Cardiovasc Diabetol. 2024 Aug 10;23(1):296. doi: 10.1186/s12933-024-02367-z.

ABSTRACT

BACKGROUND: Cardiac autonomic neuropathy (CAN) in diabetes mellitus (DM) is independently associated with cardiovascular (CV) events and CV death. Diagnosis of this complication of DM is time-consuming and not routinely performed in the clinical practice, in contrast to fundus retinal imaging which is accessible and routinely performed. Whether artificial intelligence (AI) utilizing retinal images collected through diabetic eye screening can provide an efficient diagnostic method for CAN is unknown.

METHODS: This was a single center, observational study in a cohort of patients with DM as a part of the Cardiovascular Disease in Patients with Diabetes: The Silesia Diabetes-Heart Project (NCT05626413). To diagnose CAN, we used standard CV autonomic reflex tests. In this analysis we implemented AI-based deep learning techniques with non-mydriatic 5-field color fundus imaging to identify patients with CAN. Two experiments have been developed utilizing Multiple Instance Learning and primarily ResNet 18 as the backbone network. Models underwent training and validation prior to testing on an unseen image set.

RESULTS: In an analysis of 2275 retinal images from 229 patients, the ResNet 18 backbone model demonstrated robust diagnostic capabilities in the binary classification of CAN, correctly identifying 93% of CAN cases and 89% of non-CAN cases within the test set. The model achieved an area under the receiver operating characteristic curve (AUCROC) of 0.87 (95% CI 0.74-0.97). For distinguishing between definite or severe stages of CAN (dsCAN), the ResNet 18 model accurately classified 78% of dsCAN cases and 93% of cases without dsCAN, with an AUCROC of 0.94 (95% CI 0.86-1.00). An alternate backbone model, ResWide 50, showed enhanced sensitivity at 89% for dsCAN, but with a marginally lower AUCROC of 0.91 (95% CI 0.73-1.00).

CONCLUSIONS: AI-based algorithms utilising retinal images can differentiate with high accuracy patients with CAN. AI analysis of fundus images to detect CAN may be implemented in routine clinical practice to identify patients at the highest CV risk.

TRIAL REGISTRATION: This is a part of the Silesia Diabetes-Heart Project (Clinical-Trials.gov Identifier: NCT05626413).

PMID:39127709 | DOI:10.1186/s12933-024-02367-z

J Affect Disord. 2024 Aug 8:S0165-0327(24)01215-1. doi: 10.1016/j.jad.2024.08.002. Online ahead of print.

ABSTRACT

BACKGROUND AND PURPOSE: Diagnosis of depression is based on tests performed by psychiatrists and information provided by patients or their relatives. In the field of machine learning (ML), numerous models have been devised to detect depression automatically through the analysis of speech audio signals. While deep learning approaches often achieve superior classification accuracy, they are notably resource-intensive. This research introduces an innovative, multilevel hybrid feature extraction-based classification model, specifically designed for depression detection, which exhibits reduced time complexity.

MATERIALS AND METHODS: MODMA dataset consisting of 29 healthy and 23 Major depressive disorder audio signals was used. The constructed model architecture integrates multilevel hybrid feature extraction, iterative feature selection, and classification processes. During the Hybrid Handcrafted Feature (HHF) generation stage, a combination of textural and statistical methods was employed to extract low-level features from speech audio signals. To enhance this process for high-level feature creation, a Multilevel Discrete Wavelet Transform (MDWT) was applied. This technique produced wavelet subbands, which were then input into the hybrid feature extractor, enabling the extraction of both high and low-level features. For the selection of the most pertinent features from these extracted vectors, Iterative Neighborhood Component Analysis (INCA) was utilized. Finally, in the classification phase, a one-dimensional nearest neighbor classifier, augmented with ten-fold cross-validation, was implemented to achieve detailed, results.

RESULTS: The HHF-based speech audio signal classification model attained excellent performance, with the 94.63 % classification accuracy.

CONCLUSIONS: The findings validate the remarkable proficiency of the introduced HHF-based model in depression classification, underscoring its computational efficiency.

PMID:39127304 | DOI:10.1016/j.jad.2024.08.002

J Cardiovasc Magn Reson. 2024 Aug 8:101081. doi: 10.1016/j.jocmr.2024.101081. Online ahead of print.

