Densely Connected Convolutional Networks Research Articles

Identification of CO2 , SO2 , and a Mixture of Both Gases Using Optical Imaging Combined with Convolutional Neural Network (CNN)

CO2 and SO2 gases are utilized in various industrial applications and are subjects of environmental research. However, these gases are considered toxic and pose dangers at certain concentrations. Therefore, it is crucial to monitor and control the exposure to these gases in the environment to prevent reaching hazardous levels that could endanger both humans and the environment. A non-contact detection and monitoring system is essential to minimize the adverse effects of direct gas exposure. In this research, a non-contact detection system for CO2, SO2, and mixed gases was developed using optical imaging analysis generated by infrared cameras. The images were captured using the FLIR Vue Pro-R infrared camera, with infrared absorbing gas sourced from a 50-watt tungsten lamp. Visual identification of these gases through optical imaging is challenging; however, this study successfully identified these gases using a Convolutional Neural Network (CNN). The CNN architecture used in this study is DenseNet (Densely Connected Convolutional Networks), comprising 169 convolution layers. The CNN model was trained and tested on experimental optical imaging data, categorized into three classes: CO2, SO2, and a mixture of gases. A total of 1030 optical imaging data points were utilized for training. Training was conducted using the AdamW optimization function over 28 epochs. The evaluation of results yielded accuracy, precision, recall, and F1-score metrics. The novelty of this study lies in the successful identification of CO2, SO2, and their mixture by the CNN model with an accuracy of 85%. Precision, recall, and F1-score values are all 0.85. These results indicate that the CNN model effectively distinguishes optical imaging of each gas (CO2, SO2, and their mixture) consistently and accurately. Consequently, it can be concluded that the CNN model performs well in distinguishing between these gases in optical imaging analysis.

Automated detection of vertebral body misalignments in orthogonal kV and MV guided radiotherapy: application to a comprehensive retrospective dataset

Objective. In image-guided radiotherapy (IGRT), off-by-one vertebral body misalignments are rare but potentially catastrophic. In this study, a novel detection method for such misalignments in IGRT was investigated using densely-connected convolutional networks (DenseNets) for applications towards real-time error prevention and retrospective error auditing. Approach. A total of 4213 images acquired from 527 radiotherapy patients aligned with planar kV or MV radiographs were used to develop and test error-detection software modules. Digitally reconstructed radiographs (DRRs) and setup images were retrieved and co-registered according to the clinically applied alignment contained in the DICOM REG files. A semi-automated algorithm was developed to simulate patient positioning errors on the anterior-posterior (AP) and lateral (LAT) images shifted by one vertebral body. A DenseNet architecture was designed to classify either AP images individually or AP and LAT image pairs. Receiver-operator characteristic curves (ROC) and areas under the curves (AUC) were computed to evaluate the classifiers on test subsets. Subsequently, the algorithm was applied to the entire dataset in order to retrospectively determine the absolute off-by-one vertebral body error rate for planar radiograph guided RT at our institution from 2011–2021. Main results. The AUCs for the kV models were 0.98 for unpaired AP and 0.99 for paired AP-LAT. The AUC for the MV AP model was 0.92. For a specificity of 95%, the paired kV model achieved a sensitivity of 99%. Application of the model to the entire dataset yielded a per-fraction off-by-one vertebral body error rate of 0.044% [0.0022%, 0.21%] for paired kV IGRT including one previously unreported error. Significance. Our error detection algorithm was successful in classifying vertebral body positioning errors with sufficient accuracy for retrospective quality control and real-time error prevention. The reported positioning error rate for planar radiograph IGRT is unique in being determined independently of an error reporting system.

Deep learning-enabled detection of hypoxic-ischemic encephalopathy after cardiac arrest in CT scans: a comparative study of 2D and 3D approaches.

To establish a deep learning model for the detection of hypoxic-ischemic encephalopathy (HIE) features on CT scans and to compare various networks to determine the best input data format. 168 head CT scans of patients after cardiac arrest were retrospectively identified and classified into two categories: 88 (52.4%) with radiological evidence of severe HIE and 80 (47.6%) without signs of HIE. These images were randomly divided into a training and a test set, and five deep learning models based on based on Densely Connected Convolutional Networks (DenseNet121) were trained and validated using different image input formats (2D and 3D images). All optimized stacked 2D and 3D networks could detect signs of HIE. The networks based on the data as 2D image data stacks provided the best results (S100: AUC: 94%, ACC: 79%, S50: AUC: 93%, ACC: 79%). We provide visual explainability data for the decision making of our AI model using Gradient-weighted Class Activation Mapping. Our proof-of-concept deep learning model can accurately identify signs of HIE on CT images. Comparing different 2D- and 3D-based approaches, most promising results were achieved by 2D image stack models. After further clinical validation, a deep learning model of HIE detection based on CT images could be implemented in clinical routine and thus aid clinicians in characterizing imaging data and predicting outcome.

