Automatic Screening Research Articles

Towards data-efficient mechanical design of bicontinuous composites using generative AI

The distribution of material phases is crucial to determine the composite's mechanical property. While the full structure-mechanics relationship of highly ordered material distributions can be studied with finite number of cases, this relationship is difficult to be revealed for complex irregular distributions, preventing design of such material structures to meet certain mechanical requirements. The noticeable developments of artificial intelligence (AI) algorithms in material design enables to detect the hidden structure-mechanics correlations which is essential for designing composite of complex structures. It is intriguing how these tools can assist composite design. Here, we focus on the rapid generation of bicontinuous composite structures together with the stress distribution in loading. We find that generative AI, enabled through fine-tuned Low Rank Adaptation models, can be trained with a few inputs to generate both synthetic composite structures and the corresponding von Mises stress distribution. The results show that this technique is convenient in generating massive composites designs with useful mechanical information that dictate stiffness, fracture and robustness of the material with one model, and such has to be done by several different experimental or simulation tests. This research offers valuable insights for the improvement of composite design with the goal of expanding the design space and automatic screening of composite designs for improved mechanical functions.

Towards an automatic early screening system for autism spectrum disorder in toddlers based on eye-tracking

AbstractAccording to official estimations, autism spectrum disorder (ASD) affects around 1% of European newborns. The high level of dependency of ASD-affected subjects entails an extremely high social and economic cost. However, early intervention can drastically improve children’s development and thus reduce their dependency. One of the main common characteristics of subjects with ASD is difficulties with social interaction, which determines how they react to certain stimuli. This behavior can be automatically detected by analyzing their gaze. This study explores and evaluates the feasibility of automatic screening for ASD in toddlers under 24 months of age based on this specific behavior. We applied a matched pairs experimental design and a set of test videos, using a set of variables extracted from gaze analysis from toddlers using eye-tracking devices. The different videos try to capture social engagement, social information gathering gaze exchanges, and gaze following. We used the data to make a thorough comparison of machine learning algorithms (nine learning schemes), including some that were used in related prior research, and others that are popular in classification problems. The results show that several of the tested algorithms provided notable performance.

MSA-GCN: A Multi-information Selection Aggregation Graph Convolutional Network for Breast Tumor Grading.

Physicians typically combine multi-modal data to make a graded diagnosis of breast tumors. However, most existing breast tumor grading methods rely solely on image information, resulting in limited accuracy in grading. This paper proposes a Multi-information Selection Aggregation Graph Convolutional Networks (MSA-GCN) for breast tumor grading. Firstly, to fully utilize phenotypic data reflecting the clinical and pathological characteristics of tumors, an automatic combination screening and weight encoder is proposed for phenotypic data, which can construct a population graph with improved structural information. Then, a graph structure is designed through similarity learning to reflect the correlation between patient image features. Finally, a multi-information selection aggregation mechanism is employed in the graph convolution model to extract the effective features of multi-modal data and enhance the classification performance of the model. The proposed method is evaluated on different clinical datasets from the Digital Database for Screening Mammography (DDSM) and INbreast. The average classification accuracies are 90.74% and 85.35%, respectively, surpassing the performance of existing methods. In conclusion, our method effectively fuses image and non-image information, leading to a significant improvement in the accuracy of breast tumor grading.

Developing and validating clinical models to identify candidates for allergic rhinitis pre-exposure prophylaxis

Purpose Few risk-forecasting models of allergic rhinitis (AR) exist that may aid AR pre-exposure prophylaxis (PrEP) in clinical practice. Therefore, this study aimed to develop and validate an effective clinical model for identifying candidates for AR PrEP using a routine medical questionnaire. Methods This study was conducted in 10 Chinese provinces with 13 medical centers (n = 877) between 2019 and 2021. Clinical characteristics and exposure history were collected via face-to-face interviews. Well-trained physicians diagnosed patients with AR based on skin prick test results and clinical performance. The least absolute shrinkage and selection operator model was used to identify potential risk factors for AR, and the logistic regression model was used to construct the risk-forecasting model. Predictive power and model reliability were assessed using area under the receiver operating characteristic curve and calibration curves, respectively. Results This study diagnosed 625 patients with AR who had positive responses to at least one indoor or outdoor allergen and 460 to at least one outdoor pollen allergen. Two nomograms were established to identify two types of AR with various sensitization patterns. Both models had an area under curve of approximately 0.7 in the development and internal validation datasets. Additionally, our findings found good agreement for the calibration curves of both models. Conclusion Early identification of candidates for AR PrEP using routine medical information may improve the deployment of limited resources and effective health management. Our models showed good performance in predicting AR; therefore, they can serve as potential automatic screening tools to identify AR PrEP candidates.

