ABSTRACT
Aromatic components of C8-C15 are playing indispensable roles in multi-functional properties of jet fuel. Here, we reported the controllable alkylation of benzene with mixed olefins of ethylene and propylene toward C8-C15 aromatic hydrocarbons for jet fuels over the bifunctional Ga/ZSM-5 catalyst. The resultant 2Ga/ZSM-5 exhibited a superior selectivity of 86.4% (yield of 55.5%) to C8-C15 range aromatics, at benzene conversion of 40.3%, ethylene and propylene conversion of 99.5% and 99.2%, respectively. The incorporation of Ga species could effectively weaken the strong acid sites of ZSM-5 and endow 2Ga/ZSM-5 catalyst with appropriate acidity, therefore facilitating the benzene alkylation process and suppressing the undesired hydrogen transfer or aromatization side reactions as well, thus improving the yield of desired C8-C15aromatics for jet fuels. This work provided insight into the development of promising bifunctional catalyst for the oriented transformation of biomass-derived chemicals to aviation fuels.
ABSTRACT
Iron-based Fischer-Tropsch synthesis (FTS) catalyst is widely used for syngas conversion, but its iron carbide active phase is easily oxidized into Fe3O4 by the water produced during reaction, leading to the deterioration of catalytic performance. Here, we show an efficient strategy for protecting the iron carbide active phase of FTS catalyst by surface hydrophobization. The hydrophobic surface can reduce the water concentration in the core vicinity of catalyst during syngas conversion, and thus inhibit the oxidation of iron species by water, which enhances the C - C coupling ability of catalyst and promotes the formation of long-chain olefins. More significantly, it is unraveled that appropriate shell thickness plays a crucial role in stabilizing the iron carbide active phase without Fe3O4 formation and achieving good catalytic performance.
ABSTRACT
1-Aminohydantoin (AHD), the residual marker of nitrofurantoin, is usually detected after derivatisation using the derivatisation reagent 2-nitrobenzaldehyde. Avoiding the antibody recognition of the derivatisation reagent is essential for the accurate detection of AHD residues. In this paper, a novel hapten called hapten D was designed, and then, a monoclonal antibody that did not recognise 2-nitrobenzaldehyde was prepared based on this novel hapten. An ultra-sensitive indirect competitive enzyme linked-immunosorbent assay (icELISA) was established under optimal conditions. The 50% inhibition concentration and limit of detection of AHD were 0.056 and 0.0060 ng/mL, respectively, which improved the sensitivity by 9-37-fold compared with the previously reported icELISA methods. The average recovery rates were 88.1%-97.3%, and the coefficient of variation was <8.6%. The accuracy and reliability of the icELISA were verified using liquid chromatography-tandem mass spectrometry. These results demonstrated that the developed icELISA is a useful and reliable tool.
Subject(s)
Antibodies, Monoclonal , Enzyme-Linked Immunosorbent Assay , Hydantoins , Nitrofurantoin , Antibodies, Monoclonal/chemistry , Enzyme-Linked Immunosorbent Assay/methods , Nitrofurantoin/chemistry , Nitrofurantoin/analysis , Hydantoins/chemistry , Hydantoins/analysis , Animals , Limit of Detection , Food Contamination/analysis , Mice , Haptens/chemistry , Haptens/immunology , Female , Mice, Inbred BALB CABSTRACT
Many studies have been carried out on ultrasound computed tomography (USCT) for its ability to offer quantitative measurements of tissue sound speed. Full waveform inversion (FWI) is a technique for reconstructing high-resolution sound speed images by iteratively minimizing the difference between the observed ultrasound data and the synthetic data based on the waveform equation. However, FWI suffers from cycle-skipping, which usually causes FWI convergence at a local minimum. Cycle-skipping occurs when the phase difference between the observed data and the synthetic data exceeds half a cycle. The simplest way to avoid cycle-skipping is to use low-frequency information for reconstruction. Nevertheless, in imaging systems, the response bandwidth of the probe is limited, and reliable low-frequency information often exceeds the response band. Therefore, it is a challenge to perform FWI imaging and avoid cycle-skipping problems without low-frequency information. In this paper, we propose a frequency shift envelope-based global correlation norm (FSEGCN), where an artificial source wavelet with a lower frequency is adopted to calculate synthetic data. FSEGCN compared with FWI, envelope inversion (EI), global correlation norm (GCN), envelope-based global correlation norm (EGCN) through concentric circle phantom without low-frequency information. The experimental results demonstrated the capability of the proposed method to recover the sound speed close to the exact model in the absence of low-frequency information, whereas FWI, EI, GCN, and EGCN cannot. Experiments on phantoms of the human head and calf show that artificial source wavelets can reduce image artifacts and enhance reconstruction robustness, when original low-frequency information is absent.
