BMC Bioinformatics. 2024 Nov 6;25(1):348. doi: 10.1186/s12859-024-05961-w.

ABSTRACT

BACKGROUND: Antibodies play a crucial role in disease treatment, leveraging their ability to selectively interact with the specific antigen. However, screening antibody gene sequences for target antigens via biological experiments is extremely time-consuming and labor-intensive. Several computational methods have been developed to predict antibody-antigen interaction while suffering from the lack of characterizing the underlying structure of the antibody.

RESULTS: Beneficial from the recent breakthroughs in deep learning for antibody structure prediction, we propose a novel neural network architecture to predict antibody-antigen interaction. We first introduce AbAgIPA: an antibody structure prediction network to obtain the antibody backbone structure, where the structural features of antibodies and antigens are encoded into representation vectors according to the amino acid physicochemical features and Invariant Point Attention (IPA) computation methods. Finally, the antibody-antigen interaction is predicted by global max pooling, feature concatenation, and a fully connected layer. We evaluated our method on antigen diversity and antigen-specific antibody-antigen interaction datasets. Additionally, our model exhibits a commendable level of interpretability, essential for understanding underlying interaction mechanisms.

CONCLUSIONS: Quantitative experimental results demonstrate that the new neural network architecture significantly outperforms the best sequence-based methods as well as the methods based on residue contact maps and graph convolution networks (GCNs). The source code is freely available on GitHub at https://github.com/gmthu66/AbAgIPA .

PMID:39506679 | DOI:10.1186/s12859-024-05961-w

Neuropsychopharmacology. 2024 Nov 6. doi: 10.1038/s41386-024-02021-y. Online ahead of print.

ABSTRACT

Brain alterations associated with illness severity in schizophrenia remain poorly understood. Establishing linkages between imaging biomarkers and symptom expression may enhance mechanistic understanding of acute psychotic illness. Constructing models using MRI and clinical features together to maximize model validity may be particularly useful for these purposes. A multi-task deep learning model for standard case/control recognition incorporated with psychosis symptom severity regression was constructed with anatomic MRI collected from 286 patients with drug-naïve first-episode schizophrenia and 330 healthy controls from two datasets, and validated with an independent dataset including 40 first-episode schizophrenia. To evaluate the contribution of regression to the case/control recognition, a single-task classification model was constructed. Performance of unprocessed anatomical images and of predefined imaging features obtained using voxel-based morphometry (VBM) and surface-based morphometry (SBM), were examined and compared. Brain regions contributing to the symptom severity regression and illness identification were identified. Models developed with unprocessed images achieved greater group separation than either VBM or SBM measurements, differentiating schizophrenia patients from healthy controls with a balanced accuracy of 83.0% with sensitivity = 76.1% and specificity = 89.0%. The multi-task model also showed superior performance to single-task classification model without considering clinical symptoms. These findings showed high replication in the site-split validation and external validation analyses. Measurements in parietal, occipital and medial frontal cortex and bilateral cerebellum had the greatest contribution to the multi-task model. Incorporating illness severity regression in pattern recognition algorithms, our study developed an MRI-based model that was of high diagnostic value in acutely ill schizophrenia patients, highlighting clinical relevance of the model.

PMID:39506100 | DOI:10.1038/s41386-024-02021-y

Nat Comput Sci. 2024 Nov 6. doi: 10.1038/s43588-024-00720-6. Online ahead of print.

ABSTRACT

RNAs represent a class of programmable biomolecules capable of performing diverse biological functions. Recent studies have developed accurate RNA three-dimensional structure prediction methods, which may enable new RNAs to be designed in a structure-guided manner. Here, we develop a structure-to-sequence deep learning platform for the de novo generative design of RNA aptamers. We show that our approach can design RNA aptamers that are predicted to be structurally similar, yet sequence dissimilar, to known light-up aptamers that fluoresce in the presence of small molecules. We experimentally validate several generated RNA aptamers to have fluorescent activity, show that these aptamers can be optimized for activity in silico, and find that they exhibit a mechanism of fluorescence similar to that of known light-up aptamers. Our results demonstrate how structural predictions can guide the targeted and resource-efficient design of new RNA sequences.