Histological Image-based Ensemble Model to Identify Myenteric Plexitis and Predict Endoscopic Postoperative Recurrence in Crohn's Disease: A Multicentre, Retrospective Study.

Myenteric plexitis is correlated with postoperative recurrence of Crohn's disease when relying on traditional statistical methods. However, comprehensive assessment of myenteric plexus remains challenging. This study aimed to develop and validate a deep learning system to predict postoperative recurrence through automatic screening and identification of features of the muscular layer and myenteric plexus. We retrospectively reviewed 205 patients who underwent bowel resection surgery from two hospitals. Patients were divided into a training cohort [n = 108], an internal validation cohort [n = 47], and an external validation cohort [n = 50]. A total of 190 960 patches from 278 whole-slide images of surgical specimens were analysed using the ResNet50 encoder, and 6144 features were extracted after transfer learning. We used five robust algorithms to construct classification models. The performances of the models were evaluated based on the area under the receiver operating characteristic curve [AUC] in three cohorts. The stacking model achieved satisfactory accuracy in predicting postoperative recurrence of CD in the training cohort (AUC: 0.980; 95% confidence interval [CI] 0.960-0.999), internal validation cohort [AUC: 0.908; 95% CI 0.823-0.992], and external validation cohort [AUC: 0.868; 95% CI 0.761-0.975]. The accuracy for identifying the severity of myenteric plexitis was 0.833, 0.745, and 0.694 in the training, internal validation and external validation cohorts, respectively. Our work initially established an interpretable stacking model based on features of the muscular layer and myenteric plexus extracted from histological images to identify the severity of myenteric plexitis and predict postoperative recurrence of CD.

Image preprocessing‐based ensemble deep learning classification of diabetic retinopathy

AbstractDiabetic retinopathy (DR) can cause irreversible eye damage, even blindness. The prognosis improves with early diagnosis. According to the International Classification of Diabetic Retinopathy Severity Scale (ICDRSS), DR has five stages. Modern, cost‐effective techniques for automatic DR screening and staging of fundus images are based on deep learning (DL). To obtain higher classification accuracy, the combination of several diverse individual DL models into one ensemble could be used. A new approach to model diversity in an ensemble is proposed by manipulating the training input data involving original and four variants of preprocessed image datasets. There are publicly available datasets with labels for all five stages, but some contain poor‐quality images. In contrast, this algorithm was trained on images from a six‐class DDR dataset, including the class of poor‐quality ungradable images, to enhance the classification performance. The solution was evaluated on the APTOS dataset, containing only ICDRSS classes. Classification results of the ensemble model were presented on two different ensemble convolutional neural network (CNN) models, based on Xception and EfficientNetB4 architectures using two fusion approaches. Our proposed ensemble models outperformed all other single deep learning architectures regarding overall accuracy and Cohen's Kappa, with the best results using the EfficientNetB4 architecture.

Towards a Corpus (and Language)-Independent Screening of Parkinson's Disease from Voice and Speech through Domain Adaptation.