Subject(s)
Image Processing, Computer-Assisted , Phantoms, Imaging , Ultrasonography , Image Processing, Computer-Assisted/methods , Ultrasonography/methods , Humans , Tomography, X-Ray Computed/methodsABSTRACT
Direct synthesis of dimethyl carbonate (DMC) from CO2 plays an important role in carbon neutrality, but its efficiency is still far from the practical application, due to the limited understanding of the reaction mechanism and rational design of efficient catalyst. Herein, abundant electron-enriched lattice oxygen species were introduced into CeO2 catalyst by constructing the point defects and crystal-terminated phases in the crystal reconstruction process. Benefitting from the acid-base properties modulated by the electron-enriched lattice oxygen, the optimized CeO2 catalyst exhibited a much higher DMC yield of 22.2 mmol g-1 than the reported metal-oxide-based catalysts at the similar conditions. Mechanistic investigations illustrated that the electron-enriched lattice oxygen can provide abundant sites for CO2 adsorption and activation, and was advantageous of the formation of the weakly adsorbed active methoxy species. These were facilitating to the coupling of methoxy and CO2 for the key *CH3OCOO intermediate formation. More importantly, the weakened adsorption of *CH3OCOO on the electron-enriched lattice oxygen can switch the rate-determining-step (RDS) of DMC synthesis from *CH3OCOO formation to *CH3OCOO dissociation, and lower the corresponding activation barriers, thus giving rise to a high performance. This work provides insights into the underlying reaction mechanism for DMC synthesis from CO2 and methanol and the design of highly efficient catalysts.
ABSTRACT
OBJECTIVE: Point-scatterer detection plays a key role in medical ultrasound B-mode imaging. Speckle noise and insufficient spatial resolution are important factors affecting point-scatterer detection. To address this issue, normalized spatial autocorrelation in ultrasound B-mode imaging (NSACB) is proposed. METHODS: First, the acquired data are pre-processed by adding Gaussian white noise (GWN) with a certain signal-to-Gaussian white noise ratio (SGWNR). Next, normalized spatial autocorrelation is applied to the pre-processed data, and the data are divided into several new signals with different spatial lags. Then, the new signals are performed unsigned delay multiply and sum. Finally, the NSACB beamformed data are bandpass filtered by extracting the frequency component around twice the center frequency. Simulated and in vitro experiments were designed for validation. RESULTS: Simulations revealed that the lateral resolution of NSACB measured by the -6-dB mainlobe width can reach as high as 11.11% of delay and sum (DAS), 25.01% of filtered delay multiply and sum (F-DMAS) and 50% of LAG-FDMAS-SCF. The sidelobe level of the NSACB can be reduced at most by 28 dB. Experimental results of simple and complex scatterer phantoms indicate the image resolution of the proposed NSACB can even reach up to 18.76% of DAS, 27.28% of F-DMAS and 14.29% of LAG-FDMAS-SCF. Compared with these methods, the proposed NSACB can reduce the sidelobe level at least by 18 dB. CONCLUSION: Although the proposed method causes loss of the ability to observe hypo-echoic structures, these results suggest future work to determine the ability to detect breast microcalcifications, kidney stones, biopsy needle tracking and other scenarios requiring scatterer detection.