PMID:39506080 | DOI:10.1038/s43588-024-00720-6

Sci Rep. 2024 Nov 6;14(1):27000. doi: 10.1038/s41598-024-78557-2.

ABSTRACT

Traditional English corpora mainly collect information from a single modality, but lack information from multimodal information, resulting in low quality of corpus information and certain problems with recognition accuracy. To solve the above problems, this paper proposes to introduce depth information into multimodal corpora, and studies the construction method of English multimodal corpora that integrates electronic images and depth information, as well as the speech recognition method of the corpus. The multimodal fusion strategy adopted integrates speech signals and image information, including key visual information such as the speaker's lip movements and facial expressions, and uses deep learning technology to mine acoustic and visual features. The acoustic model in the Kaldi toolkit is used for experimental research.Through experimental research, the following conclusions were drawn: Under 15-dimensional lip features, the accuracy of corpus A under monophone model was 2.4% higher than that of corpus B under monophone model when the SNR (signal-to-noise ratio) was 10dB, and the accuracy of corpus A under the triphone model at the signal-to-noise ratio of 10dB was 1.7% higher than that of corpus B under the triphone model at the signal-to-noise ratio of 10dB. Under the 32-dimensional lip features, the speech recognition effect of corpus A under the monophone model at the SNR of 10dB was 1.4% higher than that of corpus B under the monophone model at the SNR of 10dB, and the accuracy of corpus A under the triphone model at the SNR of 10dB was 2.6% higher than that of corpus B under the triphone model at the SNR of 10dB. The English multimodal corpus with image and depth information has a high accuracy, and the depth information helps to improve the accuracy of the corpus.

PMID:39506055 | DOI:10.1038/s41598-024-78557-2

Sci Rep. 2024 Nov 6;14(1):26878. doi: 10.1038/s41598-024-78081-3.

ABSTRACT

It aimed to analyze the value of deep learning algorithm combined with magnetic resonance imaging (MRI) in the risk diagnosis and prognosis of endometrial cancer (EC). Based on the deep learning convolutional neural network (CNN) architecture residual network with 101 layers (ResNet-101), spatial attention and channel attention modules were introduced to optimize the model. A retrospective collection of MRI image data from 210 EC patients was used for model segmentation and reconstruction, with 140 cases as the test set and 70 cases as the validation set. The performance was compared with traditional ResNet-101 model, ResNet-101 model based on spatial attention mechanism (SA-ResNet-101), and ResNet-101 model based on channel attention mechanism (CA-ResNet-101), using accuracy (AC), precision (PR), recall (RE), and F1 score as evaluation metrics. Among the 70 cases in the validation set, there were 45 cases of low-risk EC and 25 cases of high-risk EC. Using ROC curve analysis, it was found that the area under the curve (AUC) for the diagnosis of high-risk EC of the proposed model in this article (0.918) was visibly larger as against traditional ResNet-101 (0.613), SA-ResNet-101 (0.760), and CA-ResNet-101 models (0.758). The AC, PR, RE, and F1 values of the proposed model for the diagnosis of EC risk were visibly higher (P < 0.05). In the validation set, postoperative recurrence occurred in 13 cases and did not occur in 57 cases. Using ROC curve analysis, it was found that the AUC for postoperative recurrence prediction of the patients by the proposed model (0.926) was visibly larger as against traditional ResNet-101 (0.620), SA-ResNet-101 (0.729), and CA-ResNet-101 models (0.767). The AC, PR, RE, and F1 values of the proposed model for postoperative recurrence prediction were visibly higher (P < 0.05). The proposed model in this article, assisted by MRI, presented superior performance in diagnosing high-risk EC patients, with higher sensitivity (Sen) and specificity (Spe), and also demonstrated excellent predictive AC in postoperative recurrence prediction.

PMID:39506051 | DOI:10.1038/s41598-024-78081-3

Sci Rep. 2024 Nov 6;14(1):26993. doi: 10.1038/s41598-024-78041-x.

