RESUMEN

In today's biometric and commercial settings, state-of-the-art image processing relies solely on artificial intelligence and machine learning which provides a high level of accuracy. However, these principles are deeply rooted in abstract, complex "black-box systems". When applied to forensic image identification, concerns about transparency and accountability emerge. This study explores the impact of two challenging factors in automated facial identification: facial expressions and head poses. The sample comprised 3D faces with nine prototype expressions, collected from 41 participants (13 males, 28 females) of European descent aged 19.96 to 50.89 years. Pre-processing involved converting 3D models to 2D color images (256 × 256 px). Probes included a set of 9 images per individual with head poses varying by 5° in both left-to-right (yaw) and up-and-down (pitch) directions for neutral expressions. A second set of 3,610 images per individual covered viewpoints in 5° increments from -45° to 45° for head movements and different facial expressions, forming the targets. Pair-wise comparisons using ArcFace, a state-of-the-art face identification algorithm yielded 54,615,690 dissimilarity scores. Results indicate that minor head deviations in probes have minimal impact. However, the performance diminished as targets deviated from the frontal position. Right-to-left movements were less influential than up and down, with downward pitch showing less impact than upward movements. The lowest accuracy was for upward pitch at 45°. Dissimilarity scores were consistently higher for males than for females across all studied factors. The performance particularly diverged in upward movements, starting at 15°. Among tested facial expressions, happiness and contempt performed best, while disgust exhibited the lowest AUC values.

Asunto(s)

RESUMEN

Objectives: This study aimed to develop an approach to enhance the model precision by artificial images. Materials and Methods: Given an epidemiological study designed to predict 1 response using f features with M samples, each feature was converted into a pixel with certain value. Permutated these pixels into F orders, resulting in F distinct artificial image sample sets. Based on the experience of image recognition techniques, appropriate training images results in higher precision model. In the preliminary experiment, a binary response was predicted by 76 features, the sample set included 223 patients and 1776 healthy controls. Results: We randomly selected 10 000 artificial sample sets to train the model. Models' performance (area under the receiver operating characteristic curve values) depicted a bell-shaped distribution. Conclusion: The model construction strategy developed in the research has potential to capture feature order related information and enhance model predictability.

RESUMEN

The identification of medical images is an essential task in computer-aided diagnosis, medical image retrieval and mining. Medical image data mainly include electronic health record data and gene information data, etc. Although intelligent imaging provided a good scheme for medical image analysis over traditional methods that rely on the handcrafted features, it remains challenging due to the diversity of imaging modalities and clinical pathologies. Many medical image identification methods provide a good scheme for medical image analysis. The concepts pertinent of methods, such as the machine learning, deep learning, convolutional neural networks, transfer learning, and other image processing technologies for medical image are analyzed and summarized in this paper. We reviewed these recent studies to provide a comprehensive overview of applying these methods in various medical image analysis tasks, such as object detection, image classification, image registration, segmentation, and other tasks. Especially, we emphasized the latest progress and contributions of different methods in medical image analysis, which are summarized base on different application scenarios, including classification, segmentation, detection, and image registration. In addition, the applications of different methods are summarized in different application area, such as pulmonary, brain, digital pathology, brain, skin, lung, renal, breast, neuromyelitis, vertebrae, and musculoskeletal, etc. Critical discussion of open challenges and directions for future research are finally summarized. Especially, excellent algorithms in computer vision, natural language processing, and unmanned driving will be applied to medical image recognition in the future.

Asunto(s)