End-to-end deep learning models have shown promising results for the automatic screening of Parkinson's disease by voice and speech. However, these models often suffer degradation in their performance when applied to scenarios involving multiple corpora. In addition, they also show corpus-dependent clusterings. These facts indicate a lack of generalisation or the presence of certain shortcuts in the decision, and also suggest the need for developing new corpus-independent models. In this respect, this work explores the use of domain adversarial training as a viable strategy to develop models that retain their discriminative capacity to detect Parkinson's disease across diverse datasets. The paper presents three deep learning architectures and their domain adversarial counterparts. The models were evaluated with sustained vowels and diadochokinetic recordings extracted from four corpora with different demographics, dialects or languages, and recording conditions. The results showed that the space distribution of the embedding features extracted by the domain adversarial networks exhibits a higher intra-class cohesion. This behaviour is supported by a decrease in the variability and inter-domain divergence computed within each class. The findings suggest that domain adversarial networks are able to learn the common characteristics present in Parkinsonian voice and speech, which are supposed to be corpus, and consequently, language independent. Overall, this effort provides evidence that domain adaptation techniques refine the existing end-to-end deep learning approaches for Parkinson's disease detection from voice and speech, achieving more generalizable models.

Dromedary camel's welfare: literature from 1980 to 2023 with a text mining and topic analysis approach.

Dromedary camels are the preferable livestock species in the arid and semi-arid regions of the world. Most of the world's camel populations are managed under a subsistence/extensive system maintained by migratory pastoralists but intensification is getting more frequent. Even though recently the welfare of camels has been receiving more attention, in many countries there are no regulations to protect their health and welfare. The objectives of this article were to explore the main research topics related to camel welfare, their distribution over time and to highlight research gaps. A literature search was performed to identify records published in English from January 1980 to March 2023 on Dromedary camel welfare via Scopus®, using "Camel welfare," "Camel behaviour," "She-camel" and "Camel management" as search words. A total of 234 records were retained for analysis after automatic and manual screening procedures. Descriptive statistics, text mining (TM) and topic analysis (TA) were performed. The result shows that even though there were fluctuations between years, records on camel welfare have increased exponentially over time. Asia was the region where most of the corresponding authors were located. The first five most frequent words were, "milk," "calv," "behaviour," "femal," and "breed," the least frequent word was "stabl." TA resulted in the five most relevant topics dealing with "Calf management and milk production," "Camel health and management system," "Female and male reproduction," "Camel behaviour and feeding," and "Camel welfare." The topics that contained the oldest records were "female and male reproduction" and "camel health and management system" (in 1980 and 1983, respectively), while the topic named "camel behaviour and feeding" had the first article published in 2000. Overall, even though topics related to camel behaviour and welfare are receiving more attention from academia, research is still needed to fully understand how to safeguard welfare in Dromedary camels.

Fully automatic screening for atrial fibrillation using a photoplethysmography based wristband and a novel smart infrastructure

Abstract Background Atrial Fibrillation (AF) is the leading cause of stroke, and its prevalence is expected to rise. While early detection is crucial in preventing thromboembolic events, its often asymptomatic and paroxysmal nature makes it challenging to capture using conventional electrocardiography (ECG). Photoplethysmography (PPG) enabled devices, such as smartwatches, allow continuous monitoring of the heart rhythm and accurate screening for AF. Purpose In this study, we aimed to build an IT-infrastructure for fully automated screening of AF in hospitalised patients at risk using a PPG-based wrist worn wearable device. Methods We conducted an investigator-initiated, prospective, clinical trial to evaluate the performance of a PPG-based wristband coupled to an automatic data process device hub in hospitalised patients. Participants with a CHA2DS2-VASc score of two or higher and no prior history of AF were eligible for the study. Consenting participants received a wristband with a built-in PPG sensor to be worn during the entire hospital stay. Data from the wristband was automatically transmitted, stored, analysed and visualised. When absolute arrhythmia (AA) was detected in the PPG signal, a subsequent seven-day Holter ECG was performed to confirm the diagnosis. Results A total of 346 patients were included in the analysis (age 72 ± 10 years; 175 females (50.6%); mean CHA2DS2-VASc score of 3.5 ± 1.3). No data was lost in the process. The mean monitoring time was 4.3 ± 4.4 days, 45.9% was evaluable. In twelve patients (3.5%, CI: 1.5% - 5.4%) absolute arrhythmia (AA) was detected in the PPG-signal. The mean screening time until first detection of AA was 14.7 ± 16.8 hours. The highest detection rate was observed within the first 24 hours of screening (p = 0.004) (Figure 1). The AA burden per hour was 1.9 times greater during the night (22:00 - 06:00) compared to the day (p = 0.003) (Figure 2). Of the 346 patients, 304 (87.9%) were compliant in wearing the wristband for the entire hospitalisation. In nine patients a Holter ECG was recorded, no AF was identified. Conclusions We successfully established an infrastructure for automated AF screening in hospitalised patients using wrist-worn PPG devices, enabling real-time visualisation of incoming data and facilitating therapeutic interventions with minimal delay. To our knowledge this is the first hospital to implement a fully automated AF screening infrastructure based on PPG devices. In a population at risk, AA was identified in 3.5% of all patients. The high compliance rate of the wearable device further supports its feasibility and utility in clinical practice.Figure 1:Time to first detection of AAFigure 2:Total AA burden per hour