ABSTRACT

BACKGROUND: Time-resolved, three-dimensional phase-contrast magnetic resonance imaging (4D flow MRI) plays an important role in assessing cardiovascular diseases. However, the manual or semi-automatic segmentation of aortic vessel boundaries in 4D flow data introduces variability and limits reproducibility of aortic hemodynamics visualization and quantitative flow-related parameter computation. This paper explores the potential of deep learning to improve 4D flow MRI segmentation by developing models for automatic segmentation and analyzes the impact of the training data on the generalization of the model across different sites, scanner vendors, sequences, and pathologies.

METHODS: The study population consists of 260 4D flow MRI datasets, including subjects without known aortic pathology, healthy volunteers, and patients with bicuspid aortic valve (BAV) examined at different hospitals. The dataset was split to train segmentation models on subsets with different representations of characteristics such as pathology, gender, age, scanner model, vendor, and field strength. An enhanced 3D U-net convolutional neural network (CNN) architecture with residual units was trained for 2D+t aortic cross-sectional segmentation. The model performance was evaluated using Dice score, Hausdorff distance, and average symmetric surface distance on test data, datasets with characteristics not represented in the training set (model-specific), and an overall evaluation set. Standard diagnostic flow parameters were computed and compared with manual segmentation results using Bland-Altman analysis and interclass correlation.

RESULTS: The representation of technical factors such as scanner vendor and field strength in the training dataset had the strongest influence on the overall segmentation performance. Age had a greater impact than gender. Models solely trained on BAV patients' datasets performed well on datasets of healthy subjects but not vice versa.

CONCLUSION: This study highlights the importance of considering a heterogeneous dataset for the training of widely applicable automatic CNN segmentations in 4D flow MRI, with a particular focus on the inclusion of different pathologies and technical aspects of data acquisition.

PMID:39127260 | DOI:10.1016/j.jocmr.2024.101081

Sci Total Environ. 2024 Aug 8:175406. doi: 10.1016/j.scitotenv.2024.175406. Online ahead of print.

ABSTRACT

The Eddy Covariance (EC) method allows for monitoring carbon, water, and energy fluxes between Earth's surface and atmosphere. Due to its varying interdependent data streams and abundance of data as a whole, EC is naturally suited to Artificial Intelligence (AI) approaches. The integration of AI and EC will likely play a crucial role in the climate change mitigation and adaptation goals defined in the Sustainable Development Goals (SDGs) of the Agenda 2030. To aid this, we present a scoping review in which the novelty of various AI techniques in monitoring fluxes through the EC method from the past two decades has been collected. Overall, we find a clear positive trend in the quantity of research in this area, particularly in the last five years. We also find a lack of uniformity in available techniques, due to the diverse technologies and variables employed across environmental conditions and ecosystems. We highlight the most applied Machine Learning (ML) models, over the 71 algorithms identified in the scoping review, such as Random Forest (RF), Support Vector Machine (SVM), Artificial Neural Network (ANN), Support Vector Regression (SVR), and K-Nearest Neigbor (KNN). We suggest that future progress in this field requires an international, collaborative effort involving computer scientists and ecologists. Modern Deep Learning (DL) techniques such as Transformers and generative AI must be investigated to find how they may benefit our field. A forward-looking strategy must be formed for the optimal utilization of AI combined with EC to define future actions in flux monitoring in the face of climate change.

PMID:39127196 | DOI:10.1016/j.scitotenv.2024.175406

Biomed Phys Eng Express. 2024 Aug 10. doi: 10.1088/2057-1976/ad6dcd. Online ahead of print.

ABSTRACT

Target volumes for radiotherapy are usually contoured manually, which can be time-consuming and prone to inter- and intra-observer variability. Automatic contouring by convolutional neural networks (CNN) can be fast and consistent but may produce unrealistic contours or miss relevant structures. We evaluate approaches for increasing the quality and assessing the uncertainty of CNN-generated contours of head and neck cancers with PET/CT as input. 

