BranchLabelNet: Anatomical Human Airway Labeling Approach using a Dividing-and-Grouping Multi-Label Classification.

Anatomical airway labeling is crucial for precisely identifying airways displaying symptoms such as constriction, increased wall thickness, and modified branching patterns, facilitating the diagnosis and treatment of pulmonary ailments. This study introduces an innovative airway labeling methodology, BranchLabelNet, which accounts for the fractal nature of airways and inherent hierarchical branch nomenclature. In developing this methodology, branch-related parameters, including position vectors, generation levels, branch lengths, areas, perimeters, and more, are extracted from a dataset of 1000 chest computed tomography (CT) images. To effectively manage this intricate branch data, we employ an n-ary tree structure that captures the complicated relationships within the airway tree. Subsequently, we employ a divide-and-group deep learning approach for multi-label classification, streamlining the anatomical airway branch labeling process. Additionally, we address the challenge of class imbalance in the dataset by incorporating the Tomek Links algorithm to maintain model reliability and accuracy. Our proposed airway labeling method provides robust branch designations and achieves an impressive average classification accuracy of 95.94% across fivefold cross-validation. This approach is adaptable for addressing similar complexities in general multi-label classification problems within biomedical systems.

ID1high/activin Ahigh glioblastoma cells contribute to resistance to anti-angiogenesis therapy through malformed vasculature.

Although bevacizumab (BVZ), a representative drug for anti-angiogenesis therapy (AAT), is used as a first-line treatment for patients with glioblastoma (GBM), its efficacy is notably limited. Whereas several mechanisms have been proposed to explain the acquisition of AAT resistance, the specific underlying mechanisms have yet to be sufficiently ascertained. Here, we established that inhibitor of differentiation 1 (ID1)high/activin Ahigh glioblastoma cell confers resistance to BVZ. The bipotent effect of activin A during its active phase was demonstrated to reduce vasculature dependence in tumorigenesis. In response to a temporary exposure to activin A, this cytokine was found to induce endothelial-to-mesenchymal transition via the Smad3/Slug axis, whereas prolonged exposure led to endothelial apoptosis. ID1 tumors showing resistance to BVZ were established to be characterized by a hypovascular structure, hyperpermeability, and scattered hypoxic regions. Using a GBM mouse model, we demonstrated that AAT resistance can be overcome by administering therapy based on a combination of BVZ and SB431542, a Smad2/3 inhibitor, which contributed to enhancing survival. These findings offer valuable insights that could contribute to the development of new strategies for treating AAT-resistant GBM.

Relationship of computed tomography-based measurements with symptom perception and quality of life in patients with severe asthma.

BACKGROUND: Symptom perception and quality of life (QOL) are important domains for properly managing severe asthma. This study aimed to assess the relationship between airway structural and parenchymal variables measured using chest computed tomography (CT) and subjective symptom perception and QOL in patients with severe asthma enrolled in the Korean Severe Asthma Registry. METHODS: This study used CT-based objective measurements, including airway wall thickness (WT), hydraulic diameter, functional small airway disease (fSAD), and emphysematous lung (Emph), to assess their association with subjective symptom (cough, dyspnea, wheezing, and sputum) perception measured using the visual analog scale, and QOL measured by the Severe Asthma Questionnaire (SAQ). RESULTS: A total of 94 patients with severe asthma were enrolled in this study. The WT and fSAD% were significantly positively associated with cough and dyspnea, respectively. For QOL, WT and Emph% showed significant negative associations with the SAQ. However, there was no significant association between lung function and symptom perception or between lung function and QOL. CONCLUSION: Overall, WT, fSAD%, and Emph% measured using chest CT were associated with subjective symptom perception and QOL in patients with severe asthma. This study provides a basis for clarifying the clinical correlates of imaging-derived metrics and for understanding the mechanisms of respiratory symptom perception.

Human-airway surface mesh smoothing based on graph convolutional neural networks.