RESUMEN

Background: Dietary monitoring is critical to maintaining human health. Social media platforms are widely used for daily recording and communication for individuals' diets and activities. The textual content shared on social media offers valuable resources for dietary monitoring. Objective: This study aims to describe the development of iFood, an applet providing personal dietary monitoring based on social media content, and validate its usability, which will enable efficient personal dietary monitoring. Methods: The process of the development and validation of iFood is divided into four steps: Diet datasets construction, diet record and analysis, diet monitoring applet design, and diet monitoring applet usability assessment. The diet datasets were constructed with the data collected from Weibo, Meishijie, and diet guidelines, which will be used as the basic knowledge for further model training in the phase of diet record and analysis. Then, the friendly user interface was designed to link users with backend functions. Finally, the applet was deployed as a WeChat applet and 10 users from the Beijing Union Medical College have been recruited to validate the usability of iFood. Results: Three dietary datasets, including User Visual-Textual Dataset, Dietary Information Expansion Dataset, and Diet Recipe Dataset have been constructed. The performance of 4 models for recognizing diet and fusing unimodality data was 40.43%(dictionary-based model), 18.45%(rule-based model), 59.95%(Inception-ResNet-v2), and 51.38% (K-nearest neighbor), respectively. Furthermore, we have designed a user-friendly interface for the iFood applet and conducted a usability assessment, which resulted in an above-average usability score. Conclusions: iFood is effective for managing individual dietary behaviors through its seamless integration with social media data. This study suggests that future products could utilize social media data to promote healthy lifestyles.

RESUMEN

According to the survey statistics, most traffic accidents are caused by the driver's behavior and status irregularities. Because there is no multi-level dangerous state grading system at home and abroad, this paper proposes a complex state grading system for real-time detection and dynamic tracking of the driver's state. The system uses OpenMV as the acquisition camera combined with the cradle head tracking system to collect the driver's current driving image in real-time dynamically, combines the YOLOX algorithm with the OpenPose algorithm to judge the driver's dangerous driving behavior by detecting unsafe objects in the cab and the driver's posture, and combines the improved Retinaface face detection algorithm with the Dlib feature-point algorithm to discriminate the fatigue driving state of the driver. The experimental results show that the accuracy of the three driver danger levels (R1, R2, and R3) obtained by the proposed system reaches 95.8%, 94.5%, and 96.3%, respectively. The experimental results of this system have a specific practical significance in driver-distracted driving warnings.

RESUMEN

BACKGROUND: C-arm fluoroscopy, as an effective diagnosis and treatment method for spine surgery, can help doctors perform surgery procedures more precisely. In clinical surgery, the surgeon often determines the specific surgical location by comparing C-arm X-ray images with digital radiography (DR) images. However, this heavily relies on the doctor's experience. OBJECTIVE: In this study, we design a framework for automatic vertebrae detection as well as vertebral segment matching (VDVM) for the identification of vertebrae in C-arm X-ray images. METHODS: The proposed VDVM framework is mainly divided into two parts: vertebra detection and vertebra matching. In the first part, a data preprocessing method is used to improve the image quality of C-arm X-ray images and DR images. The YOLOv3 model is then used to detect the vertebrae, and the vertebral regions are extracted based on their position. In the second part, the Mobile-Unet model is first used to segment the vertebrae contour of the C-arm X-ray image and DR image based on vertebral regions respectively. The inclination angle of the contour is then calculated using the minimum bounding rectangle and corrected accordingly. Finally, a multi-vertebra strategy is applied to measure the visual information fidelity for the vertebral region, and the vertebrae are matched based on the measured results. RESULTS: We use 382 C-arm X-ray images and 203 full length X-ray images to train the vertebra detection model, and achieve a mAP of 0.87 in the test dataset of 31 C-arm X-ray images and 0.96 in the test dataset of 31 lumbar DR images. Finally, we achieve a vertebral segment matching accuracy of 0.733 on 31 C-arm X-ray images. CONCLUSIONS: A VDVM framework is proposed, which performs well for the detection of vertebrae and achieves good results in vertebral segment matching.

Asunto(s)

RESUMEN

Mosquitoes are one of the deadliest insects, causing harm to humans worldwide. Preemptive prevention and forecasting are important to prevent mosquito-borne diseases. However, current mosquito identification is mostly conducted manually, which consumes time, wastes labor, and causes human error. In this study, we developed an automatic image analysis method to identify mosquito species using a deep learning-based object detection technique. Color and fluorescence images of live mosquitoes were acquired using a mosquito capture device and were used to develop a deep learning-based object detection model. Among the deep learning-based object identification models, the combination of a swine transformer and a faster region-convolutional neural network model demonstrated the best performance, with a 91.7% F1-score. This indicates that the proposed automatic identification method can be rapidly applied for efficient analysis of species and populations of vector-borne mosquitoes with reduced labor in the field.