Approach. Two patient cohorts with head and neck squamous cell carcinoma and baseline 18F-fluorodeoxyglucose positron emission tomography and computed tomography images (FDG-PET/CT) were collected retrospectively from two centers. The union of manual contours of the gross primary tumor and involved nodes was used to train CNN models for generating automatic contours. The impact of image preprocessing, image augmentation, transfer learning and CNN complexity, architecture, and dimension (2D or 3D) on model performance and generalizability across centers was evaluated. A Monte Carlo dropout technique was used to quantify and visualize the uncertainty of the automatic contours. 

Main results. CNN models provided contours with good overlap with the manually contoured ground truth (median Dice Similarity Coefficient: 0.75 - 0.77), consistent with reported inter-observer variations and previous auto-contouring studies. Image augmentation and model dimension, rather than model complexity, architecture, or advanced image preprocessing, had the largest impact on model performance and cross-center generalizability. Transfer learning on a limited number of patients from a separate center increased model generalizability without decreasing model performance on the original training cohort. High model uncertainty was associated with false positive and false negative voxels as well as low Dice coefficients.

Significance. High quality automatic contours can be obtained using deep learning architectures that are not overly complex. Uncertainty estimation of the predicted contours shows potential for highlighting regions of the contour requiring manual revision or flagging segmentations requiring manual inspection and intervention.&#xD.

PMID:39127060 | DOI:10.1088/2057-1976/ad6dcd

EJNMMI Res. 2024 Aug 10;14(1):72. doi: 10.1186/s13550-024-01128-z.

ABSTRACT

BACKGROUND: [18F]FDG PET denoising by SubtlePET™ using deep learning artificial intelligence (AI) was previously found to induce slight modifications in lesion and reference organs' quantification and in lesion detection. As a next step, we aimed to evaluate its clinical impact on [18F]FDG PET solid tumour treatment response assessments, while comparing "standard PET" to "AI denoised half-duration PET" ("AI PET") during follow-up.

RESULTS: 110 patients referred for baseline and follow-up standard digital [18F]FDG PET/CT were prospectively included. "Standard" EORTC and, if applicable, PERCIST response classifications by 2 readers between baseline standard PET1 and follow-up standard PET2 as a "gold standard" were compared to "mixed" classifications between standard PET1 and AI PET2 (group 1; n = 64), or between AI PET1 and standard PET2 (group 2; n = 46). Separate classifications were established using either standardized uptake values from ultra-high definition PET with or without AI denoising (simplified to "UHD") or EANM research limited v2 (EARL2)-compliant values (by Gaussian filtering in standard PET and using the same filter in AI PET). Overall, pooling both study groups, in 11/110 (10%) patients at least one EORTCUHD or EARL2 or PERCISTUHD or EARL2 mixed vs. standard classification was discordant, with 369/397 (93%) concordant classifications, unweighted Cohen's kappa = 0.86 (95% CI: 0.78-0.94). These modified mixed vs. standard classifications could have impacted management in 2% of patients.

CONCLUSIONS: Although comparing similar PET images is preferable for therapy response assessment, the comparison between a standard [18F]FDG PET and an AI denoised half-duration PET is feasible and seems clinically satisfactory.

PMID:39126532 | DOI:10.1186/s13550-024-01128-z

J Fish Biol. 2024 Aug 10. doi: 10.1111/jfb.15887. Online ahead of print.

ABSTRACT

Photographic identification (photo ID) is an established method that is used to count animals and track individuals' movements. This method performs well with some species of elasmobranchs (i.e., sharks, skates, and rays) where individuals have distinctive skin patterns. However, the unique skin patterns used for ID must be stable through time to allow re-identification of individuals in future sampling events. More recently, artificial intelligence (AI) models have substantially decreased the labor-intensive process of matching photos in extensive photo ID libraries and increased the reliability of photo ID. Here, photo ID and AI are used for the first time to identify epaulette sharks (Hemiscyllium ocellatum) at different life stages for approximately 2 years. An AI model was developed to assess and compare the reliability of human-classified ID patterns in juvenile and neonate sharks. The model also tested the persistence of unique patterns in adult sharks. Results indicate that immature life stages are unreliable for pattern identification, using both human and AI approaches, due to the plasticity of these subadult growth forms. Mature sharks maintain their patterns through time and can be identified by AI models with approximately 86% accuracy. The approach outlined in this study has the potential of validating the stability of ID patterns through time; however, testing on wild populations and long-term datasets is needed. This study's novel deep neural network development strategy offers a streamlined and accessible framework for generating a reliable model from a small data set, without requiring high-performance computing. Since many photo ID studies commence with limited datasets and resources, this AI model presents practical solutions to such constraints. Overall, this approach has the potential to address challenges associated with long-term photo ID data sets and the application of AI for shark identification.