A high-efficiency integrated strategy for the targeted discovery of novel barrigenol-type triterpenoid saponins from the shell of Xanthoceras sorbifolium Bunge by offline two-dimensional chromatography-Orbitrap mass spectrometry, neutral loss data acquisition, and predicted natural product screening

Natural products are considered as potential sources of leading compounds and play an important role in drug discovery. The liquid chromatography-mass spectrometry (LC-MS) technique is a powerful tool for compound-guided isolation from natural products. In this study, a high-efficiency integrated strategy was adopted to improve the new leading compounds discovery, including offline two-dimensional (2D) LC to extend the peak capacities, target neutral loss (NL) data-dependent acquisition (DDA) for barrigenol-type triterpenoids saponins and automatic screening through predicted natural product screening (PNPS) in TraceFinder. To validate the integrated strategy, the shell of Xanthoceras sorbifolium Bunge (XSB) was taken as a case. An offline 2D-LC system was constructed with hydrophilic interaction chromatography (HILIC) and reversed phase (RP) C18 column, and orthogonality of 0.66 and peak capacity of 3494. The 2D-LC system improved chromatographic baseline separation and peak resolution. Full MS/all ion fragmentation (AIF)/NL dd-MS2 DDA was employed for the detection of the barrigenol saponins. PNPS strategy was adopted and markedly extended the screening coverage. The combined strategy showed about 5 times improvement in the screening capability. The PNPS screening process, using TraceFinder software, discovered a total of 752 barrigenol saponins from the shell of XSB, including 707 potentially new barrigenol saponins, accounting for 94.02%. The feasibility of the strategy was also confirmed by the isolation of two novel barrigenol saponins, the structures of which were unambiguously identified using nuclear magnetic resonance (NMR). Furthermore, this strategy could also be applied to rapidly discover new bioactive constituents from other herbal medicines or other natural sources, especially the barrigenol saponins constituents.

DeepSkillNER: An automatic screening and ranking of resumes using hybrid deep learning and enhanced spectral clustering approach

DeepSkillNER: An automatic screening and ranking of resumes using hybrid deep learning and enhanced spectral clustering approach

Improving cardiovascular risk prediction through machine learning modelling of irregularly repeated electronic health records.

Existing electronic health records (EHRs) often consist of abundant but irregular longitudinal measurements of risk factors. In this study, we aim to leverage such data to improve the risk prediction of atherosclerotic cardiovascular disease (ASCVD) by applying machine learning (ML) algorithms, which can allow automatic screening of the population. A total of 215 744 Chinese adults aged between 40 and 79 without a history of cardiovascular disease were included (6081 cases) from an EHR-based longitudinal cohort study. To allow interpretability of the model, the predictors of demographic characteristics, medication treatment, and repeatedly measured records of lipids, glycaemia, obesity, blood pressure, and renal function were used. The primary outcome was ASCVD, defined as non-fatal acute myocardial infarction, coronary heart disease death, or fatal and non-fatal stroke. The eXtreme Gradient boosting (XGBoost) algorithm and Least Absolute Shrinkage and Selection Operator (LASSO) regression models were derived to predict the 5-year ASCVD risk. In the validation set, compared with the refitted Chinese guideline-recommended Cox model (i.e. the China-PAR), the XGBoost model had a significantly higher C-statistic of 0.792, (the differences in the C-statistics: 0.011, 0.006-0.017, P < 0.001), with similar results reported for LASSO regression (the differences in the C-statistics: 0.008, 0.005-0.011, P < 0.001). The XGBoost model demonstrated the best calibration performance (men: Dx = 0.598, P = 0.75; women: Dx = 1.867, P = 0.08). Moreover, the risk distribution of the ML algorithms differed from that of the conventional model. The net reclassification improvement rates of XGBoost and LASSO over the Cox model were 3.9% (1.4-6.4%) and 2.8% (0.7-4.9%), respectively. Machine learning algorithms with irregular, repeated real-world data could improve cardiovascular risk prediction. They demonstrated significantly better performance for reclassification to identify the high-risk population correctly.

