Tomography. 2024 Apr 30;10(5):660-673. doi: 10.3390/tomography10050051.

ABSTRACT

Background: The arterial input function (AIF) is vital for myocardial blood flow quantification in cardiac MRI to indicate the input time-concentration curve of a contrast agent. Inaccurate AIFs can significantly affect perfusion quantification. Purpose: When only saturated and biased AIFs are measured, this work investigates multiple ways of leveraging tissue curve information, including using AIF + tissue curves as inputs and optimizing the loss function for deep neural network training. Methods: Simulated data were generated using a 12-parameter AIF mathematical model for the AIF. Tissue curves were created from true AIFs combined with compartment-model parameters from a random distribution. Using Bloch simulations, a dictionary was constructed for a saturation-recovery 3D radial stack-of-stars sequence, accounting for deviations such as flip angle, T2* effects, and residual longitudinal magnetization after the saturation. A preliminary simulation study established the optimal tissue curve number using a bidirectional long short-term memory (Bi-LSTM) network with just AIF loss. Further optimization of the loss function involves comparing just AIF loss, AIF with compartment-model-based parameter loss, and AIF with compartment-model tissue loss. The optimized network was examined with both simulation and hybrid data, which included in vivo 3D stack-of-star datasets for testing. The AIF peak value accuracy and ktrans results were assessed. Results: Increasing the number of tissue curves can be beneficial when added tissue curves can provide extra information. Using just the AIF loss outperforms the other two proposed losses, including adding either a compartment-model-based tissue loss or a compartment-model parameter loss to the AIF loss. With the simulated data, the Bi-LSTM network reduced the AIF peak error from -23.6 ± 24.4% of the AIF using the dictionary method to 0.2 ± 7.2% (AIF input only) and 0.3 ± 2.5% (AIF + ten tissue curve inputs) of the network AIF. The corresponding ktrans error was reduced from -13.5 ± 8.8% to -0.6 ± 6.6% and 0.3 ± 2.1%. With the hybrid data (simulated data for training; in vivo data for testing), the AIF peak error was 15.0 ± 5.3% and the corresponding ktrans error was 20.7 ± 11.6% for the AIF using the dictionary method. The hybrid data revealed that using the AIF + tissue inputs reduced errors, with peak error (1.3 ± 11.1%) and ktrans error (-2.4 ± 6.7%). Conclusions: Integrating tissue curves with AIF curves into network inputs improves the precision of AI-driven AIF corrections. This result was seen both with simulated data and with applying the network trained only on simulated data to a limited in vivo test dataset.

PMID:38787011 | DOI:10.3390/tomography10050051

Metabolites. 2024 May 19;14(5):290. doi: 10.3390/metabo14050290.

ABSTRACT

NMR is widely considered the gold standard for organic compound structure determination. As such, NMR is routinely used in organic compound identification, drug metabolite characterization, natural product discovery, and the deconvolution of metabolite mixtures in biofluids (metabolomics and exposomics). In many cases, compound identification by NMR is achieved by matching measured NMR spectra to experimentally collected NMR spectral reference libraries. Unfortunately, the number of available experimental NMR reference spectra, especially for metabolomics, medical diagnostics, or drug-related studies, is quite small. This experimental gap could be filled by predicting NMR chemical shifts for known compounds using computational methods such as machine learning (ML). Here, we describe how a deep learning algorithm that is trained on a high-quality, "solvent-aware" experimental dataset can be used to predict 1H chemical shifts more accurately than any other known method. The new program, called PROSPRE (PROton Shift PREdictor) can accurately (mean absolute error of <0.10 ppm) predict 1H chemical shifts in water (at neutral pH), chloroform, dimethyl sulfoxide, and methanol from a user-submitted chemical structure. PROSPRE (pronounced "prosper") has also been used to predict 1H chemical shifts for >600,000 molecules in many popular metabolomic, drug, and natural product databases.

PMID:38786767 | DOI:10.3390/metabo14050290

Metabolites. 2024 May 7;14(5):266. doi: 10.3390/metabo14050266.