PMID:39126281 | DOI:10.1111/jfb.15887

Int J Mol Sci. 2024 Aug 4;25(15):8502. doi: 10.3390/ijms25158502.

ABSTRACT

With the widespread adoption of next-generation sequencing technologies, the speed and convenience of genome sequencing have significantly improved, and many biological genomes have been sequenced. However, during the assembly of small genomes, we still face a series of challenges, including repetitive fragments, inverted repeats, low sequencing coverage, and the limitations of sequencing technologies. These challenges lead to unknown gaps in small genomes, hindering complete genome assembly. Although there are many existing assembly software options, they do not fully utilize the potential of artificial intelligence technologies, resulting in limited improvement in gap filling. Here, we propose a novel method, DLGapCloser, based on deep learning, aimed at assisting traditional tools in further filling gaps in small genomes. Firstly, we created four datasets based on the original genomes of Saccharomyces cerevisiae, Schizosaccharomyces pombe, Neurospora crassa, and Micromonas pusilla. To further extract effective information from the gene sequences, we also added homologous genomes to enrich the datasets. Secondly, we proposed the DGCNet model, which effectively extracts features and learns context from sequences flanking gaps. Addressing issues with early pruning and high memory usage in the Beam Search algorithm, we developed a new prediction algorithm, Wave-Beam Search. This algorithm alternates between expansion and contraction phases, enhancing efficiency and accuracy. Experimental results showed that the Wave-Beam Search algorithm improved the gap-filling performance of assembly tools by 7.35%, 28.57%, 42.85%, and 8.33% on the original results. Finally, we established new gap-filling standards and created and implemented a novel evaluation method. Validation on the genomes of Saccharomyces cerevisiae, Schizosaccharomyces pombe, Neurospora crassa, and Micromonas pusilla showed that DLGapCloser increased the number of filled gaps by 8.05%, 15.3%, 1.4%, and 7% compared to traditional assembly tools.

PMID:39126071 | DOI:10.3390/ijms25158502

Int J Mol Sci. 2024 Aug 1;25(15):8426. doi: 10.3390/ijms25158426.

ABSTRACT

Protein structure prediction is important for understanding their function and behavior. This review study presents a comprehensive review of the computational models used in predicting protein structure. It covers the progression from established protein modeling to state-of-the-art artificial intelligence (AI) frameworks. The paper will start with a brief introduction to protein structures, protein modeling, and AI. The section on established protein modeling will discuss homology modeling, ab initio modeling, and threading. The next section is deep learning-based models. It introduces some state-of-the-art AI models, such as AlphaFold (AlphaFold, AlphaFold2, AlphaFold3), RoseTTAFold, ProteinBERT, etc. This section also discusses how AI techniques have been integrated into established frameworks like Swiss-Model, Rosetta, and I-TASSER. The model performance is compared using the rankings of CASP14 (Critical Assessment of Structure Prediction) and CASP15. CASP16 is ongoing, and its results are not included in this review. Continuous Automated Model EvaluatiOn (CAMEO) complements the biennial CASP experiment. Template modeling score (TM-score), global distance test total score (GDT_TS), and Local Distance Difference Test (lDDT) score are discussed too. This paper then acknowledges the ongoing difficulties in predicting protein structure and emphasizes the necessity of additional searches like dynamic protein behavior, conformational changes, and protein-protein interactions. In the application section, this paper introduces some applications in various fields like drug design, industry, education, and novel protein development. In summary, this paper provides a comprehensive overview of the latest advancements in established protein modeling and deep learning-based models for protein structure predictions. It emphasizes the significant advancements achieved by AI and identifies potential areas for further investigation.

PMID:39125995 | DOI:10.3390/ijms25158426

Int J Mol Sci. 2024 Jul 28;25(15):8239. doi: 10.3390/ijms25158239.