ABSTRACT

Assessing street walkability is a critical agenda in urban planning and multidisciplinary research, as it facilitates public health, community cohesion, and urban sustainability. Existing evaluation systems primarily focus on objective measurements, often neglecting subjective assessments and the diverse walking needs influenced by different urban spatial elements. This study addresses these gaps by constructing a comprehensive evaluation framework that integrates both subjective and objective dimensions, combining three neighbourhood indicators: Macro-Scale Index, Micro-Scale Index, and Street Walking Preferences Index. A normalization weighting method synthesizes these indicators into a comprehensive index. We applied this framework to assess the street environment within Beijing's Fifth Ring Road. The empirical results demonstrate that: (1) The framework reliably reflects the distribution of walkability. (2) The three indicators show both similarities and differences, underscoring the need to consider the distinct roles of community and street-level elements and the interaction between subjective and objective dimensions. (3) In high-density cities with ring-road development patterns, the Macro-Scale Index closely aligns with the Comprehensive Index, demonstrating its accuracy in reflecting walkability. The proposed framework and findings offer new insights for street walkability research and theoretical support for developing more inclusive, sustainable and walkable cities.

PMID:39506013 | DOI:10.1038/s41598-024-78041-x

Sci Rep. 2024 Nov 6;14(1):26903. doi: 10.1038/s41598-024-77585-2.

ABSTRACT

Early detection and treatment of skin cancer are important for patient recovery and survival. Dermoscopy images can help clinicians for timely identification of cancer, but manual diagnosis is time-consuming, costly, and prone to human error. To conduct this, an innovative deep learning-based approach has been proposed for automatic melanoma detection. The proposed method involves preprocessing dermoscopy images to remove artifacts, enhance contrast, and cancel noise, followed by feeding them into an optimized Convolutional Neural Network (CNN). The CNN is trained using an innovative metaheuristic called the Improved Chameleon Swarm Algorithm (CSA) to optimize its performance. The approach has been validated using the SIIM-ISIC Melanoma dataset and the results have been confirmed through rigorous evaluation metrics. Simulation results demonstrate the efficacy of the proposed method in accurately diagnosing melanoma from dermoscopy images by highlighting its potential as a valuable tool for clinicians in early cancer detection.

PMID:39505992 | DOI:10.1038/s41598-024-77585-2

Sci Rep. 2024 Nov 6;14(1):26937. doi: 10.1038/s41598-024-78424-0.

ABSTRACT

Radiotherapy has been demonstrated to be one of the most significant treatments for cervical cancer, during which accurate and efficient delineation of target volumes is critical. To alleviate the data demand of deep learning and promote the establishment and promotion of auto-segmentation models in small and medium-sized oncology departments and single centres, we proposed an auto-segmentation algorithm to determine the cervical cancer target volume in small samples based on multi-decoder and semi-supervised learning (MDSSL), and we evaluated the accuracy via an independent test cohort. In this study, we retrospectively collected computed tomography (CT) datasets from 71 pelvic cervical cancer patients, and a 3:4 ratio was used for the training and testing sets. The clinical target volumes (CTVs) of the primary tumour area (CTV1) and pelvic lymph drainage area (CTV2) were delineated. For definitive radiotherapy (dRT), the primary gross target volume (GTVp) was simultaneously delineated. According to the data characteristics for small samples, the MDSSL network structure based on 3D U-Net was established to train the model by combining clinical anatomical information, which was compared with other segmentation methods, including supervised learning (SL) and transfer learning (TL). The dice similarity coefficient (DSC), 95% Hausdorff distance (HD95) and average surface distance (ASD) were used to evaluate the segmentation performance. The ability of the segmentation algorithm to improve the efficiency of online adaptive radiation therapy (ART) was assessed via geometric indicators and a subjective evaluation of radiation oncologists (ROs) in prospective clinical applications. Compared with the SL model and TL model, the proposed MDSSL model displayed the best DSC, HD95 and ASD overall, especially for the GTVp of dRT. We calculated the above geometric indicators in the range of the ground truth (head-foot direction). In the test set, the DSC, HD95 and ASD of the MDSSL model were 0.80/5.85 mm/0.95 mm for CTV1 of post-operative radiotherapy (pRT), 0.84/ 4.88 mm/0.73 mm for CTV2 of pRT, 0.84/6.58 mm/0.89 mm for GTVp of dRT, 0.85/5.36 mm/1.35 mm for CTV1 of dRT, and 0.84/4.09 mm/0.73 mm for CTV2 of dRT, respectively. In a prospective clinical study of online ART, the target volume modification time (MTime) was 3-5 min for dRT and 2-4 min for pRT, and the main duration of CTV1 modification was approximately 2 min. The introduction of the MDSSL method successfully improved the accuracy of auto-segmentation for the cervical cancer target volume in small samples, showed good consistency with RO delineation and satisfied clinical requirements. In this prospective online ART study, the application of the segmentation model was demonstrated to be useful for reducing the target volume delineation time and improving the efficiency of the online ART workflow, which can contribute to the development and promotion of cervical cancer online ART.