ABSTRACT

A major limitation of most metabolomics datasets is the sparsity of pathway annotations for detected metabolites. It is common for less than half of the identified metabolites in these datasets to have a known metabolic pathway involvement. Trying to address this limitation, machine learning models have been developed to predict the association of a metabolite with a "pathway category", as defined by a metabolic knowledge base like KEGG. Past models were implemented as a single binary classifier specific to a single pathway category, requiring a set of binary classifiers for generating the predictions for multiple pathway categories. This past approach multiplied the computational resources necessary for training while diluting the positive entries in the gold standard datasets needed for training. To address these limitations, we propose a generalization of the metabolic pathway prediction problem using a single binary classifier that accepts the features both representing a metabolite and representing a pathway category and then predicts whether the given metabolite is involved in the corresponding pathway category. We demonstrate that this metabolite-pathway features pair approach not only outperforms the combined performance of training separate binary classifiers but demonstrates an order of magnitude improvement in robustness: a Matthews correlation coefficient of 0.784 ± 0.013 versus 0.768 ± 0.154.

PMID:38786743 | DOI:10.3390/metabo14050266

Metabolites. 2024 Apr 30;14(5):258. doi: 10.3390/metabo14050258.

ABSTRACT

Accurate risk prediction for myocardial infarction (MI) is crucial for preventive strategies, given its significant impact on global mortality and morbidity. Here, we propose a novel deep-learning approach to enhance the prediction of incident MI cases by incorporating metabolomics alongside clinical risk factors. We utilized data from the KORA cohort, including the baseline S4 and follow-up F4 studies, consisting of 1454 participants without prior history of MI. The dataset comprised 19 clinical variables and 363 metabolites. Due to the imbalanced nature of the dataset (78 observed MI cases and 1376 non-MI individuals), we employed a generative adversarial network (GAN) model to generate new incident cases, augmenting the dataset and improving feature representation. To predict MI, we further utilized multi-layer perceptron (MLP) models in conjunction with the synthetic minority oversampling technique (SMOTE) and edited nearest neighbor (ENN) methods to address overfitting and underfitting issues, particularly when dealing with imbalanced datasets. To enhance prediction accuracy, we propose a novel GAN for feature-enhanced (GFE) loss function. The GFE loss function resulted in an approximate 2% improvement in prediction accuracy, yielding a final accuracy of 70%. Furthermore, we evaluated the contribution of each clinical variable and metabolite to the predictive model and identified the 10 most significant variables, including glucose tolerance, sex, and physical activity. This is the first study to construct a deep-learning approach for producing 7-year MI predictions using the newly proposed loss function. Our findings demonstrate the promising potential of our technique in identifying novel biomarkers for MI prediction.

PMID:38786735 | DOI:10.3390/metabo14050258

J Imaging. 2024 May 17;10(5):122. doi: 10.3390/jimaging10050122.

ABSTRACT

Desmoid tumors (DTs) are non-metastasizing and locally aggressive soft-tissue mesenchymal neoplasms. Those that become enlarged often become locally invasive and cause significant morbidity. DTs have a varied pattern of clinical presentation, with up to 50-60% not growing after diagnosis and 20-30% shrinking or even disappearing after initial progression. Enlarging tumors are considered unstable and progressive. The management of symptomatic and enlarging DTs is challenging, and primarily consists of chemotherapy. Despite wide surgical resection, DTs carry a rate of local recurrence as high as 50%. There is a consensus that contrast-enhanced magnetic resonance imaging (MRI) or, alternatively, computerized tomography (CT) is the preferred modality for monitoring DTs. Each uses Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST 1.1), which measures the largest diameter on axial, sagittal, or coronal series. This approach, however, reportedly lacks accuracy in detecting response to therapy and fails to detect tumor progression, thus calling for more sophisticated methods. The objective of this study was to detect unique features identified by deep learning that correlate with the future clinical course of the disease. Between 2006 and 2019, 51 patients (mean age 41.22 ± 15.5 years) who had a tissue diagnosis of DT were included in this retrospective single-center study. Each had undergone at least three MRI examinations (including a pretreatment baseline study), and each was followed by orthopedic oncology specialists for a median of 38.83 months (IQR 44.38). Tumor segmentations were performed on a T2 fat-suppressed treatment-naive MRI sequence, after which the segmented lesion was extracted to a three-dimensional file together with its DICOM file and run through deep learning software. The results of the algorithm were then compared to clinical data collected from the patients' medical files. There were 28 males (13 stable) and 23 females (15 stable) whose ages ranged from 19.07 to 83.33 years. The model was able to independently predict clinical progression as measured from the baseline MRI with an overall accuracy of 93% (93 ± 0.04) and ROC of 0.89 ± 0.08. Artificial intelligence may contribute to risk stratification and clinical decision-making in patients with DT by predicting which patients are likely to progress.