ABSTRACT

Multifactorial diseases demand therapeutics that can modulate multiple targets for enhanced safety and efficacy, yet the clinical approval of multitarget drugs remains rare. The integration of machine learning (ML) and deep learning (DL) in drug discovery has revolutionized virtual screening. This study investigates the synergy between ML/DL methodologies, molecular representations, and data augmentation strategies. Notably, we found that SVM can match or even surpass the performance of state-of-the-art DL methods. However, conventional data augmentation often involves a trade-off between the true positive rate and false positive rate. To address this, we introduce Negative-Augmented PU-bagging (NAPU-bagging) SVM, a novel semi-supervised learning framework. By leveraging ensemble SVM classifiers trained on resampled bags containing positive, negative, and unlabeled data, our approach is capable of managing false positive rates while maintaining high recall rates. We applied this method to the identification of multitarget-directed ligands (MTDLs), where high recall rates are critical for compiling a list of interaction candidate compounds. Case studies demonstrate that NAPU-bagging SVM can identify structurally novel MTDL hits for ALK-EGFR with favorable docking scores and binding modes, as well as pan-agonists for dopamine receptors. The NAPU-bagging SVM methodology should serve as a promising avenue to virtual screening, especially for the discovery of MTDLs.

PMID:39125808 | DOI:10.3390/ijms25158239

Int J Mol Sci. 2024 Jul 26;25(15):8146. doi: 10.3390/ijms25158146.

ABSTRACT

As part of the field of DNA methylation identification, this study tackles the challenge of enhancing recognition performance by introducing a specialized deep learning framework called DeepPGD. DNA methylation, a crucial biological modification, plays a vital role in gene expression analyses, cellular differentiation, and the study of disease progression. However, accurately and efficiently identifying DNA methylation sites remains a pivotal concern in the field of bioinformatics. The issue addressed in this paper is the presence of methylation in DNA, which is a binary classification problem. To address this, our research aimed to develop a deep learning algorithm capable of more precisely identifying these sites. The DeepPGD framework combined a dual residual structure involving Temporal convolutional networks (TCNs) and bidirectional long short-term memory (BiLSTM) networks to effectively extract intricate DNA structural and sequence features. Additionally, to meet the practical requirements of DNA methylation identification, extensive experiments were conducted across a variety of biological species. The experimental results highlighted DeepPGD's exceptional performance across multiple evaluation metrics, including accuracy, Matthews' correlation coefficient (MCC), and the area under the curve (AUC). In comparison to other algorithms in the same domain, DeepPGD demonstrated superior classification and predictive capabilities across various biological species datasets. This significant advancement in algorithmic prowess not only offers substantial technical support, but also holds potential for research and practical implementation within the DNA methylation identification domain. Moreover, the DeepPGD framework shows potential for application in genomics research, biomedicine, and disease diagnostics, among other fields.

PMID:39125714 | DOI:10.3390/ijms25158146

Diagnostics (Basel). 2024 Aug 5;14(15):1691. doi: 10.3390/diagnostics14151691.

ABSTRACT

Breast cancer is a prevalent malignancy characterized by the uncontrolled growth of glandular epithelial cells, which can metastasize through the blood and lymphatic systems. Microcalcifications, small calcium deposits within breast tissue, are critical markers for early detection of breast cancer, especially in non-palpable carcinomas. These microcalcifications, appearing as small white spots on mammograms, are challenging to identify due to potential confusion with other tissues. This study hypothesizes that a hybrid feature extraction approach combined with Convolutional Neural Networks (CNNs) can significantly enhance the detection and localization of microcalcifications in mammograms. The proposed algorithm employs Gabor, Prewitt, and Gray Level Co-occurrence Matrix (GLCM) kernels for feature extraction. These features are input to a CNN architecture designed with maxpooling layers, Rectified Linear Unit (ReLU) activation functions, and a sigmoid response for binary classification. Additionally, the Top Hat filter is used for precise localization of microcalcifications. The preprocessing stage includes enhancing contrast using the Volume of Interest Look-Up Table (VOI LUT) technique and segmenting regions of interest. The CNN architecture comprises three convolutional layers, three ReLU layers, and three maxpooling layers. The training was conducted using a balanced dataset of digital mammograms, with the Adam optimizer and binary cross-entropy loss function. Our method achieved an accuracy of 89.56%, a sensitivity of 82.14%, and a specificity of 91.47%, outperforming related works, which typically report accuracies around 85-87% and sensitivities between 76 and 81%. These results underscore the potential of combining traditional feature extraction techniques with deep learning models to improve the detection and localization of microcalcifications. This system may serve as an auxiliary tool for radiologists, enhancing early detection capabilities and potentially reducing diagnostic errors in mass screening programs.