PMID:39505991 | DOI:10.1038/s41598-024-78424-0

Sci Rep. 2024 Nov 6;14(1):26920. doi: 10.1038/s41598-024-76445-3.

ABSTRACT

Our previous research papers have shown the potential of deep-learning models for real-time detection and control of porosity defects in 3D printing, specifically in the laser powder bed fusion (LPBF) process. Extending these models to identify other defects like surface deformation poses a challenge due to the scarcity of available data. This study introduces the use of Transfer Learning (TL) to train models on limited data for high accuracy in detecting surface deformations, marking the first attempt to apply a model trained on one defect type to another. Our approach demonstrates the power of transfer learning in adapting a model known for porosity detection in LPBF to identify surface deformations with high accuracy (94%), matching the performance of the best existing models but with significantly less complexity. This results in faster training and evaluation, ideal for real-time systems with limited computing capabilities. We further employed Gradient-weighted Class Activation Mapping (Grad-CAM) to visualize the model's decision-making, highlighting the areas influencing defect detection. This step is vital for developing a trustworthy model, showcasing the effectiveness of our approach in broadening the model's applicability while ensuring reliability and efficiency.

PMID:39505970 | DOI:10.1038/s41598-024-76445-3

Sci Rep. 2024 Nov 6;14(1):26970. doi: 10.1038/s41598-024-78040-y.

ABSTRACT

The axillary lymph node status remains an important prognostic factor in breast cancer, and nodal staging using sentinel lymph node biopsy (SLNB) is routine. Randomized clinical trials provide evidence supporting de-escalation of axillary surgery and omission of SLNB in patients at low risk. However, identifying sentinel lymph node macrometastases (macro-SLNMs) is crucial for planning treatment tailored to the individual patient. This study is the first to explore the capacity of deep learning (DL) models to identify macro-SLNMs based on preoperative clinicopathological characteristics. We trained and validated five multivariable models using a population-based cohort of 18,185 patients. DL models outperform logistic regression, with Transformer showing the strongest results, under the constraint that the sensitivity is no less than 90%, reflecting the sensitivity of SLNB. This highlights the feasibility of noninvasive macro-SLNM prediction using DL. Feature importance analysis revealed that patients with similar characteristics exhibited different nodal status predictions, indicating the need for additional predictors for further improvement.

PMID:39505964 | DOI:10.1038/s41598-024-78040-y

Sci Rep. 2024 Nov 6;14(1):26919. doi: 10.1038/s41598-024-78028-8.

ABSTRACT

Cardiovascular diseases are a cause of death making it crucial to accurately diagnose them. Electrocardiography plays a role in detecting heart issues such as heart attacks, bundle branch blocks and irregular heart rhythms. Manual analysis of ECGs is prone to mistakes and time consuming, underscoring the importance of automated methods. This study uses AI models like AlexNet and a dual branch model for categorizing ECG signals from the PTB Diagnostic ECG Database. AlexNet achieved a validation accuracy of 98.64% and a test set accuracy of 99% while the dual branch fusion network model achieved a test set accuracy of 99%. Data preprocessing involved standardizing, balancing and reshaping ECG signals. These models exhibited precision, sensitivity and specificity. In comparison to state of the arts' models such as Hybrid AlexNet SVM and DCNN LSTM our proposed models displayed performance. The high accuracy rates of 99% underscore their potential for ECG classification. These results validate the advantages of incorporating learning models into setups for automated ECG analysis providing adaptable solutions for various healthcare settings including rural areas.

PMID:39505940 | DOI:10.1038/s41598-024-78028-8

Brief Bioinform. 2024 Sep 23;25(6):bbae550. doi: 10.1093/bib/bbae550.

