medical images to identify brain tumor

The latest developments in medical imaging and computer-aided solutions for image processing problems attract attention of various researchers to impart their research in the medical imaging field. Designing and developing efficient algorithms to present the medical information effectively have become critical areas of research in this field. A 3-Stage Hybrid Computer Aided Design system is introduced to identify Brain tumor in earlier stages by extracting meaningful information from multimodal medical images. The preferred multi-modality images is Magnetic Resonance Imaging (MRI) and Computed Tomography (CT). The CAD system proposed in this paper can eliminate the dependency on human operators as it is an efficient software-based system. The first stage in this model consists of image pre-processing with Wavelet and Curvelet transforms. The second stage of the CAD system is an image segmentation process, which involves a combination of Wavelet Transform and Watershed Technique. The third stage involves image fusion, where the individually segmented CT and MRI images are fused together to obtain an integrated complementary information from two different images. This is followed by decomposing CT and MRI images using the Dual Tree Complex Wavelet Transform (DTCWT) and Nonsubsampled Contourlet Transform (NSCT).

Reliability and accuracy of straightforward measurements for liver volume determination in ultrasound and computed tomography compared to real volumetry

To evaluate the suitability of volume index measurement (VI) by either ultrasound (US) or computed tomography (CT) for the assessment of liver volume. Fifty-nine patients, 21 women, with a mean age of 66.8 ± 12.6 years underwent US of the liver followed immediately by abdominal CT. In US and CT imaging dorsoventral, mediolateral and craniocaudal liver diameters in their maximum extensions were assessed by two observers. VI was calculated by multiplication of the diameters divided by a constant (3.6). The liver volume determined by a manual segmentation in CT (“true liver volume”) served as gold standard. True liver volume and calculated VI determined by US and CT were compared using Bland–Altman analysis. Mean differences of VI between observers were − 34.7% (− 90.1%; 20.7%) for the US-based and 1.1% (− 16.1%; 18.2%) for the CT-based technique, respectively. Liver volumes determined by semi-automated segmentation, US-based VI and CT-based VI, were as follows: 1.500 ± 347cm3; 863 ± 371cm3; 1.509 ± 432cm3. Results showed a great discrepancy between US-based VI and true liver volume with a mean bias of 58.3 ± 66.9%, and high agreement between CT-based VI and true liver volume with a low mean difference of 4.4 ± 28.3%. Volume index based on CT diameters is a reliable, fast and simple approach for estimating liver volume and can therefore be recommended for clinical practice. The usage of US-based volume index for assessment of liver volume should not be used due to its low accuracy of US in measurement of liver diameters.

Psoas Cross-Sectional Measurements Using Manual CT Segmentation before and after Endovascular Aortic Repair (EVAR).

Sarcopenia has been associated with an increased incidence of adverse outcomes, including higher mortality, after endovascular aortic repair (EVAR). We aim to use computed tomography (CT) to quantify changes in total psoas muscles area (PMA) and psoas muscle density (PMD) after EVAR, and to evaluate the reproducibility of both measurements. PMA and PMD were assessed via manual segmentation of the psoas muscle on pre- and post-operative CT scans belonging to consecutive patients who underwent EVAR. Wilcoxon test was used to compare PMA and PMD before and after EVAR, and inter- and intra-reader agreements of both methods were evaluated through Bland–Altman analysis. A total of 50 patients, 42 of them males (84%), were included in the study. PMA changes from 1243 mm2 (1006–1445 mm2) to 1102 mm2 (IQR 937–1331 mm2), after EVAR (p < 0.001). PMD did not vary between pre-EVAR (33 HU, IQR 26.5–38.7 HU) and post-EVAR (32 HU, IQR 26–37 HU, p = 0.630). At inter-reader Bland–Altman analysis, PMA showed a bias of 64.0 mm2 and a coefficient of repeatability (CoR) of 359.2 mm2, whereas PMD showed a bias of −2.43 HU and a CoR of 6.19 HU. At intra-reader Bland–Altman analysis, PMA showed a bias of −81.1 mm2 and a CoR of 394.6 mm2, whereas PMD showed a bias of 1.41 HU and a CoR of 6.36 HU. In conclusion, PMA decreases after EVAR. A good intra and inter-reader reproducibility was observed for both PMA and PMD. We thus propose to use PMA during the follow-up of patients who underwent EVAR to monitor muscle depletion after surgery.

An unsupervised domain adaptation brain CT segmentation method across image modalities and diseases

Computed tomography (CT) is the primary diagnostic tool for brain diseases. To determine the appropriate treatment plan, it is necessary to ascertain the patient’s bleeding volume. Automatic segmentation algorithms for hemorrhagic lesions can significantly improve efficiency and avoid treatment delays. However, for deep supervised learning algorithms, a large amount of labeled training data is usually required, making them difficult to apply clinically. In this study, we propose an unsupervised domain adaptation method that is an unsupervised domain adaptation segmentation model that can be trained across modalities and diseases. We call it AMD-DAS for brain CT hemorrhage segmentation tasks. This circumvents the heavy data labeling task by converting the source domain data (MRI with glioma) to our task’s required data (CT with Intraparenchymal hemorrhage (IPH)). Our model implements a two-stage domain adaptation process to achieve this objective. In the first stage, we train a pseudo-CT image synthesis network using the CycleGAN architecture through a matching mechanism and domain adaptation approach. In the second stage, we use the model trained in the first stage to synthesize the pseudo-CT images. We use the pseudo-CT with source domain labels and real CT images to train a domain-adaptation segmentation model. Our method exhibits a better performance than the basic one-stage domain adaptation segmentation method (+11.55 Dice score) and achieves an 86.93 Dice score in the IPH unsupervised segmentation task. Our model can be trained without using a ground-truth label, therefore increasing its application potential. Our implementation is publicly available at https://github.com/GuanghuiFU/AMD-DAS-Brain-CT-Segmentation.