PMID:39125567 | DOI:10.3390/diagnostics14151691

Diagnostics (Basel). 2024 Aug 2;14(15):1677. doi: 10.3390/diagnostics14151677.

ABSTRACT

In this work, several machine learning (ML) algorithms, both classical ML and modern deep learning, were investigated for their ability to improve the performance of a pipeline for the segmentation and classification of prostate lesions using MRI data. The algorithms were used to perform a binary classification of benign and malignant tissue visible in MRI sequences. The model choices include support vector machines (SVMs), random decision forests (RDFs), and multi-layer perceptrons (MLPs), along with radiomic features that are reduced by applying PCA or mRMR feature selection. Modern CNN-based architectures, such as ConvNeXt, ConvNet, and ResNet, were also evaluated in various setups, including transfer learning. To optimize the performance, different approaches were compared and applied to whole images, as well as gland, peripheral zone (PZ), and lesion segmentations. The contribution of this study is an investigation of several ML approaches regarding their performance in prostate cancer (PCa) diagnosis algorithms. This work delivers insights into the applicability of different approaches for this context based on an exhaustive examination. The outcome is a recommendation or preference for which machine learning model or family of models is best suited to optimize an existing pipeline when the model is applied as an upstream filter.

PMID:39125553 | DOI:10.3390/diagnostics14151677

Diagnostics (Basel). 2024 Aug 1;14(15):1669. doi: 10.3390/diagnostics14151669.

ABSTRACT

Advancements in artificial intelligence (AI) for point-of-care ultrasound (POCUS) have ushered in new possibilities for medical diagnostics in low-resource settings. This review explores the current landscape of AI applications in POCUS across these environments, analyzing studies sourced from three databases-SCOPUS, PUBMED, and Google Scholars. Initially, 1196 records were identified, of which 1167 articles were excluded after a two-stage screening, leaving 29 unique studies for review. The majority of studies focused on deep learning algorithms to facilitate POCUS operations and interpretation in resource-constrained settings. Various types of low-resource settings were targeted, with a significant emphasis on low- and middle-income countries (LMICs), rural/remote areas, and emergency contexts. Notable limitations identified include challenges in generalizability, dataset availability, regional disparities in research, patient compliance, and ethical considerations. Additionally, the lack of standardization in POCUS devices, protocols, and algorithms emerged as a significant barrier to AI implementation. The diversity of POCUS AI applications in different domains (e.g., lung, hip, heart, etc.) illustrates the challenges of having to tailor to the specific needs of each application. By separating out the analysis by application area, researchers will better understand the distinct impacts and limitations of AI, aligning research and development efforts with the unique characteristics of each clinical condition. Despite these challenges, POCUS AI systems show promise in bridging gaps in healthcare delivery by aiding clinicians in low-resource settings. Future research endeavors should prioritize addressing the gaps identified in this review to enhance the feasibility and effectiveness of POCUS AI applications to improve healthcare outcomes in resource-constrained environments.

PMID:39125545 | DOI:10.3390/diagnostics14151669

JMIR Med Inform. 2024 Aug 9;12:e57097. doi: 10.2196/57097.

ABSTRACT

BACKGROUND: Activities of daily living (ADL) are essential for independence and personal well-being, reflecting an individual's functional status. Impairment in executing these tasks can limit autonomy and negatively affect quality of life. The assessment of physical function during ADL is crucial for the prevention and rehabilitation of movement limitations. Still, its traditional evaluation based on subjective observation has limitations in precision and objectivity.