Feature analysis and automatic classification of B-mode ultrasound images of fatty liver

Fatty liver refers to the pathological changes caused by excessive fat accumulation in hepatocytes with various causes. The medical analysis of B-mode ultrasound images(BMUI) is an important method for the identification of the degree of fatty liver. In this paper, a transfer learning method based on Densely Connected Convolutional Networks (DenseNet) and Light Gradient Boosting Machine (LightGBM) was proposed to automatically recognize the four classifications of BMUI of fatty liver. In comparison with the traditional gray and texture feature analysis, this method has the advantages of avoiding the manual extraction the region of interest of BMUI required by the traditional machine learning method for the analysis of feature performance. The complete BMUI were used to train the pathological feature classification model of fatty liver, and this method not only provides the comparison of ultrasonic echo intensity in the near field and far field of BMUI, but also retains the internal texture features and gray information of fatty liver. On the experimental data set, the classification model combined with DenseNet201 network and LightGBM proposed in this paper improved the recognition accuracy by 17.5% on average compared with the support vector machine and migration models based on ResNet101 network. In comparison with the two models above, the size was reduced by 99% on average, indicating remarkably improvement.

Number of necessary training examples for Neural Networks with different number of trainable parameters

In this work, the network complexity should be reduced with a concomitant reduction in the number of necessary training examples. The focus thus was on the dependence of proper evaluation metrics on the number of adjustable parameters of the considered deep neural network. The used data set encompassed Hematoxylin and Eosin (H&E) colored cell images provided by various clinics. We used a deep convolutional neural network to get the relation between a model’s complexity, its concomitant set of parameters, and the size of the training sample necessary to achieve a certain classification accuracy. The complexity of the deep neural networks was reduced by pruning a certain amount of filters in the network. As expected, the unpruned neural network showed best performance. The network with the highest number of trainable parameter achieved, within the estimated standard error of the optimized cross-entropy loss, best results up to 30% pruning. Strongly pruned networks are highly viable and the classification accuracy declines quickly with decreasing number of training patterns. However, up to a pruning ratio of 40%, we found a comparable performance of pruned and unpruned deep convolutional neural networks (DCNN) and densely connected convolutional networks (DCCN).

Mine MIMO Depth Receiver: An Intelligent Receiving Model Based on Densely Connected Convolutional Networks.

Multiple-input multiple-output (MIMO) systems suffer from high BER in the mining environment. In this paper, the mine MIMO depth receiver model is proposed. The model uses densely connected convolutional networks for feature extraction and constructs multiple binary classifiers to recover the original information. Compared with conventional MIMO receivers, the model has no error accumulation caused by processes such as decoding and demodulation. The experimental results show that the model has better performance than conventional decoding methods under different modulation codes and variations in the number of transmitting terminals. Furthermore, we demonstrate that the model can still achieve effective decoding and recover the original information with some data loss at the receiver.

Densely connected convolutional networks-based COVID-19 screening model.

The extensively utilized tool to detect novel coronavirus (COVID-19) is a real-time polymerase chain reaction (RT-PCR). However, RT-PCR kits are costly and consume critical time, around 6 to 9 hours to classify the subjects as COVID-19(+) or COVID-19(-). Due to the less sensitivity of RT-PCR, it suffers from high false-negative results. To overcome these issues, many deep learning models have been implemented in the literature for the early-stage classification of suspected subjects. To handle the sensitivity issue associated with RT-PCR, chest CT scans are utilized to classify the suspected subjects as COVID-19 (+), tuberculosis, pneumonia, or healthy subjects. The extensive study on chest CT scans of COVID-19 (+) subjects reveals that there are some bilateral changes and unique patterns. But the manual analysis from chest CT scans is a tedious task. Therefore, an automated COVID-19 screening model is implemented by ensembling the deep transfer learning models such as Densely connected convolutional networks (DCCNs), ResNet152V2, and VGG16. Experimental results reveal that the proposed ensemble model outperforms the competitive models in terms of accuracy, f-measure, area under curve, sensitivity, and specificity.