BACKGROUND AND OBJECTIVE: A detailed representation of the airway geometry in the respiratory system is critical for predicting precise airflow and pressure behaviors in computed tomography (CT)-image-based computational fluid dynamics (CFD). The CT-image-based geometry often contains artifacts, noise, and discontinuities due to the so-called stair step effect. Hence, an advanced surface smoothing is necessary. The existing smoothing methods based on the Laplacian operator drastically shrink airway geometries, resulting in the loss of information related to smaller branches. This study aims to introduce an unsupervised airway-mesh-smoothing learning (AMSL) method that preserves the original geometry of the three-dimensional (3D) airway for accurate CT-image-based CFD simulations. METHOD: The AMSL method jointly trains two graph convolutional neural networks (GCNNs) defined on airway meshes to filter vertex positions and face normal vectors. In addition, it regularizes a combination of loss functions such as reproducibility, smoothness and consistency of vertex positions, and normal vectors. The AMSL adopts the concept of a deep mesh prior model, and it determines the self-similarity for mesh restoration without using a large dataset for training. Images of the airways of 20 subjects were smoothed by the AMSL method, and among them, the data of two subjects were used for the CFD simulations to assess the effect of airway smoothing on flow properties. RESULTS: In 18 of 20 benchmark problems, the proposed smoothing method delivered better results compared with the conventional or state-of-the-art deep learning methods. Unlike the traditional smoothing, the AMSL successfully constructed 20 smoothed airways with airway diameters that were consistent with the original CT images. Besides, CFD simulations with the airways obtained by the AMSL method showed much smaller pressure drop and wall shear stress than the results obtained by the traditional method. CONCLUSIONS: The airway model constructed by the AMSL method reproduces branch diameters accurately without any shrinkage, especially in the case of smaller airways. The accurate estimation of airway geometry using a smoothing method is critical for estimating flow properties in CFD simulations.

Structural and functional features of asthma participants with fixed airway obstruction using CT imaging and 1D computational fluid dynamics: A feasibility study.

Asthma with fixed airway obstruction (FAO) is associated with significant morbidity and rapid decline in lung function, making its treatment challenging. Quantitative computed tomography (QCT) along with data postprocessing is a useful tool to obtain detailed information on airway structure, parenchymal function, and computational flow features. In this study, we aim to identify the structural and functional differences between asthma with and without FAO. The FAO group was defined by a ratio of forced expiratory volume in 1 s (FEV1 ) to forced vital capacity (FVC), FEV1 /FVC <0.7. Accordingly, we obtained two sets of QCT images at inspiration and expiration of asthma subjects without (N = 24) and with FAO (N = 12). Structural and functional QCT-derived airway variables were extracted, including normalized hydraulic diameter, normalized airway wall thickness, functional small airway disease, and emphysema percentage. A one-dimensional (1D) computational fluid dynamics (CFD) model considering airway deformation was used to compare the pressure distribution between the two groups. The computational pressures showed strong correlations with the pulmonary function test (PFT)-based metrics. In conclusion, asthma participants with FAO had worse lung functions and higher-pressure drops than those without FAO.

Migrasomal autophagosomes relieve endoplasmic reticulum stress in glioblastoma cells.

BACKGROUND: Glioblastoma (GBM) is more difficult to treat than other intractable adult tumors. The main reason that GBM is so difficult to treat is that it is highly infiltrative. Migrasomes are newly discovered membrane structures observed in migrating cells. Thus, they can be generated from GBM cells that have the ability to migrate along the brain parenchyma. However, the function of migrasomes has not yet been elucidated in GBM cells. RESULTS: Here, we describe the composition and function of migrasomes generated along with GBM cell migration. Proteomic analysis revealed that LC3B-positive autophagosomes were abundant in the migrasomes of GBM cells. An increased number of migrasomes was observed following treatment with chloroquine (CQ) or inhibition of the expression of STX17 and SNAP29, which are involved in autophagosome/lysosome fusion. Furthermore, depletion of ITGA5 or TSPAN4 did not relieve endoplasmic reticulum (ER) stress in cells, resulting in cell death. CONCLUSIONS: Taken together, our study suggests that increasing the number of autophagosomes, through inhibition of autophagosome/lysosome fusion, generates migrasomes that have the capacity to alleviate cellular stress.