PMID:38786576 | DOI:10.3390/jimaging10050122

J Imaging. 2024 May 15;10(5):121. doi: 10.3390/jimaging10050121.

ABSTRACT

In graph theory, the weighted Laplacian matrix is the most utilized technique to interpret the local and global properties of a complex graph structure within computer vision applications. However, with increasing graph nodes, the Laplacian matrix's dimensionality also increases accordingly. Therefore, there is always the "curse of dimensionality"; In response to this challenge, this paper introduces a new approach to reducing the dimensionality of the weighted Laplacian matrix by utilizing the Gershgorin circle theorem by transforming the weighted Laplacian matrix into a strictly diagonal domain and then estimating rough eigenvalue inclusion of a matrix. The estimated inclusions are represented as reduced features, termed GC features; The proposed Gershgorin circle feature extraction (GCFE) method was evaluated using three publicly accessible computer vision datasets, varying image patch sizes, and three different graph types. The GCFE method was compared with eight distinct studies. The GCFE demonstrated a notable positive Z-score compared to other feature extraction methods such as I-PCA, kernel PCA, and spectral embedding. Specifically, it achieved an average Z-score of 6.953 with the 2D grid graph type and 4.473 with the pairwise graph type, particularly on the E_Balanced dataset. Furthermore, it was observed that while the accuracy of most major feature extraction methods declined with smaller image patch sizes, the GCFE maintained consistent accuracy across all tested image patch sizes. When the GCFE method was applied to the E_MNSIT dataset using the K-NN graph type, the GCFE method confirmed its consistent accuracy performance, evidenced by a low standard deviation (SD) of 0.305. This performance was notably lower compared to other methods like Isomap, which had an SD of 1.665, and LLE, which had an SD of 1.325; The GCFE outperformed most feature extraction methods in terms of classification accuracy and computational efficiency. The GCFE method also requires fewer training parameters for deep-learning models than the traditional weighted Laplacian method, establishing its potential for more effective and efficient feature extraction in computer vision tasks.

PMID:38786575 | DOI:10.3390/jimaging10050121

J Imaging. 2024 May 9;10(5):116. doi: 10.3390/jimaging10050116.

ABSTRACT

Hyperfluorescence (HF) and reduced autofluorescence (RA) are important biomarkers in fundus autofluorescence images (FAF) for the assessment of health of the retinal pigment epithelium (RPE), an important indicator of disease progression in geographic atrophy (GA) or central serous chorioretinopathy (CSCR). Autofluorescence images have been annotated by human raters, but distinguishing biomarkers (whether signals are increased or decreased) from the normal background proves challenging, with borders being particularly open to interpretation. Consequently, significant variations emerge among different graders, and even within the same grader during repeated annotations. Tests on in-house FAF data show that even highly skilled medical experts, despite previously discussing and settling on precise annotation guidelines, reach a pair-wise agreement measured in a Dice score of no more than 63-80% for HF segmentations and only 14-52% for RA. The data further show that the agreement of our primary annotation expert with herself is a 72% Dice score for HF and 51% for RA. Given these numbers, the task of automated HF and RA segmentation cannot simply be refined to the improvement in a segmentation score. Instead, we propose the use of a segmentation ensemble. Learning from images with a single annotation, the ensemble reaches expert-like performance with an agreement of a 64-81% Dice score for HF and 21-41% for RA with all our experts. In addition, utilizing the mean predictions of the ensemble networks and their variance, we devise ternary segmentations where FAF image areas are labeled either as confident background, confident HF, or potential HF, ensuring that predictions are reliable where they are confident (97% Precision), while detecting all instances of HF (99% Recall) annotated by all experts.