Research on Intrusion Detection Technology Based on Densely Connected Convolutional Neural Networks

Convolutional Neural Network (CNN) is widely used in the field of intrusion detection technology. It is generally believed that the deeper the network structure, the more accurate in feature extraction and detection accuracy. However, it is accompanied with the problems of gradient dispersion, insufficient generalization ability and low accuracy of parameters. In view of the above problems, the Densely Connected Convolutional Network (DCCNet) is applied into the intrusion detection technology, and achieve the purpose of improving the detection accuracy by using the hybrid loss function. Experiments are performed with the KDD 99 data set, and the experimental results are compared with the commonly used LeNet neural network and VggNet neural network structure. Finally, the analysis shows that the accuracy of detection is improved, and the problem of gradient vanishing during training is alleviated.

Logo Recognition with the Use of Deep Convolutional Neural Networks

Automatic logo recognition is gaining importance due to the increasing number of its applications. Unlike other object recognition tasks, logo recognition is more challenging because of the limited amount of the available original data. In this paper, the transfer leaning technique was applied to a Deep Convolutional Neural Network model to guarantee logo recognition using a small computational overhead. The proposed method was based on the Densely Connected Convolutional Networks (DenseNet). The experimental results show that for the FlickrLogos-32 logo recognition dataset, our proposed method performs comparably with state-of-the-art methods while using fewer parameters.

Semi-supervised method for image texture classification of pituitary tumors via CycleGAN and optimized feature extraction

BackgroundAccurately determining the softness level of pituitary tumors preoperatively by using their image textures can provide a basis for surgical options and prognosis. Existing methods for this problem require manual intervention, which could hinder the efficiency and accuracy considerably.MethodsWe present an automatic method for diagnosing the texture of pituitary tumors using unbalanced sequence image data. Firstly, for the small sample problem in our pituitary tumor MRI image dataset where T1 and T2 sequence data are unbalanced (due to data missing) and under-sampled, our method uses a CycleGAN (Cycle-Consistent Adversarial Networks) model for domain conversion to obtain fully sampled MRI spatial sequence. Then, it uses a DenseNet (Densely Connected Convolutional Networks)-ResNet(Deep Residual Networks) based Autoencoder framework to optimize the feature extraction process for pituitary tumor image data. Finally, to take advantage of sequence data, it uses a CRNN (Convolutional Recurrent Neural Network) model to classify pituitary tumors based on their predicted softness levels.ResultsExperiments show that our method is the best in terms of efficiency and accuracy (91.78%) compared to other methods.ConclusionsWe propose a semi-supervised method for grading pituitary tumor texture. This method can accurately determine the softness level of pituitary tumors, which provides convenience for surgical selection and prognosis, and improves the diagnostic efficiency of pituitary tumors.

Finger-Vein Recognition Based on Densely Connected Convolutional Network Using Score-Level Fusion With Shape and Texture Images

Biometrics using finger-veins is a recognition method based on the shape of veins in fingers, and it has the advantage of difficulty to be forged. However, a shade is inevitably produced due to the bones and fingernails, and a change in illumination occurs when acquiring the images of finger-veins. Previous studies have conducted finger-vein recognition using a single-type texture image or finger-vein segmented image (shape image). A texture image provides numerous features, but it is vulnerable to the changes in illumination during recognition and contains noises in regions other than the finger-vein region. A shape image is less affected by noises; however, the recognition accuracy is significantly reduced due to fewer features available and mis-segmented regions caused by shades. In this study, therefore, rough finger-vein regions in an image are detected to reduce the effect of mis-segmented regions, to complement the drawbacks of shape image-based finger-vein recognition. Furthermore, score-level fusion is performed for two output scores of deep convolutional neural network extracted from the texture and shape images, which can reduce the sensitivity to noise, while diverse features provided in the texture image are used efficiently. Two open databases, the Shandong University homologous multi-modal traits finger-vein database and Hong Kong Polytech University finger image database, are used for experiments, and the proposed method shows better recognition performance than the state-of-the-art method.

An Automatic Method for Water Extraction From High Spatial Resolution GF-1 Imagery Based On A Deep Learning Algorithm

Traditional remote sensing methods for water extraction generally rely on complex spectral analysis and subjective thresholds, which do not satisfy the requirements for automated high-accuracy water detection from high spatial resolution images. In this paper, a novel framework of Convolutional Neural Networks (CNNs) is proposed for automatic water extraction based on Densely Connected Convolutional Networks (DenseNet) and a U-net structure. The proposed CNN model is trained on 1300 GF-1 images and tested on another 50 GF-1 images. Six of the GF-1 images are used for experiments to demonstrate the accuracy of the results and compare them with those from NDWI and supervised classification methods. The results show that our method has better computation efficiency and classification accuracy. The new framework has remarkable advantages for automatic water extraction from high-resolution images, especially in maintaining the integrity of small water bodies and in overcoming interference from clouds, shadows in built-up-areas, and mountainous areas with snow cover.