OBJECTIVE: The primary objective of this study is to use innovative technology, specifically wearable inertial sensors combined with artificial intelligence techniques, to objectively and accurately evaluate human performance in ADL. It is proposed to overcome the limitations of traditional methods by implementing systems that allow dynamic and noninvasive monitoring of movements during daily activities. The approach seeks to provide an effective tool for the early detection of dysfunctions and the personalization of treatment and rehabilitation plans, thus promoting an improvement in the quality of life of individuals.

METHODS: To monitor movements, wearable inertial sensors were developed, which include accelerometers and triaxial gyroscopes. The developed sensors were used to create a proprietary database with 6 movements related to the shoulder and 3 related to the back. We registered 53,165 activity records in the database (consisting of accelerometer and gyroscope measurements), which were reduced to 52,600 after processing to remove null or abnormal values. Finally, 4 deep learning (DL) models were created by combining various processing layers to explore different approaches in ADL recognition.

RESULTS: The results revealed high performance of the 4 proposed models, with levels of accuracy, precision, recall, and F1-score ranging between 95% and 97% for all classes and an average loss of 0.10. These results indicate the great capacity of the models to accurately identify a variety of activities, with a good balance between precision and recall. Both the convolutional and bidirectional approaches achieved slightly superior results, although the bidirectional model reached convergence in a smaller number of epochs.

CONCLUSIONS: The DL models implemented have demonstrated solid performance, indicating an effective ability to identify and classify various daily activities related to the shoulder and lumbar region. These results were achieved with minimal sensorization-being noninvasive and practically imperceptible to the user-which does not affect their daily routine and promotes acceptance and adherence to continuous monitoring, thus improving the reliability of the data collected. This research has the potential to have a significant impact on the clinical evaluation and rehabilitation of patients with movement limitations, by providing an objective and advanced tool to detect key movement patterns and joint dysfunctions.

PMID:39121473 | DOI:10.2196/57097

ACS Appl Mater Interfaces. 2024 Aug 9. doi: 10.1021/acsami.4c07851. Online ahead of print.

ABSTRACT

Applying machine-learning techniques for imbalanced data sets presents a significant challenge in materials science since the underrepresented characteristics of minority classes are often buried by the abundance of unrelated characteristics in majority of classes. Existing approaches to address this focus on balancing the counts of each class using oversampling or synthetic data generation techniques. However, these methods can lead to loss of valuable information or overfitting. Here, we introduce a deep learning framework to predict minority-class materials, specifically within the realm of metal-insulator transition (MIT) materials. The proposed approach, termed boosting-CGCNN, combines the crystal graph convolutional neural network (CGCNN) model with a gradient-boosting algorithm. The model effectively handled extreme class imbalances in MIT material data by sequentially building a deeper neural network. The comparative evaluations demonstrated the superior performance of the proposed model compared to other approaches. Our approach is a promising solution for handling imbalanced data sets in materials science.

PMID:39121441 | DOI:10.1021/acsami.4c07851

Phys Rev Lett. 2024 Jul 26;133(4):046802. doi: 10.1103/PhysRevLett.133.046802.

ABSTRACT

Freestanding ferroelectric oxide membranes emerge as a promising platform for exploring the interplay between topological polar ordering and dipolar interactions that are continuously tunable by strain. Our investigations combining density functional theory (DFT) and deep-learning-assisted molecular dynamics simulations demonstrate that DFT-predicted strain-driven morphotropic phase boundary involving monoclinic phases manifest as diverse domain structures at room temperatures, featuring continuous distributions of dipole orientations and mobile domain walls. Detailed analysis of dynamic structures reveals that the enhanced piezoelectric response observed in stretched PbTiO_{3} membranes results from small-angle rotations of dipoles at domain walls, distinct from conventional polarization rotation mechanism and adaptive phase theory inferred from static structures. We identify a ferroelectric topological structure, termed "dipole spiral," which exhibits a giant intrinsic piezoelectric response (>320 pC/N). This helical structure, possessing a rotational zero-energy mode, unlocks new possibilities for exploring chiral phonon dynamics and dipolar Dzyaloshinskii-Moriya-like interactions.

PMID:39121403 | DOI:10.1103/PhysRevLett.133.046802

PLoS One. 2024 Aug 9;19(8):e0307822. doi: 10.1371/journal.pone.0307822. eCollection 2024.