ABSTRACT

Increasing studies have revealed the critical roles of human microbiome in a wide variety of disorders. Identification of disease-associated microbes might improve our knowledge and understanding of disease pathogenesis and treatment. Computational prediction of microbe-disease associations would provide helpful guidance for further biomedical screening, which has received lots of research interest in bioinformatics. In this study, a deep learning-based computational approach entitled SGJMDA is presented for predicting microbe-disease associations. Specifically, SGJMDA first fuses multiple similarities of microbes and diseases using a nonlinear strategy, and extracts feature information from homogeneous networks composed of the fused similarities via a graph convolution network. Second, a heterogeneous microbe-disease network is built to further capture the structural information of microbes and diseases by employing multi-neighborhood graph convolution network and jumping knowledge network. Finally, potential microbe-disease associations are inferred through computing the linear correlation coefficients of their embeddings. Results from cross-validation experiments show that SGJMDA outperforms 6 state-of-the-art computational methods. Furthermore, we carry out case studies on three important diseases using SGJMDA, in which 19, 20, and 11 predictions out of their top 20 results are successfully checked by the latest databases, respectively. The excellent performance of SGJMDA suggests that it could be a valuable and promising tool for inferring disease-associated microbes.

PMID:39504483 | DOI:10.1093/bib/bbae550

Brief Bioinform. 2024 Sep 23;25(6):bbae552. doi: 10.1093/bib/bbae552.

ABSTRACT

The accurate prediction of peptide-major histocompatibility complex (MHC) class I binding probabilities is a critical endeavor in immunoinformatics, with broad implications for vaccine development and immunotherapies. While recent deep neural network based approaches have showcased promise in peptide-MHC (pMHC) prediction, they have two shortcomings: (i) they rely on hand-crafted pseudo-sequence extraction, (ii) they do not generalize well to different datasets, which limits the practicality of these approaches. While existing methods rely on a 34 amino acid pseudo-sequence, our findings uncover the involvement of 147 positions in direct interactions between MHC and peptide. We further show that neural architectures can learn the intricacies of pMHC binding using even full sequences. To this end, we present PerceiverpMHC that is able to learn accurate representations on full-sequences by leveraging efficient transformer based architectures. Additionally, we propose IMGT/RobustpMHC that harnesses the potential of unlabeled data in improving the robustness of pMHC binding predictions through a self-supervised learning strategy. We extensively evaluate RobustpMHC on eight different datasets and showcase an overall improvement of over 6% in binding prediction accuracy compared to state-of-the-art approaches. We compile CrystalIMGT, a crystallography-verified dataset presenting a challenge to existing approaches due to significantly different pMHC distributions. Finally, to mitigate this distribution gap, we further develop a transfer learning pipeline.

PMID:39504482 | DOI:10.1093/bib/bbae552

Brief Bioinform. 2024 Sep 23;25(6):bbae555. doi: 10.1093/bib/bbae555.

ABSTRACT

The increasing single-cell RNA sequencing (scRNA-seq) data enable researchers to explore cellular heterogeneity and gene expression profiles, offering a high-resolution view of the transcriptome at the single-cell level. However, the dropout events, which are often present in scRNA-seq data, remaining challenges for downstream analysis. Although a number of studies have been developed to recover single-cell expression profiles, their performance may be hindered due to not fully exploring the inherent relations between genes. To address the issue, we propose scDTL, a deep transfer learning based approach for scRNA-seq data imputation by harnessing the bulk RNA-sequencing information. We firstly employ a denoising autoencoder trained on bulk RNA-seq data as the initial imputation model, and then leverage a domain adaptation framework that transfers the knowledge learned by the bulk imputation model to scRNA-seq learning task. In addition, scDTL employs a parallel operation with a 1D U-Net denoising model to provide gene representations of varying granularity, capturing both coarse and fine features of the scRNA-seq data. Finally, we utilize a cross-channel attention mechanism to fuse the features learned from the transferred bulk imputation model and U-Net model. In the evaluation, we conduct extensive experiments to demonstrate that scDTL could outperform other state-of-the-art methods in the quantitative comparison and downstream analyses.