Segmentation and thickness calculation of glomerular basement membrane using RADS-Net in glomerular microscopic images

Glomerular ultrastructure needs to be observed under transmission electron microscopy during pathological diagnosis of renal biopsies. Evaluating the glomerular basement membranes (GBM) is necessary for the renal pathological diagnosis. There is little work that performs well for full-flow automated analysis of GBM in whole-glomerular electron microscopy images, including GBM segmentation, location, and thickness calculation. To address this problem, we propose a network architecture, RADS-Net, whose segmentation module combines the advantages of ViT and CNN to achieve better performance in GBM and electron dense deposit (EDD) segmentation tasks. It also contains an uncertainty evaluation module and a location module to exclude unsuitable regions for measurement and extract accurate GBM contours. We propose a GBM thickness calculation method based on statistical learning method, skeletonization algorithm, and improved region growth method. Systematic experiments demonstrate that our proposed segmentation network achieves superior accuracy compared to other competing methods. Additionally, the computed GBM thickness, after an automatic screening of suitable fields of view, meets the requirements for pathology evaluation.

GLCM-Based FBLS: A Novel Broad Learning System for Knee Osteopenia and Osteoprosis Screening in Athletes

Due to the physical strain experienced during intense workouts, athletes are at a heightened risk of developing osteopenia and osteoporosis. These conditions not only impact their overall health but also their athletic performance. The current clinical screening methods for osteoporosis are limited by their high radiation dose, complex post-processing requirements, and the significant time and resources needed for implementation. This makes it challenging to incorporate them into athletes’ daily training routines. Consequently, our objective was to develop an innovative automated screening approach for detecting osteopenia and osteoporosis using X-ray image data. Although several automated screening methods based on deep learning have achieved notable results, they often suffer from overfitting and inadequate datasets. To address these limitations, we proposed a novel model called the GLCM-based fuzzy broad learning system (GLCM-based FBLS). Initially, texture features of X-ray images were extracted using the gray-level co-occurrence matrix (GLCM). Subsequently, these features were combined with the fuzzy broad learning system to extract crucial information and enhance the accuracy of predicting osteoporotic conditions. Finally, we applied the proposed method to the field of osteopenia and osteoporosis screening. By comparing this model with three advanced deep learning models, we have verified the effectiveness of GLCM-based FBLS in the automatic screening of osteoporosis for athletes.

Discriminative ensemble meta-learning with co-regularization for rare fundus diseases diagnosis.

Deep neural networks (DNNs) have been widely applied in the medical image community, contributing to automatic ophthalmic screening systems for some common diseases. However, the incidence of fundus diseases patterns exhibits a typical long-tailed distribution. In clinic, a small number of common fundus diseases have sufficient observed cases for large-scale analysis while most of the fundus diseases are infrequent. For these rare diseases with extremely low-data regimes, it is challenging to train DNNs to realize automatic diagnosis. In this work, we develop an automatic diagnosis system for rare fundus diseases, based on the meta-learning framework. The system incorporates a co-regularization loss and the ensemble-learning strategy into the meta-learning framework, fully leveraging the advantage of multi-scale hierarchical feature embedding. We initially conduct comparative experiments on our newly-constructed lightweight multi-disease fundus images dataset for the few-shot recognition task (namely, FundusData-FS). Moreover, we verify the cross-domain transferability from miniImageNet to FundusData-FS, and further confirm our method's good repeatability. Rigorous experiments demonstrate that our method can detect rare fundus diseases, and is superior to the state-of-the-art methods. These investigations demonstrate that the potential of our method for the real clinical practice is promising.