ABSTRACT

Accurately extracting the Region of Interest (ROI) of a palm print was crucial for subsequent palm print recognition. However, under unconstrained environmental conditions, the user's palm posture and angle, as well as the background and lighting of the environment, were not controlled, making the extraction of the ROI of palm print a major challenge. In existing research methods, traditional ROI extraction methods relied on image segmentation and were difficult to apply to multiple datasets simultaneously under the aforementioned interference. However, deep learning-based methods typically did not consider the computational cost of the model and were difficult to apply to embedded devices. This article proposed a palm print ROI extraction method based on lightweight networks. Firstly, the YOLOv5-lite network was used to detect and preliminarily locate the palm, in order to eliminate most of the interference from complex backgrounds. Then, an improved UNet was used for keypoints detection. This network model reduced the number of parameters compared to the original UNet model, improved network performance, and accelerated network convergence. The output of this model combined Gaussian heatmap regression and direct regression and proposed a joint loss function based on JS loss and L2 loss for supervision. During the experiment, a mixed database consisting of 5 databases was used to meet the needs of practical applications. The results showed that the proposed method achieved an accuracy of 98.3% on the database, with an average detection time of only 28ms on the GPU, which was superior to other mainstream lightweight networks, and the model size was only 831k. In the open-set test, with a success rate of 93.4%, an average detection time of 5.95ms on the GPU, it was far ahead of the latest palm print ROI extraction algorithm and could be applied in practice.

PMID:39121173 | DOI:10.1371/journal.pone.0307822

IEEE Trans Med Imaging. 2024 Aug 9;PP. doi: 10.1109/TMI.2024.3441494. Online ahead of print.

ABSTRACT

Chest radiography, commonly known as CXR, is frequently utilized in clinical settings to detect cardiopulmonary conditions. However, even seasoned radiologists might offer different evaluations regarding the seriousness and uncertainty associated with observed abnormalities. Previous research has attempted to utilize clinical notes to extract abnormal labels for training deep-learning models in CXR image diagnosis. However, these methods often neglected the varying degrees of severity and uncertainty linked to different labels. In our study, we initially assembled a comprehensive new dataset of CXR images based on clinical textual data, which incorporated radiologists' assessments of uncertainty and severity. Using this dataset, we introduced a multi-relationship graph learning framework that leverages spatial and semantic relationships while addressing expert uncertainty through a dedicated loss function. Our research showcases a notable enhancement in CXR image diagnosis and the interpretability of the diagnostic model, surpassing existing state-of-the-art methodologies. The dataset address of disease severity and uncertainty we extracted is: https://physionet.org/content/cad-chest/1.0/.

PMID:39120990 | DOI:10.1109/TMI.2024.3441494

IEEE Trans Neural Netw Learn Syst. 2024 Aug 9;PP. doi: 10.1109/TNNLS.2024.3436944. Online ahead of print.

ABSTRACT

Multi-instance learning (MIL) is a widely applied technique in practical applications that involve complex data structures. MIL can be broadly categorized into two types: traditional methods and those based on deep learning. These approaches have yielded significant results, especially regarding their problem-solving strategies and experiment validation, providing valuable insights for researchers in the MIL field. However, considerable knowledge is often trapped within the algorithm, leading to subsequent MIL algorithms that rely solely on the model's data fitting to predict unlabeled samples. This results in a significant loss of knowledge and impedes the development of more powerful models. In this article, we propose a novel data-driven knowledge fusion for deep MIL (DKMIL) algorithm. DKMIL adopts a completely different idea from existing deep MIL methods by analyzing the decision-making of key samples in the dataset (referred to as the data-driven) and using the knowledge fusion module designed to extract valuable information from these samples to assist the model's learning. In other words, this module serves as a new interface between data and the model, providing strong scalability and enabling prior knowledge from existing algorithms to enhance the model's learning ability. Furthermore, to adapt the downstream modules of the model to more knowledge-enriched features extracted from the data-driven knowledge fusion (DDKF) module, we propose a two-level attention (TLA) module that gradually learns shallow-and deep-level features of the samples to achieve more effective classification. We will prove the scalability of the knowledge fusion module and verify the efficiency of the proposed architecture by conducting experiments on 62 datasets across five categories.

PMID:39120987 | DOI:10.1109/TNNLS.2024.3436944