Subject(s)
Algorithms , Image Processing, Computer-Assisted , Image Processing, Computer-Assisted/methods , Ultrasonography/methods , Phantoms, Imaging , Signal-To-Noise RatioABSTRACT
Semi-supervised segmentation is highly significant in 3D medical image segmentation. The typical solutions adopt a teacher-student dual-model architecture, and they constrain the two models' decision consistency on the same segmentation task. However, the scarcity of medical samples can lower the diversity of tasks, reducing the effectiveness of consistency constraint. The issue can further worsen as the weights of the models gradually become synchronized. In this work, we have proposed to construct diverse joint-tasks using masked image modelling for enhancing the reliability of the consistency constraint, and develop a novel architecture consisting of a single teacher but multiple students to enjoy the additional knowledge decoupled from the synchronized weights. Specifically, the teacher and student models 'see' varied randomly-masked versions of an input, and are trained to segment the same targets but reconstruct different missing regions concurrently. Such joint-task of segmentation and reconstruction can have the two learners capture related but complementary features to derive instructive knowledge when constraining their consistency. Moreover, two extra students join the original one to perform an inter-student learning. The three students share the same encoding but different decoding designs, and learn decoupled knowledge by constraining their mutual consistencies, preventing themselves from suboptimally converging to the biased predictions of the dictatorial teacher. Experimental on four medical datasets show that our approach performs better than six mainstream semi-supervised methods. Particularly, our approach achieves at least 0.61% and 0.36% higher Dice and Jaccard values, respectively, than the most competitive approach on our in-house dataset. The code will be released at https://github.com/zxmboshi/DDL.
Subject(s)
Algorithms , Imaging, Three-Dimensional , Humans , Imaging, Three-Dimensional/methods , Supervised Machine Learning , Tomography, X-Ray Computed/methodsABSTRACT
In this study, a highly sensitive monoclonal antibody (mAb) was developed for the detection of aflatoxin B1 (AFB1) in maize and feed. Additionally, indirect competitive enzyme-linked immunosorbent assay (ic-ELISA) and time-resolved fluorescence immunoassay assay (TRFICA) were established. Firstly, the hapten AFB1-CMO was synthesized and conjugated with carrier proteins to prepare the immunogen for mouse immunization. Subsequently, mAb was generated using the classical hybridoma technique. The lowest half-maximal inhibitory concentration (IC50) of ic-ELISA was 38.6 ng/kg with a linear range of 6.25-100 ng/kg. The limits of detections (LODs) were 6.58 ng/kg and 5.54 ng/kg in maize and feed, respectively, with the recoveries ranging from 72% to 94%. The TRFICA was developed with a significantly reduced detection time of only 21 min, from sample processing to reading. Additionally, the limits of detection (LODs) for maize and feed were determined to be 62.7 ng/kg and 121 ng/kg, respectively. The linear ranges were 100-4000 ng/kg, with the recoveries ranging from 90% to 98%. In conclusion, the development of AFB1 mAb and the establishment of ic-ELISA for high-throughput sample detection, as well as TRFICA for rapid detection presented robust tools for versatile AFB1 detection in different scenarios.