Computed tomography-based imaging biomarker identifies coal workers' pneumoconiosis.

Rationale: The increase in the incidence and the diagnostic limitations of pneumoconiosis have emerged as a public health concern. This study aimed to conduct a computed tomography (CT)- based quantitative analysis to understand differences in imaging results of pneumoconiosis according to disease severity. Methods: According to the International Labor Organization (ILO) guidelines, coal workers' pneumoconiosis (CWP) are classified into five categories. CT images were obtained only at full inspiration and were quantitatively evaluated for airway structural variables such as bifurcation angle (θ), hydraulic diameter (Dh), wall thickness (WT), and circularity (Cr). Parenchymal functional variables include abnormal regions (emphysema, ground-glass opacities, consolidation, semi consolidation, and fibrosis) and blood vessel volume. Through the propensity score matching method, the confounding effects were decreased. Results: Category 4 demonstrated a reduced θ in TriLUL, a thicker airway wall in both the Trachea and Bronint compared to Category 0, and a decreased Cr in Bronint. Category 4 presented with higher abnormal regions except for ground-glass opacity and a narrower pulmonary blood vessel volume. A negative correlation was found between abnormal areas with lower Hounsfield units (HU) than the normal lung and the ratio of forced expiratory volume in 1 s/forced vital capacity, with narrowed pulmonary blood vessel volume which is positively correlated with abnormal areas with upper HU than the normal lung. Conclusion: This study provided valuable insight into pneumoconiosis progression through a comparison of quantitative CT images based on severity. Furthermore, as there has been paucity of studies on the pulmonary blood vessel volume of the CWP, in this study, a correlation between reduced pulmonary blood vessel volume and regions with low HU values holds significant importance.

Aloe-derived nanovesicles attenuate inflammation and enhance tight junction proteins for acute colitis treatment.

Inflammatory bowel disease (IBD) is a chronic recurrent inflammatory disease of the digestive tract that causes pain and weight loss and also increases the risk of colon cancer. Inspired by the benefits of plant-derived nanovesicles and aloe, we herein report aloe-derived nanovesicles, including aloe vera-derived nanovesicles (VNVs), aloe arborescens-derived nanovesicles (ANVs), and aloe saponaria-derived nanovesicles (SNVs) and evaluate their therapeutic potential and molecular mechanisms in a dextran sulfate sodium (DSS)-induced acute experimental colitis mouse model. Aloe-derived nanovesicles not only facilitate markedly reduced DSS-induced acute colonic inflammation, but also enable the restoration of tight junction (TJ) and adherent junction (AJ) proteins to prevent gut permeability in DSS-induced acute colonic injury. These therapeutic effects are ascribed to the anti-inflammatory and anti-oxidant effects of aloe-derived nanovesicles. Therefore, aloe-derived nanovesicles are a safe treatment option for IBD.

An unsupervised image registration method employing chest computed tomography images and deep neural networks.

BACKGROUND: Deformable image registration is crucial for multiple radiation therapy applications. Fast registration of computed tomography (CT) lung images is challenging because of the large and nonlinear deformation between inspiration and expiration. With advancements in deep learning techniques, learning-based registration methods are considered efficient alternatives to traditional methods in terms of accuracy and computational cost. METHOD: In this study, an unsupervised lung registration network (LRN) with cycle-consistent training is proposed to align two acquired CT-derived lung datasets during breath-holds at inspiratory and expiratory levels without utilizing any ground-truth registration results. Generally, the LRN model uses three loss functions: image similarity, regularization, and Jacobian determinant. Here, LRN was trained on the CT datasets of 705 subjects and tested using 10 pairs of public CT DIR-Lab datasets. Furthermore, to evaluate the effectiveness of the registration technique, target registration errors (TREs) of the LRN model were compared with those of the conventional algorithm (sum of squared tissue volume difference; SSTVD) and a state-of-the-art unsupervised registration method (VoxelMorph). RESULTS: The results showed that the LRN with an average TRE of 1.78 ± 1.56 mm outperformed VoxelMorph with an average TRE of 2.43 ± 2.43 mm, which is comparable to that of SSTVD with an average TRE of 1.66 ± 1.49 mm. In addition, estimating the displacement vector field without any folding voxel consumed less than 2 s, demonstrating the superiority of the learning-based method with respect to fiducial marker tracking and the overall soft tissue alignment with a nearly real-time speed. CONCLUSIONS: Therefore, this proposed method shows significant potential for use in time-sensitive pulmonary studies, such as lung motion tracking and image-guided surgery.