PMID:38786570 | DOI:10.3390/jimaging10050116

Biofouling. 2024 May 24:1-15. doi: 10.1080/08927014.2024.2353034. Online ahead of print.

ABSTRACT

Thermal bubble-driven micro-pumps are an upcoming micro-actuator technology that can be directly integrated into micro/mesofluidic channels, have no moving parts, and leverage existing mass production fabrication approaches. These micro-pumps consist of a high-power micro-resistor that boils fluid in microseconds to create a high-pressure vapor bubble which performs mechanical work. As such, these micro-pumps hold great promise for micro/mesofluidic systems such as lab-on-a-chip technologies. However, to date, no current work has studied the interaction of these micro-pumps with biofluids such as blood and protein-rich fluids. In this study, the effects of organic fouling due to egg albumin and bovine whole blood are characterized using stroboscopic high-speed imaging and a custom deep learning neural network based on transfer learning of RESNET-18. It was found that the growth of a fouling film inhibited vapor bubble formation. A new metric to quantify the extent of fouling was proposed using the decrease in vapor bubble area as a function of the number of micro-pump firing events. Fouling due to egg albumin and bovine whole blood was found to significantly degrade pump performance as well as the lifetime of thermal bubble-driven micro-pumps to less than 104 firings, which may necessitate the use of protective thin film coatings to prevent the buildup of a fouling layer.

PMID:38785127 | DOI:10.1080/08927014.2024.2353034

Stud Health Technol Inform. 2024 May 23;314:103-107. doi: 10.3233/SHTI240067.

ABSTRACT

The Prediabetes impacts one in every three individuals, with a 10% annual probability of transitioning to type 2 diabetes without lifestyle changes or medical interventions. It's crucial to manage glycemic health to deter the progression to type 2 diabetes. In the United States, 13% of individuals (18 years of age and older) have diabetes, while 34.5% meet the criteria for prediabetes. Diabetes mellitus and prediabetes are more common in older persons. Currently, nevertheless, there aren't many noninvasive, commercially accessible methods for tracking glycemic status to help with prediabetes self-management. This study tackles the task of forecasting glucose levels using personalized prediabetes data through the utilization of the Long Short-Term Memory (LSTM) model. Continuous monitoring of interstitial glucose levels, heart rate measurements, and dietary records spanning a week were collected for analysis. The efficacy of the proposed model has been assessed using evaluation metrics including Root Mean Square Error (RMSE), Mean Squared Error (MSE), Mean Absolute Error (MAE), and the coefficient of determination (R2).

PMID:38785012 | DOI:10.3233/SHTI240067

Stud Health Technol Inform. 2024 May 23;314:98-102. doi: 10.3233/SHTI240066.

ABSTRACT

This paper explores the potential of leveraging electronic health records (EHRs) for personalized health research through the application of artificial intelligence (AI) techniques, specifically Named Entity Recognition (NER). By extracting crucial patient information from clinical texts, including diagnoses, medications, symptoms, and lab tests, AI facilitates the rapid identification of relevant data, paving the way for future care paradigms. The study focuses on Non-small cell lung cancer (NSCLC) in Italian clinical notes, introducing a novel set of 29 clinical entities that include both presence or absence (negation) of relevant information associated with NSCLC. Using a state-of-the-art model pretrained on Italian biomedical texts, we achieve promising results (average F1-score of 80.8%), demonstrating the feasibility of employing AI for extracting biomedical information in the Italian language.

PMID:38785011 | DOI:10.3233/SHTI240066

Front Neurol. 2024 May 9;15:1386728. doi: 10.3389/fneur.2024.1386728. eCollection 2024.