PMID:39504481 | DOI:10.1093/bib/bbae555

Bioinformatics. 2024 Nov 6:btae662. doi: 10.1093/bioinformatics/btae662. Online ahead of print.

ABSTRACT

SUMMARY: Memprot.GPCR-ModSim leverages our previous web-based protocol, which was limited to class-A G protein-coupled receptors, to become the first one-stop web server for the modelling and simulation of any membrane protein system. Motivated by the exponential growth of experimental structures and the breakthrough of deep-learning-based structural modelling, the server accepts as input either a membrane-protein sequence, in which case it reports the associated AlphaFold model, or a 3D (experimental, modelled) structure, including quaternary complexes with associated proteins and/or ligands of any kind. In both cases, the molecular dynamics (MD) protocol produces a membrane-embedded, solvated, and equilibrated system, ready to be used as a starting point for further MD simulations, including ligand-binding free energy calculations.

AVAILABILITY: Memprot.GPCR-ModSim web server is publicly available at https://memprot.gpcr-modsim.org/. The standalone modules for 3D modelling (PyModSim) or membrane embedding and MD equilibration (PyMemDyn) are available at the GitHub repository https://github.com/GPCR-ModSim/.

SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

PMID:39504465 | DOI:10.1093/bioinformatics/btae662

J Craniofac Surg. 2024 Nov 6. doi: 10.1097/SCS.0000000000010829. Online ahead of print.

ABSTRACT

Dacryocystography (DCG) has been used to illustrate the morphological and functional aspects of the lacrimal drainage system in the evaluation of patients with maxillofacial trauma and epiphora. This study developed deep-learning models for the automatic classification of the status of the lacrimal passage based on DCG. The authors collected 719 DCG images from 430 patients with nasolacrimal duct obstruction. The obstruction images were further manually categorized into 2 binary categories based on the location of the obstruction: (1) upper obstruction and (2) lower obstruction. An upper obstruction was defined as one occurring within the canaliculus or common canaliculus, whereas a lower obstruction was defined as one within the lacrimal sac, duct-sac junction, or nasolacrimal duct. The authors then established a deep-learning model to automatically determine whether a passage was patent or obstruction. The accuracy, precision, sensitivity, F1 score, and area under the receiver operating characteristic curve for the evaluation set of each deep-learning model were 99.3%, 98.8%, 99.5%, 99.2%, and 0.9998, respectively, for obstruction detection, and 95.5%, 93.0%, 93.0%, 93.0%, and 0.9778 for classifying the obstruction location. Both receiver operating characteristic curves were skewed toward the left-upper region, indicating the high reliability of these models. The high accuracies of the obstruction detection model (99.3%) and the obstruction classification model (95.5%) demonstrate that deep-learning models can be reliable diagnostic tools for DCG images. This deep-learning model could enhance diagnostic consistency, enable non-specialists to interpret results accurately and facilitate the efficient allocation of medical resources.

PMID:39504416 | DOI:10.1097/SCS.0000000000010829

PLoS One. 2024 Nov 6;19(11):e0306441. doi: 10.1371/journal.pone.0306441. eCollection 2024.

ABSTRACT

Non-small cell lung cancer (NSCLC) exhibits a comparatively slower rate of metastasis in contrast to small cell lung cancer, contributing to approximately 85% of the global patient population. In this work, leveraging CT scan images, we deploy a knowledge distillation technique within teaching assistant (TA) and student frameworks for NSCLC classification. We employed various deep learning models, CNN, VGG19, ResNet152v2, Swin, CCT, and ViT, and assigned roles as teacher, teaching assistant and student. Evaluation underscores exceptional model performance in performance metrics achieved via cost-sensitive learning and precise hyperparameter (alpha and temperature) fine-tuning, highlighting the model's efficiency in lung cancer tumor prediction and classification. The applied TA (ResNet152) and student (CNN) models achieved 90.99% and 94.53% test accuracies, respectively, with optimal hyperparameters (alpha = 0.7 and temperature = 7). The implementation of the TA framework improves the overall performance of the student model. After obtaining Shapley values, explainable AI is applied with a partition explainer to check each class's contribution, further enhancing the transparency of the implemented deep learning techniques. Finally, a web application designed to make it user-friendly and classify lung types in recently captured images. The execution of the three-stage knowledge distillation technique proved efficient with significantly reduced trainable parameters and training time applicable for memory-constrained edge devices.