RESUMEN

Cervical cancer is the fourth most common cancer among women, and cytopathological images are often used to screen for this cancer. However, manual examination is very troublesome and the misdiagnosis rate is high. In addition, cervical cancer nest cells are denser and more complex, with high overlap and opacity, increasing the difficulty of identification. The appearance of the computer aided automatic diagnosis system solves this problem. In this paper, a weakly supervised cervical cancer nest image identification approach using Conjugated Attention Mechanism and Visual Transformer (CAM-VT), which can analyze pap slides quickly and accurately. CAM-VT proposes conjugated attention mechanism and visual transformer modules for local and global feature extraction respectively, and then designs an ensemble learning module to further improve the identification capability. In order to determine a reasonable interpretation, comparative experiments are conducted on our datasets. The average accuracy of the validation set of three repeated experiments using CAM-VT framework is 88.92%, which is higher than the optimal result of 22 well-known deep learning models. Moreover, we conduct ablation experiments and extended experiments on Hematoxylin and Eosin stained gastric histopathological image datasets to verify the ability and generalization ability of the framework. Finally, the top 5 and top 10 positive probability values of cervical nests are 97.36% and 96.84%, which have important clinical and practical significance. The experimental results show that the proposed CAM-VT framework has excellent performance in potential cervical cancer nest image identification tasks for practical clinical work.

Asunto(s)

RESUMEN

In situ synchrotron high-energy X-ray powder diffraction (XRD) is highly utilized by researchers to analyze the crystallographic structures of materials in functional devices (e.g. battery materials) or in complex sample environments (e.g. diamond anvil cells or syntheses reactors). An atomic structure of a material can be identified by its diffraction pattern along with a detailed analysis of the Rietveld refinement which yields rich information on the structure and the material, such as crystallite size, microstrain and defects. For in situ experiments, a series of XRD images is usually collected on the same sample under different conditions (e.g. adiabatic conditions) yielding different states of matter, or is simply collected continuously as a function of time to track the change of a sample during a chemical or physical process. In situ experiments are usually performed with area detectors and collect images composed of diffraction patterns. For an ideal powder, the diffraction pattern should be a series of concentric Debye-Scherrer rings with evenly distributed intensities in each ring. For a realistic sample, one may observe different characteristics other than the typical ring pattern, such as textures or preferred orientations and single-crystal diffraction spots. Textures or preferred orientations usually have several parts of a ring that are more intense than the rest, whereas single-crystal diffraction spots are localized intense spots owing to diffraction of large crystals, typically >10 µm. In this work, an investigation of machine learning methods is presented for fast and reliable identification and separation of the single-crystal diffraction spots in XRD images. The exclusion of artifacts during an XRD image integration process allows a precise analysis of the powder diffraction rings of interest. When it is trained with small subsets of highly diverse datasets, the gradient boosting method can consistently produce high-accuracy results. The method dramatically decreases the amount of time spent identifying and separating single-crystal diffraction spots in comparison with the conventional method.

RESUMEN

Breast cancer is one of the leading causes of death among women worldwide. Histopathological images have proven to be a reliable way to find out if someone has breast cancer over time, however, it could be time consuming and require much resources when observed physically. In order to lessen the burden on the pathologists and save lives, there is need for an automated system to effectively analysis and predict the disease diagnostic. In this paper, a lightweight separable convolution network (LWSC) is proposed to automatically learn and classify breast cancer from histopathological images. The proposed architecture aims to treat the problem of low quality by extracting the visual trainable features of the histopathological image using a contrast enhancement algorithm. LWSC model implements separable convolution layers stacked in parallel with multiple filters of different sizes in order to obtain wider receptive fields. Additionally, the factorization and the utilization of bottleneck convolution layers to reduce model dimension were introduced. These methods reduce the number of trainable parameters as well as the computational cost sufficiently with greater non-linear expressive capacity than plain convolutional networks. The evaluation results depict that the proposed LWSC model performs optimally, obtaining 97.23% accuracy, 97.71% sensitivity, and 97.93% specificity on multi-class categories. Compared with other models, the proposed LWSC obtains comparable performance.