ABSTRACT
Breast ultrasound computed tomography (USCT) has been gradually promoted to clinical application after years of rapid development. Compared with the traditional handheld ultrasound scanning method, the scanning plane of USCT is fixed at the coronal plane, and the scanning path is designed in advance; the acoustic window is not in direct contact with the breast, a lot of coupling medium (usually degassed water is used to fill the gaps between the probe and breast. The clinical application of breast USTC faces challenges: (1) the processes of water degassing, heating, filling, draining, and cleaning prolong the entire scan cycle and reduce patient throughput. (2) The breast is not stabilized and slight movements of the breast may cause motion artifacts in the USCT images. (3) The non-normal incidence of ultrasound into the breast causes reflected and transmitted signals received with a low signal-to-noise ratio (SNR) or even unable to be detected. This article proposes a coupling, stabilizing, and shaping strategy for the clinical application of USCT with a ring array transducer. The solid gel coupling agent (SGCA) is applied for coupling, and a set of SGCA moldings is designed to stabilize and shape the breast during scanning, the breast shape and size which vary from person to person are simplified into several models. The preparation time is reduced to less than 1 min by replacing disposable moldings. The results show that the breast after shaping is close to round in the coronal plane, and slopes of the breast skin are limited in the sagittal and transverse planes, the breast subcutaneous tissue (fat and glands) has a better contrast-to-noise ratio (CNR) and can be better distinguished in the reflection images than that of the breast without shaping. The mean value of the raw beamformed data which represents the reflection signal amplitude of breast subcutaneous tissue after shaping shows 1.5 times that of the breast without shaping, the signal-to-noise ratio (SNR) of the raw transmission signal data after breast shaping is overall higher than that of the breast without shaping. The application of SGCA moldings for breast coupling, stabilizing, and shaping also benefits establishing a standardized scanning process, the standardized diagnosis of the breast lesion, and the localization of breast lesions.
Subject(s)
Tomography, X-Ray Computed , Ultrasonography, Mammary , Female , Humans , Ultrasonography, Mammary/methods , Ultrasonography , Transducers , WaterABSTRACT
Contamination in food and the environment with fluoroquinolones (FQs) has become a serious threat to the global ecological balance and public health safety. Ofloxacin (OFL) is one of the most widely utilized sterilization agents in FQs. In the process of monitoring OFL, broad-spectrum monoclonal antibodies (mAb) cannot meet the demand for monospecific detection. Here, a computational chemistry-assisted hapten screening strategy was proposed in this study. Differences in the properties of antigenic epitopes were precisely extracted through a comprehensive comparative study of 16 common FQs molecules and a monospecific and ultrasensitive mAb-3B4 for OFL was successfully prepared. The screened fleroxacin (FLE) hapten was applied in a heterologous competition strategy resulting in a 20-fold improvement in the half inhibitory concentration (IC50) of mAb-3B4 to 0.0375 µg L-1 and cross-reacted only with marbofloxacin (MAR) in regulated FQs. In addition, a single-chain variable fragment (scFv) for OFL was constructed for the first time with an IC50 of 0.378 µg L-1. Molecular recognition mechanism studies validated the reliability of this strategy and revealed the key amino acid sites responsible for OFL specificity and sensitivity. Finally, ic-ELISA and GICA were established for OFL in real samples. This work provides new ideas for the preparation of monospecific mAb and improves the monitoring system of FQs.
Subject(s)
Computational Chemistry , Ofloxacin , Reproducibility of Results , Fluoroquinolones , Enzyme-Linked Immunosorbent Assay , Haptens , Anti-Bacterial Agents/chemistryABSTRACT
To monitor benzoic acid (BA) residues in liquid food samples, a monoclonal antibody (mAb)-based lateral flow immunoassay (LFA) was developed in this study. First, 2-aminobenzoic acid (2-AA), 3-aminobenzoic acid (3-AA), and 4-aminobenzoic acid (4-AA) were conjugated to BSA and used as immunogens. After cell fusion, mAb 6D8 from 4-AA-BSA performed best with an IC50 value of 0.21 mg L-1 using 3-AA-OVA as a heterogeneous antigen, which represented a 3.4-fold improvement compared with the homogeneous antigen 4-AA-BSA. Subsequently, eight kinds of CGNPs with sizes varying from 20.94 nm to 90.00 nm were synthesized for screening the suitable size to develop a sensitive LFA. Finally, a sensitive LFA based on colloidal gold (23.27 nm) nanoparticles was developed for screening BA with a cut-off value of 4 mg L-1, which could meet the requirement of BA detection in milk, Fanta, Sprite, Coca-Cola, and Smart samples.