PMID:39504338 | DOI:10.1371/journal.pone.0306441

PLoS One. 2024 Nov 6;19(11):e0309368. doi: 10.1371/journal.pone.0309368. eCollection 2024.

ABSTRACT

Unlike in the field of visual scene recognition, where tremendous advances have taken place due to the availability of very large datasets to train deep neural networks, inference from medical images is often hampered by the fact that only small amounts of data may be available. When working with very small dataset problems, of the order of a few hundred items of data, the power of deep learning may still be exploited by using a pre-trained model as a feature extractor and carrying out classic pattern recognition techniques in this feature space, the so-called few-shot learning problem. However, medical images are highly complex and variable, making it difficult for few-shot learning to fully capture and model these features. To address these issues, we focus on the intrinsic characteristics of the data. We find that, in regimes where the dimension of the feature space is comparable to or even larger than the number of images in the data, dimensionality reduction is a necessity and is often achieved by principal component analysis or singular value decomposition (PCA/SVD). In this paper, noting the inappropriateness of using SVD for this setting we explore two alternatives based on discriminant analysis (DA) and non-negative matrix factorization (NMF). Using 14 different datasets spanning 11 distinct disease types we demonstrate that at low dimensions, discriminant subspaces achieve significant improvements over SVD-based subspaces and the original feature space. We also show that at modest dimensions, NMF is a competitive alternative to SVD in this setting. The implementation of the proposed method is accessible via the following link.

PMID:39504337 | DOI:10.1371/journal.pone.0309368

PLoS One. 2024 Nov 6;19(11):e0305943. doi: 10.1371/journal.pone.0305943. eCollection 2024.

ABSTRACT

Computational modeling of artwork meaning is complex and difficult. This is because art interpretation is multidimensional and highly subjective. This paper experimentally investigated the degree to which a state-of-the-art Deep Convolutional Neural Network (DCNN), a popular Machine Learning approach, can correctly distinguish modern conceptual art work into the galleries devised by art curators. Two hypotheses were proposed to state that the DCNN model uses Exhibited Properties for classification, like shape and color, but not Non-Exhibited Properties, such as historical context and artist intention. The two hypotheses were experimentally validated using a methodology designed for this purpose. VGG-11 DCNN pre-trained on ImageNet dataset and discriminatively fine-tuned was trained on handcrafted datasets designed from real-world conceptual photography galleries. Experimental results supported the two hypotheses showing that the DCNN model ignores Non-Exhibited Properties and uses only Exhibited Properties for artwork classification. This work points to current DCNN limitations, which should be addressed by future DNN models.

PMID:39504315 | DOI:10.1371/journal.pone.0305943

IEEE J Biomed Health Inform. 2024 Nov;28(11):6641-6652. doi: 10.1109/JBHI.2024.3457969.

ABSTRACT

Sleep staging is a crucial task in sleep monitoring and diagnosis, but clinical sleep staging is both time-consuming and subjective. In this study, we proposed a novel deep learning algorithm named feature fusion temporal convolutional network (FFTCN) for automatic sleep staging using single-channel EEG data. This algorithm employed a one-dimensional convolutional neural network (1D-CNN) to extract temporal features from raw EEG, and a two-dimensional CNN (2D-CNN) to extract time-frequency features from spectrograms generated through continuous wavelet transform (CWT) at the epoch level. These features were subsequently fused and further fed into a temporal convolutional network (TCN) to classify sleep stages at the sequence level. Moreover, a two-step training strategy was used to enhance the model's performance on an imbalanced dataset. Our proposed method exhibits superior performance in the 5-class classification task for healthy subjects, as evaluated on the SHHS-1, Sleep-EDF-153, and ISRUC-S1 datasets. This work provided a straightforward and promising method for improving the accuracy of automatic sleep staging using only single-channel EEG, and the proposed method exhibited great potential for future applications in professional sleep monitoring, which could effectively alleviate the workload of sleep technicians.

PMID:39504300 | DOI:10.1109/JBHI.2024.3457969