RESUMEN

Because of its unique characteristics of small specific gravity, high strength, and corrosion resistance, the carbon fiber sucker rod has been widely used in petroleum production. However, there is still a lack of corresponding online testing methods to detect its integrity during the process of manufacturing. Ultrasonic nondestructive testing has become one of the most accepted methods for inspection of homogeneous and fixed-thickness composites, or layered and fixed-interface-shape composites, but a carbon fiber sucker rod with multi-layered structures and irregular interlayer interfaces increases the difficulty of testing. In this paper, a novel defect detection method based on multi-sensor information fusion and a deep belief network (DBN) model was proposed to identify online its defects. A water-immersed ultrasonic array with 32 ultrasonic probes was designed to realize the online and full-coverage scanning of carbon fiber rods in radial and axial positions. Then, a multi-sensor information fusion method was proposed to integrate amplitudes and times-of-flight of the received ultrasonic pulse-echo signals with the spatial angle information of each probe into defect images with obvious defects including small cracks, transverse cracks, holes, and chapped cracks. Three geometric features and two texture features from the defect images characterizing the four types of defects were extracted. Finally, a DBN-based defect identification model was constructed and trained to identify the four types of defects of the carbon fiber rods. The testing results showed that the defect identification accuracy of the proposed method was 95.11%.

RESUMEN

Pneumonia is a prevalent severe respiratory infection that affects the distal and alveoli airways. Across the globe, it is a serious public health issue that has caused high mortality rate of children below five years old and the aged citizens who must have had previous chronic-related ailment. Pneumonia can be caused by a wide range of microorganisms, including virus, fungus, bacteria, which varies greatly across the globe. The spread of the ailment has gained computer-aided diagnosis (CAD) attention. This paper presents a multi-channel-based image processing scheme to automatically extract features and identify pneumonia from chest X-ray images. The proposed approach intends to address the problem of low quality and identify pneumonia in CXR images. Three channels of CXR images, namely, the Local Binary Pattern (LBP), Contrast Enhanced Canny Edge Detection (CECED), and Contrast Limited Adaptive Histogram Equalization (CLAHE) CXR images are processed by deep neural networks. CXR-related features of LBP images are extracted using shallow CNN, features of the CLAHE CXR images are extracted by pre-trained inception-V3, whereas the features of CECED CXR images are extracted using pre-trained MobileNet-V3. The final feature weights of the three channels are concatenated and softmax classification is utilized to determine the final identification result. The proposed network can accurately classify pneumonia according to the experimental result. The proposed method tested on publicly available dataset reports accuracy of 98.3%, sensitivity of 98.9%, and specificity of 99.2%. Compared with the single models and the state-of-the-art models, our proposed network achieves comparable performance.

RESUMEN

It is a well-known fact that diabetic retinopathy (DR) is one of the most common causes of visual impairment between the ages of 25 and 74 around the globe. Diabetes is caused by persistently high blood glucose levels, which leads to blood vessel aggravations and vision loss. Early diagnosis can minimise the risk of proliferated diabetic retinopathy, which is the advanced level of this disease, and having higher risk of severe impairment. Therefore, it becomes important to classify DR stages. To this effect, this paper presents a weighted fusion deep learning network (WFDLN) to automatically extract features and classify DR stages from fundus scans. The proposed framework aims to treat the issue of low quality and identify retinopathy symptoms in fundus images. Two channels of fundus images, namely, the contrast-limited adaptive histogram equalization (CLAHE) fundus images and the contrast-enhanced canny edge detection (CECED) fundus images are processed by WFDLN. Fundus-related features of CLAHE images are extracted by fine-tuned Inception V3, whereas the features of CECED fundus images are extracted using fine-tuned VGG-16. Both channels' outputs are merged in a weighted approach, and softmax classification is used to determine the final recognition result. Experimental results show that the proposed network can identify the DR stages with high accuracy. The proposed method tested on the Messidor dataset reports an accuracy level of 98.5%, sensitivity of 98.9%, and specificity of 98.0%, whereas on the Kaggle dataset, the proposed model reports an accuracy level of 98.0%, sensitivity of 98.7%, and specificity of 97.8%. Compared with other models, our proposed network achieves comparable performance.