The oncogenic JAG1 intracellular domain is a transcriptional cofactor that acts in concert with DDX17/SMAD3/TGIF2.

Jagged1 (JAG1) is a Notch ligand that contact-dependently activates Notch receptors and regulates cancer progression. The JAG1 intracellular domain (JICD1) is generated from JAG1, like formation of the NOTCH1 intracellular domain (NICD1); however, the role of JICD1 in tumorigenicity has not been comprehensively elucidated. Here we show that JICD1 induces astrocytes to acquire several cancer stem cell properties, including tumor formation, invasiveness, stemness, and resistance to anticancer therapy. The transcriptome, chromatin immunoprecipitation sequencing (ChIP-seq), and proteomics analyses show that JICD1 increases SOX2 expression by forming a transcriptional complex with DDX17, SMAD3, and TGIF2. JICD1-driven tumorigenicity is directly regulated by SOX2. Our results demonstrate that, like NICD1, JICD1 acts as a transcriptional cofactor in formation of the DDX17/SMAD3/TGIF2 transcriptional complex, leading to oncogenic transformation.

Inhibitor of differentiation 1 in U87MG glioblastoma cells promotes HUVEC sprouting through endothelin-1.

Anti-angiogenesis therapy, a promising remedy against tumor progression, is now widely used to treat numerous types of cancer. Since vascular endothelial growth factor (VEGF) is the most vital factor in angiogenesis, most anti-angiogenesis drugs target the VEGF-related pathway. However, in glioblastoma (GBM), the therapeutic strategy involving the inhibition of VEGF signaling is ineffective. The present study demonstrated that the potential angiogenic function of endothelin-1 (EDN1) was upregulated by inhibitor of differentiation 1 (ID1) independent of VEGF during tumor angiogenesis. Anatomic structure transcriptomes of patients with GBM revealed that the expression levels of ID1 and EDN1 were specifically upregulated in the vascular-related region. The aortic ring assay and endothelial sprouting assay demonstrated that EDN1 more potently promoted endothelial sprouting ability than VEGF. The activity of EDN1 was induced by endothelin receptor, which seemed to mediate regulation via positive feedback. Finally, in patients with GBM who did not respond to bevacizumab, a VEGF antagonist, EDN1 expression was higher than that in bevacizumab responders. Collectively, the present study demonstrated that EDN1 is a potent angiogenic factor inducing endothelial sprouting and may be a novel target for inhibiting glioma angiogenesis.

Hemp-Derived Nanovesicles Protect Leaky Gut and Liver Injury in Dextran Sodium Sulfate-Induced Colitis.

Hemp (Cannabis sativa L.) is used for medicinal purposes owing to its anti-inflammatory and antioxidant activities. We evaluated the protective effect of nanovesicles isolated from hemp plant parts (root, seed, hemp sprout, and leaf) in dextran sulfate sodium (DSS)-induced colitis in mice. The particle sizes of root-derived nanovesicles (RNVs), seed-derived nanovesicles (SNVs), hemp sprout-derived nanovesicles (HSNVs), and leaf-derived nanovesicles (LNVs) were within the range of 100-200 nm as measured by nanoparticle tracking analysis. Acute colitis was induced in C57BL/N mice by 5% DSS in water provided for 7 days. RNVs were administered orally once a day, leading to the recovery of both the small intestine and colon lengths. RNVs, SNVs, and HSNVs restored the tight (ZO-1, claudin-4, occludin) and adherent junctions (E-cadherin and α-tubulin) in DSS-induced small intestine and colon injury. Additionally, RNVs markedly reduced NF-κB activation and oxidative stress proteins in DSS-induced small intestine and colon injury. Tight junction protein expression and epithelial cell permeability were elevated in RNV-, SNV-, and HSNV-treated T84 colon cells exposed to 2% DSS. Interestedly, RNVs, SNVs, HSNVs, and LNVs reduced ALT activity and liver regeneration marker proteins in DSS-induced liver injury. These results showed for the first time that hemp-derived nanovesicles (HNVs) exhibited a protective effect on DSS-induced gut leaky and liver injury through the gut-liver axis by inhibiting oxidative stress marker proteins.