ABSTRACT

Acuity assessments are vital for timely interventions and fair resource allocation in critical care settings. Conventional acuity scoring systems heavily depend on subjective patient assessments, leaving room for implicit bias and errors. These assessments are often manual, time-consuming, intermittent, and challenging to interpret accurately, especially for healthcare providers. This risk of bias and error is likely most pronounced in time-constrained and high-stakes environments, such as critical care settings. Furthermore, such scores do not incorporate other information, such as patients' mobility level, which can indicate recovery or deterioration in the intensive care unit (ICU), especially at a granular level. We hypothesized that wearable sensor data could assist in assessing patient acuity granularly, especially in conjunction with clinical data from electronic health records (EHR). In this prospective study, we evaluated the impact of integrating mobility data collected from wrist-worn accelerometers with clinical data obtained from EHR for estimating acuity. Accelerometry data were collected from 87 patients wearing accelerometers on their wrists in an academic hospital setting. The data was evaluated using five deep neural network models: VGG, ResNet, MobileNet, SqueezeNet, and a custom Transformer network. These models outperformed a rule-based clinical score (Sequential Organ Failure Assessment, SOFA) used as a baseline when predicting acuity state (for ground truth we labeled as unstable patients if they needed life-supporting therapies, and as stable otherwise), particularly regarding the precision, sensitivity, and F1 score. The results demonstrate that integrating accelerometer data with demographics and clinical variables improves predictive performance compared to traditional scoring systems in healthcare. Deep learning models consistently outperformed the SOFA score baseline across various scenarios, showing notable enhancements in metrics such as the area under the receiver operating characteristic (ROC) Curve (AUC), precision, sensitivity, specificity, and F1 score. The most comprehensive scenario, leveraging accelerometer, demographics, and clinical data, achieved the highest AUC of 0.73, compared to 0.53 when using SOFA score as the baseline, with significant improvements in precision (0.80 vs. 0.23), specificity (0.79 vs. 0.73), and F1 score (0.77 vs. 0.66). This study demonstrates a novel approach beyond the simplistic differentiation between stable and unstable conditions. By incorporating mobility and comprehensive patient information, we distinguish between these states in critically ill patients and capture essential nuances in physiology and functional status. Unlike rudimentary definitions, such as equating low blood pressure with instability, our methodology delves deeper, offering a more holistic understanding and potentially valuable insights for acuity assessment.

PMID:38784909 | PMC:PMC11112699 | DOI:10.3389/fneur.2024.1386728

Front Bioeng Biotechnol. 2024 May 9;12:1392513. doi: 10.3389/fbioe.2024.1392513. eCollection 2024.

ABSTRACT

OBJECTIVE: We endeavor to develop a novel deep learning architecture tailored specifically for the analysis and classification of tongue features, including color, shape, and coating. Unlike conventional methods based on architectures like VGG or ResNet, our proposed method aims to address the challenges arising from their extensive size, thereby mitigating the overfitting problem. Through this research, we aim to contribute to the advancement of techniques in tongue feature recognition, ultimately leading to more precise diagnoses and better patient rehabilitation in Traditional Chinese Medicine (TCM).

METHODS: In this study, we introduce TGANet (Tongue Feature Attention Network) to enhance model performance. TGANet utilizes the initial five convolutional blocks of pre-trained VGG16 as the backbone and integrates an attention mechanism into this backbone. The integration of the attention mechanism aims to mimic human cognitive attention, emphasizing model weights on pivotal regions of the image. During the learning process, the allocation of attention weights facilitates the interpretation of causal relationships in the model's decision-making.

RESULTS: Experimental results demonstrate that TGANet outperforms baseline models, including VGG16, ResNet18, and TSC-WNet, in terms of accuracy, precision, F1 score, and AUC metrics. Additionally, TGANet provides a more intuitive and meaningful understanding of tongue feature classification models through the visualization of attention weights.

CONCLUSION: In conclusion, TGANet presents an effective approach to tongue feature classification, addressing challenges associated with model size and overfitting. By leveraging the attention mechanism and pre-trained VGG16 backbone, TGANet achieves superior performance metrics and enhances the interpretability of the model's decision-making process. The visualization of attention weights contributes to a more intuitive understanding of the classification process, making TGANet a promising tool in tongue diagnosis and rehabilitation.

PMID:38784768 | PMC:PMC11112418 | DOI:10.3389/fbioe.2024.1392513

Cureus. 2024 May 22;16(5):e60879. doi: 10.7759/cureus.60879. eCollection 2024 May.