RESUMEN

With precision medicine as the goal, the human biobank of each country should be analyzed to determine the complete research results related to genetic diseases. In addition, with the increase in medical imaging data, automatic image processing with image recognition has been widely studied and applied in biomedicine. However, case-control data imbalance often occurs in human biobanks, which is usually solved by the statistical method SAIGE. Due to the huge amount of genetic data in human biobanks, the direct use of the SAIGE method often faces the problem of insufficient computer memory to support calculations and excessive calculation time. The other method is to use sampling to adjust the data to balance the case-control ratio, which is called Synthetic Minority Oversampling Technique (SMOTE). Our study employed the Manhattan plot and genetic disease information from the Taiwan Biobank to adjust the imbalance in the case-control ratio by SMOTE, called "TW-SMOTE." We further used a deep learning image recognition system to identify the TW-SMOTE. We found that TW-SMOTE can achieve the same results as that of SAIGE and the UK Biobank (UKB). The processing of the technical data can be equivalent to the use of data plots with a relatively large UKB sample size and achieve the same effect as that of SAIGE in addressing data imbalance.

RESUMEN

Anomaly detection is a classical problem in computer vision, namely the determination of the normal from the abnormal when datasets are highly biased toward one class (normal) due to the insufficient sample size of the other class (abnormal). We introduce a novel model that utilizes two decoders to share two encoders, respectively, forming two sets of network structures of encoder-decoder-encoder called EDE, which are used to map image distributions to predefined latent distributions and vice versa. In addition, we propose an innovative two-stage training mode. The first stage is roughly the same as the traditional autoencoder (AE) training, using the reconstruction loss of images and latent vectors for training. The second stage uses the idea of generative confrontation to send one of the two groups of reconstructed vectors into another EDE structure to generate fake images and latent vectors. This EDE structure needs to achieve two goals to distinguish the source of the data: the first is to maximize the difference between the fake image and the real image; the second is to maximize the difference between the fake latent vector and the reconstructed vector. Another EDE structure has the opposite goal. This network structure combined with special training methods not only well avoids the shortcomings of generative adversarial networks (GANs) and AEs, but also achieves state-of-the-art performance evaluated on several publicly available image datasets.

RESUMEN

The identification of different meat cuts for labeling and quality control on production lines is still largely a manual process. As a result, it is a labor-intensive exercise with the potential for not only error but also bacterial cross-contamination. Artificial intelligence is used in many disciplines to identify objects within images, but these approaches usually require a considerable volume of images for training and validation. The objective of this study was to identify five different meat cuts from images and weights collected by a trained operator within the working environment of a commercial Irish beef plant. Individual cut images and weights from 7,987 meats cuts extracted from semimembranosus muscles (i.e., Topside muscle), post editing, were available. A variety of classical neural networks and a novel Ensemble machine learning approaches were then tasked with identifying each individual meat cut; performance of the approaches was dictated by accuracy (the percentage of correct predictions), precision (the ratio of correctly predicted objects relative to the number of objects identified as positive), and recall (also known as true positive rate or sensitivity). A novel Ensemble approach outperformed a selection of the classical neural networks including convolutional neural network and residual network. The accuracy, precision, and recall for the novel Ensemble method were 99.13%, 99.00%, and 98.00%, respectively, while that of the next best method were 98.00%, 98.00%, and 95.00%, respectively. The Ensemble approach, which requires relatively few gold-standard measures, can readily be deployed under normal abattoir conditions; the strategy could also be evaluated in the cuts from other primals or indeed other species.

Asunto(s)

RESUMEN

In unpredictable disaster scenarios, it is important to recognize the situation promptly and take appropriate response actions. This study proposes a cloud computing-based data collection, processing, and analysis process that employs a crowd-sensing application. Clustering algorithms are used to define the major damage types, and hotspot analysis is applied to effectively filter critical data from crowdsourced data. To verify the utility of the proposed process, it is applied to Icheon-si and Anseong-si, both in Gyeonggi-do, which were affected by heavy rainfall in 2020. The results show that the types of incident at the damaged site were effectively detected, and images reflecting the damage situation could be classified using the application of the geospatial analysis technique. For 5 August 2020, which was close to the date of the event, the images were classified with a precision of 100% at a threshold of 0.4. For 24-25 August 2020, the image classification precision exceeded 95% at a threshold of 0.5, except for the mudslide mudflow in the Yul area. The location distribution of the classified images showed a distribution similar to that of damaged regions in unmanned aerial vehicle images.