Phenotypic clusters on computed tomography reflects asthma heterogeneity and severity.

Background: Asthma is a heterogeneous inflammatory airway disorder with various phenotypes. Quantitative computed tomography (QCT) methods can differentiate among lung diseases through accurate assessment of the location, extent, and severity of the disease. The purpose of this study was to identify asthma clusters using QCT metrics of airway and parenchymal structure, and to identify associations with visual analyses conducted by radiologists. Methods: This prospective study used input from QCT-based metrics including hydraulic diameter (D h), luminal wall thickness (WT), functional small airway disease (fSAD), and emphysematous lung (Emph) to perform a cluster analysis and made comparisons with the visual grouping analysis conducted by radiologists based on site of airway involvement and remodeling evaluated. Results: A total of 61 asthmatics of varying severities were grouped into 4 clusters. From C1 to C4, more severe lung function deterioration, higher fixed obstruction rate, and more frequent asthma exacerbations were observed in the 5-year follow-up period. C1 presented non-severe asthma with increased WT, D h of proximal airways, and fSAD. C2 was mixed with non-severe and severe asthmatics, and showed bronchodilator responses limited to the proximal airways. C3 and C4 included severe asthmatics that showed a reduced D h of the proximal airway and diminished bronchodilator responses. While C3 was characterized by severe allergic asthma without fSAD, C4 included ex-smokers with high fSAD% and Emph%. These clusters correlated well with the grouping done by radiologists and clinical outcomes. Conclusions: Four QCT imaging-based clusters with distinct structural and functional changes in proximal and small airways can stratify heterogeneous asthmatics and can be a complementary tool to predict clinical outcomes.

Cytoplasmic LMO2-LDB1 Complex Activates STAT3 Signaling through Interaction with gp130-JAK in Glioma Stem Cells.

The oncogenic role of nuclear LIM domain only 2 (LMO2) as a transcriptional regulator is well established, but its function in the cytoplasm is largely unknown. Here, we identified LMO2 as a cytoplasmic activator for signal transducer and activator of transcription 3 (STAT3) signaling in glioma stem cells (GSCs) through biochemical and bioinformatics analyses. LMO2 increases STAT3 phosphorylation by interacting with glycoprotein 130 (gp130) and Janus kinases (JAKs). LMO2-driven activation of STAT3 signaling requires the LDB1 protein and leads to increased expression of an inhibitor of differentiation 1 (ID1), a master regulator of cancer stemness. Our findings indicate that the cytoplasmic LMO2-LDB1 complex plays a crucial role in the activation of the GSC signaling cascade via interaction with gp130 and JAK1/2. Thus, LMO2-LDB1 is a bona fide oncogenic protein complex that activates either the JAK-STAT signaling cascade in the cytoplasm or direct transcriptional regulation in the nucleus.

Structural and functional alterations of subjects with cement dust exposure: A longitudinal quantitative computed tomography-based study.

Cement dust exposure (CDE) can be a risk factor for pulmonary disease, causing changes in segmental airways and parenchymal lungs. This study investigates longitudinal alterations in quantitative computed tomography (CT)-based metrics due to CDE. We obtained CT-based airway structural and lung functional metrics from CDE subjects with baseline CT and follow-up CT scans performed three years later. From the CT, we extracted wall thickness (WT) and bifurcation angle (θ) at total lung capacity (TLC) and functional residual capacity (FRC), respectively. We also computed air volume (Vair), tissue volume (Vtissue), global lung shape, percentage of emphysema (Emph%), and more. Clinical measures were used to associate with CT-based metrics. Three years after their baseline, the pulmonary function tests of CDE subjects were similar or improved, but there were significant alterations in the CT-based structural and functional metrics. The follow-up CT scans showed changes in θ at most of the central airways; increased WT at the subgroup bronchi; smaller Vair at TLC at all except the right upper and lower lobes; smaller Vtissue at all lobes in TLC and FRC except for the upper lobes in FRC; smaller global lung shape; and greater Emph% at the right upper and lower lobes. CT-based structural and functional variables are more sensitive to the early identification of CDE subjects, while most clinical lung function changes were not noticeable. We speculate that the significant long-term changes in CT are uniquely observed in CDE subjects, different from smoking-induced structural changes.