Clinical Acceptability of Artificial Intelligence-Screened Interstitial Lung Disease (AI-ILD) in Lung Cancer Patients Treated with Radiotherapy

Patients with interstitial lung disease (ILD) treated with thoracic radiotherapy (RT) are at greater risk of pulmonary toxicity. Automatic universal screening for ILD allows radiation oncologists (ROs) to risk stratify patients and implement necessary modifications to their respiratory monitoring or treatment. Automatic screening however may affect RO workload and so it is imperative to assess the clinical acceptability of this tool. We have developed a machine learning algorithm to identify patients who are at high risk of having ILD based on RT planning computed tomography (CT) images. A quality improvement (QI) project was initiated to test feasibility and acceptability of the machine learning algorithm. If positive, the results of the machine learning algorithm were made available to ROs via structured electronic reporting. ROs were prompted to review the patient and consider expert radiologist consultation if thought appropriate. All electronic surveys and qualitative comments were summarized to describe clinical acceptability. Expert radiologist established gold standard ILD status of all patients on the study. A formal review of RO feedback was collected for all screen-positive, true-positive cases. Two hundred forty cases were screened of which 45 were flagged as AI-ILD positive and the responsible RO notified. Of these 45 screen-positive cases, all continued on to RT except for 3 patients with tumor progression. From these 45, 24 surveys were completed, 21 had no prior suspicion of ILD. There were 7 true-positives, of which 1 had a survey response. Based on the survey responses, 88% of cases underwent review by the responsible RO. In 16 cases this automatic notification prompted case consultation with an expert radiologist. Expert review was performed from 10 minutes up to 53 hours after the email prompt to the radiologist, with median response time of 1.5 hours. In the 7 screen-positive, true-positive cases, only 2 were not previously known to the responsible RO. In the two cases where true-positive ILD status was previously unknown, one was a mild case of ILD and the other had previously received thoracic RT at this institution without ILD being identified, in both cases the ROs were grateful that this diagnosis was identified prior to treatment. RO confidence in the machine learning prediction was moderate due to the high proportion of false positives. Based on available survey results, more than 75% of the screen-positive cases were reviewed by the responsible RO and two-thirds of these involved expert radiology input. RO feedback was generally positive and this tool was rated as a net benefit despite the high rate of false-positives and the need for clarification.

A molecular networking-assisted automatic database screening strategy for comprehensive annotation of small molecules in complex matrices

Liquid chromatography-tandem with high-resolution mass spectrometry (LCHRMS) has proven challenging for annotating multiple small molecules within complex matrices due to the complexities of chemical structure and raw LCHRMS data, as well as limitations in previous literatures and reference spectra related to those molecules. In this study, we developed a molecular networking assisted automatic database screening (MN/auto-DBS) strategy to examine the combined effect of MS1 exact mass screening and MS2 similarity analysis. We compiled all previously reported compounds from the relevant literatures. With the development of a Python software, the in-house database (DB) was created by automatically calculating the m/z and data from experimental MS1 hits were rapid screened with DB. We then performed a feature-based molecular network analysis on the auto-MS2 data for supplementary identification of unreported compounds, including clustered FBMN and annotated GNPS compounds. Finally, the results from both strategies were merged and manually curated for correct structural assignment. To demonstrate the applicability of MN/auto-DBS, we selected the Huangqi-Danshen herb pair (HD), commonly used in prescriptions or patent medicines to treat diabetic nephropathy and cerebrovascular disease. A total of 223 compounds were annotated, including 65 molecules not previously reported in HD, such as aromatic polyketides, coumarins, and diarylheptanoids. Using MN/auto-DBS, we can profile and mine a wide range of complex matrices for potentially new compounds.