ABSTRACT

Purpose The purpose of this study was to train a deep learning-based method for the prediction of postoperative recurrence of symptoms in Chiari malformation type 1 (CM1) patients undergoing surgery. Studies suggest that certain radiological and clinical features do exist in patients with treatment failure, though these are inconsistent and poorly defined. Methodology This study was a retrospective cohort study of patients who underwent primary surgical intervention for CM1 from January 2010 to May 2020. Only patients who completed pre- and postoperative 12-item short form (SF-12) surveys were included and these were used to classify the recurrence or persistence of symptoms. Forty patients had an improvement in overall symptoms while 17 had recurrence or persistence. After magnetic resonance imaging (MRI) data augmentation, a ResNet50, pre-trained on the ImageNet dataset, was used for feature extraction, and then clustering-constrained attention multiple instance learning (CLAM), a weakly supervised multi-instance learning framework, was trained for prediction of recurrence. Five-fold cross-validation was used for the development of MRI only, clinical features only, and a combined machine learning model. Results This study included 57 patients who underwent CM1 decompression. The recurrence rate was 30%. The combined model incorporating MRI, pre-operative SF-12 physical component scale (PCS), and extent of cerebellar ectopia performed best with an area under the curve (AUC) of 0.71 and an F1 score of 0.74. Conclusion This is the first study to our knowledge to explore the prediction of postoperative recurrence of symptoms in CM1 patients using machine learning methods and represents the first step toward developing a clinically useful prognostication machine learning model. Further studies utilizing a similar deep learning approach with a larger sample size are needed to improve the performance.

PMID:38784688 | PMC:PMC11111598 | DOI:10.7759/cureus.60879

Front Comput Neurosci. 2024 May 9;18:1365727. doi: 10.3389/fncom.2024.1365727. eCollection 2024.

ABSTRACT

Automatic segmentation of vestibular schwannoma (VS) from routine clinical MRI has potential to improve clinical workflow, facilitate treatment decisions, and assist patient management. Previous work demonstrated reliable automatic segmentation performance on datasets of standardized MRI images acquired for stereotactic surgery planning. However, diagnostic clinical datasets are generally more diverse and pose a larger challenge to automatic segmentation algorithms, especially when post-operative images are included. In this work, we show for the first time that automatic segmentation of VS on routine MRI datasets is also possible with high accuracy. We acquired and publicly release a curated multi-center routine clinical (MC-RC) dataset of 160 patients with a single sporadic VS. For each patient up to three longitudinal MRI exams with contrast-enhanced T1-weighted (ceT1w) (n = 124) and T2-weighted (T2w) (n = 363) images were included and the VS manually annotated. Segmentations were produced and verified in an iterative process: (1) initial segmentations by a specialized company; (2) review by one of three trained radiologists; and (3) validation by an expert team. Inter- and intra-observer reliability experiments were performed on a subset of the dataset. A state-of-the-art deep learning framework was used to train segmentation models for VS. Model performance was evaluated on a MC-RC hold-out testing set, another public VS datasets, and a partially public dataset. The generalizability and robustness of the VS deep learning segmentation models increased significantly when trained on the MC-RC dataset. Dice similarity coefficients (DSC) achieved by our model are comparable to those achieved by trained radiologists in the inter-observer experiment. On the MC-RC testing set, median DSCs were 86.2(9.5) for ceT1w, 89.4(7.0) for T2w, and 86.4(8.6) for combined ceT1w+T2w input images. On another public dataset acquired for Gamma Knife stereotactic radiosurgery our model achieved median DSCs of 95.3(2.9), 92.8(3.8), and 95.5(3.3), respectively. In contrast, models trained on the Gamma Knife dataset did not generalize well as illustrated by significant underperformance on the MC-RC routine MRI dataset, highlighting the importance of data variability in the development of robust VS segmentation models. The MC-RC dataset and all trained deep learning models were made available online.

PMID:38784680 | PMC:PMC11111906 | DOI:10.3389/fncom.2024.1365727

Front Big Data. 2024 May 9;7:1392662. doi: 10.3389/fdata.2024.1392662. eCollection 2024.