Plum Prevents Intestinal and Hepatic Inflammation in the Acute and Chronic Models of Dextran Sulfate Sodium-Induced Mouse Colitis.

SCOPE: Inflammatory bowel disease (IBD), including ulcerative colitis (UC), is a chronic recurrent inflammatory disease of the digestive tract and increases the risk of colon cancer. METHOD AND RESULTS: This study evaluates the effects of dietary intervention with freeze-dried plum (FDP), a natural antioxidant and anti-inflammatory fruit with no toxicity on dextran sulfate sodium (DSS)-induced acute and chronic experimental colitis in a mouse model and studies the molecular mechanisms of protection through the gut-liver axis. The results show that FDP decreases the levels of inflammatory mediators, which is a nitrative stress biomarker in both acute and chronic models. FDP markedly reduces DSS-induced injury to the colonic epithelium in both acute and chronic models. In addition, FDP significantly decreases the levels of pro-oxidant markers such as CYP2E1, iNOS, and nitrated proteins (detected by anti-3-NT antibody) in DSS-induced acute and chronic colonic injury models. Furthermore, FDP markedly reduces markers of liver injury such as serum ALT/AST, antioxidant markers, and inflammatory mediators in DSS-induced acute and chronic colonic injury. CONCLUSION: These results demonstrate that the FDP exhibits a protective effect on DSS-induced acute and chronic colonic and liver injury through the gut-liver axis via antioxidant and anti-inflammatory properties.

Effects of face shield on an emitter during a cough process: A large-eddy simulation study.

Face shield is a common personal protection equipment for pandemic. In the present work, three-dimensional computational fluid dynamic (CFD) method is used to simulate a cough jet from an emitter who wears a face shield. A realistic manikin model with a simplified mouth cavity is employed. A large eddy simulation with a dynamic structure subgrid scale model is applied to model the turbulence. An Eulerian-Lagrangian approach is adopted to model the two-phase flows, with which the droplets are represented by a cloud of particles. The droplet breakup, evaporation, dispersion, drag force, and wall impingement are considered in this model. An inlet velocity profile that is based on a variable mouth opening area is considered. Special attentions have been put the vortex structure and droplet re-distribution induced by the face shield. It is found that the multiple vortices are formed when the cough jet impinges on the face shield. Some droplets move backward and others move downward after the impinging. It is also found that a small modification of the face shield significantly modifies the flow field and droplet distribution. We conclude that face shield significantly reduces the risk factor in the front of the emitter, meanwhile the risk factor in the back of the emitter increases. When the receiver standing in front of the emitter is shorter than the emitter, the risk is still very high. More attentions should be paid on the design of the face field, clothes cleaning and floor cleaning of the emitters with face shields. Based on the predicted droplet trajectory, a conceptual model for droplet flux is proposed for the scenario with the face shield.

Classification of rotator cuff tears in ultrasound images using deep learning models.