Subject(s)
Antibodies, Monoclonal , Nanoparticles , Benzoic Acid , Immunoassay , AntigensABSTRACT
Aqueous zinc-ion batteries (AZIBs) with excellent safety, low-cost and environmental friendliness have great application potential in large-scale energy storage systems and thus have received extensive research interest. Layered oxovanadium phosphate dihydrate (VOPO4·2H2O) is an appealing cathode for AZIBs due to the unique layered framework and desirable discharge plateau, but bottlenecked by low operation voltage and unstable cycling. Herein, we propose delta-oxovanadium phosphate (δ-VOPO4) without conventional pre-embedding of metal elements or organics into the structure and paired it into AZIBs for the first time. Consequently, superior to the layered counterpart, δ-VOPO4 exhibits better performance with a prominent discharge voltage of 1.46 V and a higher specific capacity of 122.6 mA h g-1 at 1C (1C = 330 mA g-1), as well as an impressive capacity retention of 90.88 mA h g-1 after 1000 cycles under 10C. By investigation of structure resolution and theoretical calculation, this work well elucidates the structure-function relationship in vanadyl phosphates, offering more chances for exploration of new cathode materials to construct high performance AZIBs.
ABSTRACT
Although considerable efforts towards directly converting syngas to liquid fuels through Fischer-Tropsch synthesis have been made, developing catalysts with low CO2 selectivity for the synthesis of high-quality gasoline remains a big challenge. Herein, we designed a bifunctional catalyst composed of hydrophobic FeNa@Si-c and HZSM-5 zeolite, which exhibited a low CO2 selectivity of 14.3 % at 49.8 % CO conversion, with a high selectivity of 62.5 % for gasoline in total products. Molecular dynamic simulations and model experiments revealed that the diffusion of water molecules through hydrophilic catalyst was bidirectional, while the diffusion through hydrophobic catalyst was unidirectional, which were crucial to tune the water-gas shift reaction and control CO2 formation. This work provides a new fundamental understanding about the function of hydrophobic modification of catalysts in syngas conversion.
ABSTRACT
Full-aperture tomography (FAT) is the major image reconstruction method for a circular ring array (CRA)-based ultrasound computed tomography (USCT) system. The FAT technique requires transferring the reconstruction process from the temporal domain to the spatial domain, during which the imaging resolution of the USCT is degraded by the spatial-domain pulse width (SDPW) of backprojection areas. To tackle this challenge, this study investigates the characteristics of the SDPW and how it degrades the image resolution. We show that the SDPW depends on the frequency of the ultrasound and the position of the transmitting elements, receiving elements and the imaging point. To quantify the deterioration of image resolution associated with the position of the transmitting and receiving elements, a SDPW broadening factor (SDPWBF) is introduced. The results of numerical simulation show a smaller SDPWBFprovides a better reflection image resolution, and the distribution of SDPWBFshows that a shorter distance between the receiving element and the transmitting element yields a smaller SDPWBF. The SDPWBFis therefore able to be an indicator of selecting the signals acquired from the transmitting and receiving elements to perform optimal image resolution. Single-scatterer phantom andinvivoexperiments demonstrate how the SDPWBFaffects the USCT image spatial resolution and signal-to-noise ratio (SNR), and the results agree well with the theoretical predictions.