ABSTRACT

In recent years, analyzing the explanation for the prediction of Graph Neural Networks (GNNs) has attracted increasing attention. Despite this progress, most existing methods do not adequately consider the inherent uncertainties stemming from the randomness of model parameters and graph data, which may lead to overconfidence and misguiding explanations. However, it is challenging for most of GNN explanation methods to quantify these uncertainties since they obtain the prediction explanation in a post-hoc and model-agnostic manner without considering the randomness of graph data and model parameters. To address the above problems, this paper proposes a novel uncertainty quantification framework for GNN explanations. For mitigating the randomness of graph data in the explanation, our framework accounts for two distinct data uncertainties, allowing for a direct assessment of the uncertainty in GNN explanations. For mitigating the randomness of learned model parameters, our method learns the parameter distribution directly from the data, obviating the need for assumptions about specific distributions. Moreover, the explanation uncertainty within model parameters is also quantified based on the learned parameter distributions. This holistic approach can integrate with any post-hoc GNN explanation methods. Empirical results from our study show that our proposed method sets a new standard for GNN explanation performance across diverse real-world graph benchmarks.

PMID:38784676 | PMC:PMC11111992 | DOI:10.3389/fdata.2024.1392662

Heliyon. 2024 May 11;10(10):e30798. doi: 10.1016/j.heliyon.2024.e30798. eCollection 2024 May 30.

ABSTRACT

Although some deep learning-based image fusion approaches have realized promising results, how to extract information-rich features from different source images while preserving them in the fused image with less distortions remains challenging issue that needs to be addressed. Here, we propose a well worked-out GAN-based scheme with multi-scale feature extractor and global-local discriminator for infrared and visible image fusion. We use Y-Net as the backbone architecture to design the generator network, and introduce the residual dense block (RDblock) to yield more realistic fused images for infrared and visible images by learning discriminative multi-scale representations that are closer to the essence of different modal images. During feature reconstruction, the cross-modality shortcuts with contextual attention (CMSCA) are employed to selectively aggregate features at different scales and different levels to construct information-rich fused images with better visual effect. To ameliorate the information content of the fused image, we not only constrain the structure and contrast information using structural similarity index, but also evaluate the intensity and gradient similarities at both feature and image levels. Two global-local discriminators that combine global GAN with PatchGAN as a unified architecture help to dig for finer differences between the generated image and reference images, which force the generator to learn both the local radiation information and pervasive global details in two source images. It is worth mentioning that image fusion is achieved during confrontation without fusion rules. Lots of assessment tests demonstrate that the reported fusion scheme achieves superior performance against state-of-the-art works in meaningful information preservation.

PMID:38784534 | PMC:PMC11112272 | DOI:10.1016/j.heliyon.2024.e30798

Heliyon. 2024 May 9;10(10):e30756. doi: 10.1016/j.heliyon.2024.e30756. eCollection 2024 May 30.

ABSTRACT

Sentiment analysis has broad use in diverse real-world contexts, particularly in the online movie industry and other e-commerce platforms. The main objective of our work is to examine the word information order and analyze the content of texts by exploring the hidden meanings of words in online movie text reviews. This study presents an enhanced method of representing text and computationally feasible deep learning models, namely the PEW-MCAB model. The methodology categorizes sentiments by considering the full written text as a unified piece. The feature vector representation is processed using an enhanced text representation called Positional embedding and pretrained Glove Embedding Vector (PEW). The learning of these features is achieved by inculcating a multichannel convolutional neural network (MCNN), which is subsequently integrated into an Attention-based Bidirectional Long Short-Term Memory (AB) model. This experiment examines the positive and negative of online movie textual reviews. Four datasets were used to evaluate the model. When tested on the IMDB, MR (2002), MRC (2004), and MR (2005) datasets, the (PEW-MCAB) algorithm attained accuracy rates of 90.3%, 84.1%, 85.9%, and 87.1%, respectively, in the experimental setting. When implemented in practical settings, the proposed structure shows a great deal of promise for efficacy and competitiveness.

PMID:38784532 | PMC:PMC11112287 | DOI:10.1016/j.heliyon.2024.e30756

Radiother Oncol. 2024 May 21:110338. doi: 10.1016/j.radonc.2024.110338. Online ahead of print.