A self-supervised anomaly detection algorithm with interpretability

Identifying the abnormal samples from a data set and determining their type are two key tasks of anomaly detection. However, the existing anomaly detection algorithms are generally faced with the defects of weak generalization ability and insufficient interpretation, the core reason of which is that they cannot mine specific features for different abnormal types. In this paper, a new anomaly detection algorithm aiming at feature selection for different abnormal types is developed. Inspired by self-supervised learning, we take the stationarity of variance changes of abnormal score similarity as a pretext task and combine it with wrapped search method. Then, the features and the corresponding parameters for different abnormal types can be screened to apply to the downstream task of anomaly integration detection. To verify the efficiency of the new algorithm, we conduct two sets of experiments to compare the new algorithm with 11 classical anomaly detection and 3 clustering anomaly detection algorithms on the data sets WDBC, WPBC and Wilt from DAMI database with the evaluation measures P@n, Adj-P@n, AP, Adj-AP and AUC. The experiment results show that, both in the identification and classification on abnormal samples, all performance measures of the new algorithm are explicitly better than that of the contrastive algorithms. Also, we apply the new algorithm to the Chinese auto insurance market, and find that the results can help managers to identify the main patterns of fraudulent claims and to summarize the feature combinations of fraud behaviors. In general, the new algorithm developed in this paper has the following advantages compared with traditional algorithms: 1) It can directly capture abnormal features and realize effective recognition of abnormal types, which effectively bridge the gap between abnormal judgement and feature screening. 2) Automatic screening of abnormal features can be completed under the condition of self-updating learning optimal strategy. 3) Only a few features are extracted from all features to reveal the abnormal characteristics, which significantly improves the interpretability and generalization ability of the algorithm and its results. In a word, through the novel self-supervised design method, feature screening is skillfully integrated into the anomaly detection task, which may provide a new way for anomaly detection research.

Automatic Diagnosis of Different Types of Retinal Vein Occlusion Based on Fundus Images

Retinal vein occlusion (RVO) is the second common cause of blindness following diabetic retinopathy. The manual screening of fundus images to detect RVO is time consuming. Deep-learning techniques have been used for screening RVO due to their outstanding performance in many applications. However, unlike other images, medical images have smaller lesions, which require a more elaborate approach. To provide patients with an accurate diagnosis, followed by timely and effective treatment, we developed an intelligent method for automatic RVO screening on fundus images. Swin Transformer learns the hierarchy of low-to high-level features like the convolutional neural network. However, Swin Transformer extracts features from fundus images through attention modules, which pay more attention to the interrelationship between the features and each other. The model is more universal, does not rely entirely on the data itself, and focuses not only on local information but has a diffusion mechanism from local to global. To suppress overfitting, we adopt a regularization strategy, label smoothing, which uses one-hot to add noise to reduce the weight of the categories of true sample labels when calculating the loss function. The choice of different models using a 5-fold cross-validation on our own datasets indicates that Swin Transformer performs better. The accuracy of classifying all datasets is 98.75 ± 0.000, and the accuracy of identifying MRVO, CRVO, BRVO, and normal, using the method proposed in the paper, is 94.49 ± 0.094, 99.98 ± 0.015, 98.88 ± 0.08, and 99.42 ± 0.012, respectively. The method will be useful to diagnose RVO and help decide grade through fundus images, which has the potency to provide patients with further diagnosis and treatment.

Automatic Screening System to Distinguish Benign/Malignant Breast-Cancer Histology Images Using Optimized Deep and Handcrafted Features

Breast Cancer (BC) has been increasing in incidence among women for a variety of reasons, and prompt detection and management are essential to reducing mortality rates. In the context of clinical-level breast cancer screening, the needle biopsy sample is used to generate Breast Histology Images (BHIs), which will then be used to confirm the results. Using a novel Deep-Learning Plan (DLP), the proposed work identifies BHI accurately and confirms the severity of BC by confirming its severity. As part of the proposed DLP implementation, four phases are involved: (i) the collection and enhancement of images, (ii) the extraction of features, (iii) the reduction of features and their integration, and (iv) binary classification and validation. The purpose of this study is to optimize deep features and machine features using particle swarm algorithms. To evaluate the performance of the proposed scheme, we compare the results obtained using individual deep features, dual deep features, and hybrid features. Using the hybrid image features in the classifier, this study has determined that ResNet18 with k-nearest neighbor provides superior classification accuracy (> 94%).