Rotator cuff tears (RCTs) are one of the most common shoulder injuries, which are typically diagnosed using relatively expensive and time-consuming diagnostic imaging tests such as magnetic resonance imaging or computed tomography. Deep learning algorithms are increasingly used to analyze medical images, but they have not been used to identify RCTs with ultrasound images. The aim of this study is to develop an approach to automatically classify RCTs and provide visualization of tear location using ultrasound images and convolutional neural networks (CNNs). The proposed method was developed using transfer learning and fine-tuning with five pre-trained deep models (VGG19, InceptionV3, Xception, ResNet50, and DenseNet121). The Bayesian optimization method was also used to optimize hyperparameters of the CNN models. A total of 194 ultrasound images from Kosin University Gospel Hospital were used to train and test the CNN models by five-fold cross-validation. Among the five models, DenseNet121 demonstrated the best classification performance with 88.2% accuracy, 93.8% sensitivity, 83.6% specificity, and AUC score of 0.832. A gradient-weighted class activation mapping (Grad-CAM) highlighted the sensitive features in the learning process on ultrasound images. The proposed approach demonstrates the feasibility of using deep learning and ultrasound images to assist RCTs' diagnosis.

Quantitative computed tomography imaging-based classification of cement dust-exposed subjects with an artificial neural network technique.

BACKGROUND AND OBJECTIVE: Cement dust exposure is likely to affect the structural and functional alterations in segmental airways and parenchymal lungs. This study develops an artificial neural network (ANN) model for identifying cement dust-exposed (CDE) subjects using quantitative computed tomography-based airway structural and functional features. METHODS: We obtained the airway features in five central and five sub-grouped segmental regions and the lung features in five lobar regions and one total lung region from 311 CDE and 298 non-CDE (NCDE) subjects. The five-fold cross-validation method was adopted to train the following classification models:ANN, support vector machine (SVM), logistic regression (LR), and decision tree (DT). For all the classification models, linear discriminant analysis (LDA) and genetic algorithm (GA) were applied for dimensional reduction and hyperparameterization, respectively. The ANN model without LDA was also optimized by the GA method to observe the effect of the dimensional reduction. RESULTS: The genetically optimized ANN model without the LDA method was the best in terms of the classification accuracy. The accuracy, sensitivity, and specificity of the GA-ANN model with four layers were greater than those of the other classification models (i.e., ANN, SVM, LR, and DT using LDA and GA methods) in the five-fold cross-validation. The average values of accuracy, sensitivity, and specificity for the five-fold cross-validation were 97.0%, 98.7%, and 98.6%, respectively. CONCLUSIONS: We demonstrated herein that a quantitative computed tomography-based ANN model could more effectively detect CDE subjects when compared to their counterpart models. By employing the model, the CDE subjects may be identified early for therapeutic intervention.

Deep Learning Techniques for Fatty Liver Using Multi-View Ultrasound Images Scanned by Different Scanners: Development and Validation Study.

BACKGROUND: Fat fraction values obtained from magnetic resonance imaging (MRI) can be used to obtain an accurate diagnosis of fatty liver diseases. However, MRI is expensive and cannot be performed for everyone. OBJECTIVE: In this study, we aim to develop multi-view ultrasound image-based convolutional deep learning models to detect fatty liver disease and yield fat fraction values. METHODS: We extracted 90 ultrasound images of the right intercostal view and 90 ultrasound images of the right intercostal view containing the right renal cortex from 39 cases of fatty liver (MRI-proton density fat fraction [MRI-PDFF] ≥ 5%) and 51 normal subjects (MRI-PDFF < 5%), with MRI-PDFF values obtained from Good Gang-An Hospital. We obtained combined liver and kidney-liver (CLKL) images to train the deep learning models and developed classification and regression models based on the VGG19 model to classify fatty liver disease and yield fat fraction values. We employed the data augmentation techniques such as flip and rotation to prevent the deep learning model from overfitting. We determined the deep learning model with performance metrics such as accuracy, sensitivity, specificity, and coefficient of determination (R2). RESULTS: In demographic information, all metrics such as age and sex were similar between the two groups-fatty liver disease and normal subjects. In classification, the model trained on CLKL images achieved 80.1% accuracy, 86.2% precision, and 80.5% specificity to detect fatty liver disease. In regression, the predicted fat fraction values of the regression model trained on CLKL images correlated with MRI-PDFF values (R2=0.633), indicating that the predicted fat fraction values were moderately estimated. CONCLUSIONS: With deep learning techniques and multi-view ultrasound images, it is potentially possible to replace MRI-PDFF values with deep learning predictions for detecting fatty liver disease and estimating fat fraction values.