Differential Diagnosis for Pancreatic Cysts in CT Scans Using Densely-Connected Convolutional Networks.

The lethal nature of pancreatic ductal adenocarcinoma (PDAC) calls for early differential diagnosis of pancreatic cysts, which are identified in up to 16% of normal subjects, and some of them may develop into PDAC. Pancreatic cysts have a large variation in size and shape, and the precise segmentation of them remains rather challenging, which restricts the computer-aided interpretation of CT images acquired for differential diagnosis. We propose a computer-aided framework for early differential diagnosis of pancreatic cysts without pre-segmenting the lesions using densely-connected convolutional networks (Dense-Net). The Dense-Net learns high-level features from whole abnormal pancreas and builds mappings between medical imaging appearance to different pathological types of pancreatic cysts. To enhance the clinical applicability, we integrate saliency maps in the framework to assist the physicians to understand the decision of the deep learning method. The test on a cohort of 206 patients with 4 pathologically confirmed subtypes of pancreatic cysts has achieved an overall accuracy of 72.8%, which is significantly higher than the baseline accuracy of 48.1%. The superior performance on this challenging dataset strongly supports the clinical potential of our developed method.

Presence Detection of Surgical Tool Via Densely Connected Convolutional Networks

Surgical tool detection is important to surgical workflow recognition. It is considered as an essential task in surgical phase recognition. Recently, Densely Connected Convolutional Networks have gained a huge success in computer vision applications, especially in object detection and image Classification. In this paper, we proposed a method to solve the surgical tool presence detection problem as a multi-label classification problem based on Densely Connected Convolutional Networks. The performance of the proposed method has been evaluated in the surgical tool presence detection challenge dataset held by Modeling and Monitoring of Computer Assisted Interventions workshop. The result shows that our proposed model has achieved significant success in detecting surgical tool and got a mean average precision of 62.9% on the testing data. The technology studied in this paper has broad application prospects in computer-aided surgical systems and is a core component of the artificial intelligence medical operating room in the future.

DA-U-Net: Densely Connected Convolutional Networks and Decoder with Attention Gate for Retinal Vessel Segmentation

The segmentation of retinal vessels is greatly significant for doctors to diagnose the fundus diseases. However, existing methods have various problems in the segmentation of the retinal vessels, such as insufficient segmentation of retinal vessels, weak anti-noise interference ability. Aiming to the shortcomings of existed methods, this paper proposes an improved model based on the U-Net networks, which contains densely-connected convolutional network and a novel attention gate (AG) model, referred as Densely-Attention-U-Net (DA-U-Net), to automatically segment the retinal blood vessels. The method can alleviate the vanishing-gradient problem, strengthen feature propagation, substantially reduce the number of parameters, and automatically learn to focus on target structures without additional supervision. By verifying the method on the DRIVE datasets, the segmentation accuracy rate is 96.09%, higher than that of U-Net and R2U-Net.

Finger-Vein Recognition Based on Deep DenseNet Using Composite Image

Finger-vein recognition has the advantages of high immutability, as finger veins are located under the skin, high user convenience, as a non-invasive and contactless capture device, is used, and high readability even when one of the fingers is damaged or not available for recognition. However, there is an issue of recognition performance degradation caused by finger positional variation, misalignment, and shading from uneven illumination. The existing hand-crafted feature-based methods have exhibited varied performance depending on how these issues were handled by pre-processing. To overcome this shortcoming of hand-crafted feature-based methods, convolutional neural network (CNN)-based recognition methods have been researched. The existing systems based on a CNN use two methods: using a difference image as the input to the network and calculating the distance between feature vectors extracted from the CNN. Difference images can be susceptible to noise as they are generated by differences in pixel values. Also, the method for calculating the distance between feature vectors cannot employ all layers of the trained network and has less accuracy than the method employing difference images. To address these issues, this paper examined a method less susceptible to noise and which uses the entire network; a composite image of two finger-vein images was used as the input to a deep, densely-connected convolutional network (DenseNet). Two open databases, namely Shandong University homologous multi-modal traits (SDUMLA-HMT) finger-vein database and The Hong Kong Polytechnic University finger image database (version 1), were used for experiments and the results show that the proposed method has greater performance than the existing methods.