Asunto(s)

RESUMEN

Ipomoea cairica (L.) sweets are an invasive weed which has caused serious harm to the biodiversity and stability of the ecosystem. It is very important to accurately and rapidly identifying and monitoring Ipomoea cairica (L.) sweets in the wild for managements taking the necessary strategies to control the Ipomoea cairica (L.) sweets to rapidly grow in the wild. However, current approaches mainly depend on manual identification, which result in high cost and low efficiency. Satellite and manned aircraft are feasible assisting approaches, but the quality of the images collected by them is not well since the ground sampling resolution is low and cloud exists. In this study, we present a novel identifying and monitoring framework and method for Ipomoea cairica (L.) sweets based on unmanned aerial vehicle (UAV) and artificial intelligence (AI). In the proposed framework, we low-costly collected the images with 8256 × 5504 pixels of the monitoring area by the UAV and the collected images are split into more small sub-images with 224 × 224 pixels for identifying model. For identifying Ipomoea cairica (L.) sweets, we also proposed a novel deep convolutional neural network which includes 12 layers. Finally, the Ipomoea cairica (L.) sweets can be efficiently monitored by painting the area containing Ipomoea cairica (L.) sweets. In our experiments, we collected 100 raw images and generated 288000 samples, and made comparison with LeNet, AlexNet, GoogleNet, VGG and ResNet for validating our framework and model. The experimental results show the proposed method is excellent, the accuracy is 93.00% and the time cost is 7.439 s. The proposed method can achieve to an efficient balance between high accuracy and low complexity. Our method is more suitable for the identification of Ipomoea cairica (L.) sweets in the wild than other methods.

Asunto(s)

RESUMEN

BACKGROUND: The novel coronavirus disease 2019 (COVID-19) constitutes a public health emergency globally. The number of infected people and deaths are proliferating every day, which is putting tremendous pressure on our social and healthcare system. Rapid detection of COVID-19 cases is a significant step to fight against this virus as well as release pressure off the healthcare system. OBJECTIVE: One of the critical factors behind the rapid spread of COVID-19 pandemic is a lengthy clinical testing time. The imaging tool, such as Chest X-ray (CXR), can speed up the identification process. Therefore, our objective is to develop an automated CAD system for the detection of COVID-19 samples from healthy and pneumonia cases using CXR images. METHODS: Due to the scarcity of the COVID-19 benchmark dataset, we have employed deep transfer learning techniques, where we examined 15 different pre-trained CNN models to find the most suitable one for this task. RESULTS: A total of 860 images (260 COVID-19 cases, 300 healthy and 300 pneumonia cases) have been employed to investigate the performance of the proposed algorithm, where 70% images of each class are accepted for training, 15% is used for validation, and rest is for testing. It is observed that the VGG19 obtains the highest classification accuracy of 89.3% with an average precision, recall, and F1 score of 0.90, 0.89, 0.90, respectively. CONCLUSION: This study demonstrates the effectiveness of deep transfer learning techniques for the identification of COVID-19 cases using CXR images.

Asunto(s)

RESUMEN

Visual development during early childhood is a vital process. Examining the visual acuity of children is essential for early detection of visual abnormalities, but performing visual examination in children is challenging. Here, we developed a human-in-the-loop deep learning (DL) paradigm that combines traditional vision examination and DL with integration of software and hardware, thus facilitating the execution of vision examinations, offsetting the shortcomings of human doctors, and improving the abilities of both DL and doctors to evaluate the vision of children. Because this paradigm contains two rounds (a human round and DL round), doctors can learn from DL and the two can mutually supervise each other such that the precision of the DL system in evaluating the visual acuity of children is improved. Based on DL-based object localization and image identification, the experiences of doctors and the videos captured in the first round, the DL system in the second round can simulate doctors in evaluating the visual acuity of children with a final accuracy of 75.54%. For comparison, we also assessed an automatic deep learning method that did not consider the experiences of doctors, but its performance was not satisfactory. This entire paradigm can evaluate the visual acuity of children more accurately than humans alone. Furthermore, the paradigm facilitates automatic evaluation of the vision of children with a wearable device.

Asunto(s)