Subject(s)
Tomography, X-Ray Computed , Ultrasonic Waves , Adult , Female , Humans , Middle Aged , Algorithms , Breast/diagnostic imaging , Phantoms, Imaging , Tomography, X-Ray Computed/methodsABSTRACT
Multimodal medical image fusion (MMIF) is highly significant in such fields as disease diagnosis and treatment. The traditional MMIF methods are difficult to provide satisfactory fusion accuracy and robustness due to the influence of such possible human-crafted components as image transform and fusion strategies. Existing deep learning based fusion methods are generally difficult to ensure image fusion effect due to the adoption of a human-designed network structure and a relatively simple loss function and the ignorance of human visual characteristics during weight learning. To address these issues, we have presented the foveated differentiable architecture search (F-DARTS) based unsupervised MMIF method. In this method, the foveation operator is introduced into the weight learning process to fully explore human visual characteristics for the effective image fusion. Meanwhile, a distinctive unsupervised loss function is designed for network training by integrating mutual information, sum of the correlations of differences, structural similarity and edge preservation value. Based on the presented foveation operator and loss function, an end-to-end encoder-decoder network architecture will be searched using the F-DARTS to produce the fused image. Experimental results on three multimodal medical image datasets demonstrate that the F-DARTS performs better than several traditional and deep learning based fusion methods by providing visually superior fused results and better objective evaluation metrics.
ABSTRACT
In the field of disease diagnosis where only a small dataset of medical images may be accessible, the light-weight convolutional neural network (CNN) has become popular because it can help to avoid the over-fitting problem and improve computational efficiency. However, the feature extraction capability of the light-weight CNN is inferior to that of the heavy-weight counterpart. Although the attention mechanism provides a feasible solution to this problem, the existing attention modules, such as the squeeze and excitation module and the convolutional block attention module, have insufficient non-linearity, thereby influencing the ability of the light-weight CNN to discover the key features. To address this issue, we have proposed a spiking cortical model based global and local (SCM-GL) attention module. The SCM-GL module analyzes the input feature maps in parallel and decomposes each map into several components according to the relation between pixels and their neighbors. The components are weighted summed to obtain a local mask. Besides, a global mask is produced by discovering the correlation between the distant pixels in the feature map. The final attention mask is generated by combining the local and global masks, and it is multiplied by the original map so that the important components can be highlighted to facilitate accurate disease diagnosis. To appreciate the performance of the SCM-GL module, this module and some mainstream attention modules have been embedded into the popular light-weight CNN models for comparison. Experiments on the classification of brain MR, chest X-ray, and osteosarcoma image datasets demonstrate that the SCM-GL module can significantly improve the classification performance of the evaluated light-weight CNN models by enhancing the ability of discovering the suspected lesions and it is generally superior to state-of-the-art attention modules in terms of accuracy, recall, specificity and F1 score.
ABSTRACT
Research into plastic-degrading bacteria and fungi is important for understanding how microorganisms can be used to address the problem of plastic pollution and for developing new approaches to sustainable waste management and bioplastic production. In the present study, we isolated 55 bacterial and 184 fungal strains degrading polycaprolactone (PCL) in plastic waste samples from Dafeng coastal salt marshes, Jiangsu, China. Of these, Jonesia and Streptomyces bacteria also showed potential to degrade other types of petroleum-based polymers. The metabarcoding results proved the existence of plastisphere as a distinct ecological niche regardless of the plastic types where 27 bacterial and 29 fungal amplicon sequence variants (ASVs) were found to be significantly (p < 0.05) enriched, including some belonging to Alternaria (Ascomycota, Fungi) and Pseudomonas (Gammaproteobacteria, Bacteria) that were also mined out by the method of cultivation. Further assembly analyses demonstrated the importance of deterministic processes especially the environmental filtering effect of carbon content and pH on bacteria as well as the carbon and cation content on fungi in shaping the plastisphere communities in this ecosystem. Thus, the unique microbiome of the plastisphere in the terrestrial-marine ecotone is enriched with microorganisms that are potentially capable of utilizing petroleum-based polymers, making it a valuable resource for screening plastic biodegraders.
Subject(s)
Ascomycota , Microbiota , Petroleum , Polymers , Plastics , Bacteria/genetics , Biodegradation, EnvironmentalABSTRACT
Photocatalytic H2 evolution is a promising technology which could be instrumental in producing clean hydrogen energy. In regard to the photocatalyst, its band structure, morphology and light utilization have a significant influence on the H2 evolution rate and stability. Herein, a three-dimensional ordered macroporous nitrogen-vacancy carbon nitride (3DOM V-CN) photocatalyst was developed by combining vacancies with 3DOM structure for visible-light photocatalytic H2 evolution. This strategy preserved the structural properties of 3DOM to improve the light utilization and the specific surface area of the photocatalysts. Moreover, constructing suitable vacancies could trap electrons to facilitate the separation of photogenerated carriers, and extend the light absorption region of the photocatalysts by adjusting band structure, thus improving photocatalytic activity. Compared with CN (0.3 mmol h-1 g-1), 3DOM V-CN demonstrated a superior photocatalytic H2 evolution rate of 2.3 mmol h-1 g-1 (λ ≥ 420 nm) while possessing excellent stability. This work provides an effective and low-cost strategy for the design of the photocatalysts with high activity and stability by simultaneously tuning the band structure and morphology.
ABSTRACT
Vessel wall volume (VWV) is a 3-D ultrasound measurement for the assessment of therapy in patients with carotid atherosclerosis. Deep learning can be used to segment the media-adventitia boundary (MAB) and lumen-intima boundary (LIB) and to quantify VWV automatically; however, it typically requires large training data sets with expert manual segmentation, which are difficult to obtain. In this study, a UNet++ ensemble approach was developed for automated VWV measurement, trained on five small data sets (n = 30 participants) and tested on 100 participants with clinically diagnosed coronary artery disease enrolled in a multicenter CAIN trial. The Dice similarity coefficient (DSC), average symmetric surface distance (ASSD), Pearson correlation coefficient (r), Bland-Altman plots and coefficient of variation (CoV) were used to evaluate algorithm segmentation accuracy, agreement and reproducibility. The UNet++ ensemble yielded DSCs of 91.07%-91.56% and 87.53%-89.44% and ASSDs of 0.10-0.11 mm and 0.33-0.39 mm for the MAB and LIB, respectively; the algorithm VWV measurements were correlated (r = 0.763-0.795, p < 0.001) with manual segmentations, and the CoV for VWV was 8.89%. In addition, the UNet++ ensemble trained on 30 participants achieved a performance similar to that of U-Net and Voxel-FCN trained on 150 participants. These results suggest that our approach could provide accurate and reproducible carotid VWV measurements using relatively small training data sets, supporting deep learning applications for monitoring atherosclerosis progression in research and clinical trials.
Subject(s)
Carotid Arteries , Imaging, Three-Dimensional , Humans , Reproducibility of Results , Imaging, Three-Dimensional/methods , Carotid Arteries/diagnostic imaging , Ultrasonography/methods , AlgorithmsABSTRACT
In the proposed network, the features were first extracted from the gonioscopically obtained anterior segment photographs using the densely-connected high-resolution network. Then the useful information is further strengthened using the hybrid attention module to improve the classification accuracy. Between October 30, 2020, and January 30, 2021, a total of 146 participants underwent glaucoma screening. One thousand seven hundred eighty original images of the ACA were obtained with the gonioscope and slit lamp microscope. After data augmentation, 4457 images are used for the training and validation of the HahrNet, and 497 images are used to evaluate our algorithm. Experimental results demonstrate that the proposed HahrNet exhibits a good performance of 96.2% accuracy, 99.0% specificity, 96.4% sensitivity, and 0.996 area under the curve (AUC) in classifying the ACA test dataset. Compared with several deep learning-based classification methods and nine human readers of different levels, the HahrNet achieves better or more competitive performance in terms of accuracy, specificity, and sensitivity. Indeed, the proposed ACA classification method will provide an automatic and accurate technology for the grading of glaucoma.