ABSTRACT

BACKGROUND: Volume of interest (VOI) segmentation is a crucial step for Radiomics analyses and radiotherapy (RT) treatment planning. Because it can be time-consuming and subject to inter-observer variability, we developed and tested a Deep Learning-based automatic segmentation (DLBAS) algorithm to reproducibly predict the primary gross tumor as VOI for Radiomics analyses in extremity soft tissue sarcomas (STS).

METHODS: A DLBAS algorithm was trained on a cohort of 157 patients and externally tested on an independent cohort of 87 patients using contrast-enhanced MRI. Manual tumor delineations by a radiation oncologist served as ground truths (GTs). A benchmark study with 20 cases from the test cohort compared the DLBAS predictions against manual VOI segmentations of two residents (ERs) and clinical delineations of two radiation oncologists (ROs). The ROs rated DLBAS predictions regarding their direct applicability.

RESULTS: The DLBAS achieved a median dice similarity coefficient (DSC) of 0.88 against the GTs in the entire test cohort (interquartile range (IQR): 0.11) and a median DSC of 0.89 (IQR 0.07) and 0.82 (IQR 0.10) in comparison to ERs and ROs, respectively. Radiomics feature stability was high with a median intraclass correlation coefficient of 0.97, 0.95 and 0.94 for GTs, ERs, and ROs, respectively. DLBAS predictions were deemed clinically suitable by the two ROs in 35% and 20% of cases, respectively.

CONCLUSION: The results demonstrate that the DLBAS algorithm provides reproducible VOI predictions for radiomics feature extraction. Variability remains regarding direct clinical applicability of predictions for RT treatment planning.

PMID:38782301 | DOI:10.1016/j.radonc.2024.110338

Am J Cardiol. 2024 May 21:S0002-9149(24)00394-1. doi: 10.1016/j.amjcard.2024.05.027. Online ahead of print.

ABSTRACT

We develop and evaluate an artificial intelligence-based algorithm that uses pre-rotation atherectomy (RA) intravascular ultrasound (IVUS) images to automatically predict regions debulked by RA. A total of 2106 IVUS cross-sections from 60 patients with de novo severely calcified coronary lesions who underwent IVUS-guided RA were consecutively collected. The two identical IVUS images of pre- and post-RA were merged, and the orientations of the debulked segments identified in the merged images were marked on the outer circle of each IVUS image. The artificial intelligence model was developed based on ResNet (deep residual learning for image recognition). The architecture connected 36 fully connected layers, each corresponding to one of the 36 orientations segmented every 10°, to a single feature extractor. In each cross-sectional analysis, our artificial intelligence model achieved an average sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of 81%, 72%, 46%, 90%, and 75%, respectively. In conclusion, the artificial intelligence-based algorithm can use information from pre-RA IVUS images to accurately predict regions debulked by RA and will assist interventional cardiologists in determining the treatment strategies for severely calcified coronary lesions.

PMID:38782227 | DOI:10.1016/j.amjcard.2024.05.027

Gene. 2024 May 21:148598. doi: 10.1016/j.gene.2024.148598. Online ahead of print.

ABSTRACT

Genes are the basic units of protein synthesis in organisms, and accurately identifying the translation initiation site (TIS) of genes is crucial for understanding the regulation, transcription, and translation processes of genes. However, the existing models cannot adequately extract the feature information in TIS sequences, and they also inadequately capture the complex hierarchical relationships among features. Therefore, a novel predictor named CapsNet-TIS is proposed in this paper. CapsNet-TIS first fully extracts the TIS sequence information using four encoding methods, including One-hot encoding, physical structure property (PSP) encoding, nucleotide chemical property (NCP) encoding, and nucleotide density (ND) encoding. Next, multi-scale convolutional neural networks are used to perform feature fusion of the encoded features to enhance the comprehensiveness of the feature representation. Finally, the fused features are classified using capsule network as the main network of the classification model to capture the complex hierarchical relationships among the features. Moreover, we improve the capsule network by introducing residual block, channel attention, and BiLSTM to enhance the model's feature extraction and sequence data modeling capabilities. In this paper, the performance of CapsNet-TIS is evaluated using TIS datasets from four species: human, mouse, bovine, and fruit fly, and the effectiveness of each part is demonstrated by performing ablation experiments. By comparing the experimental results with models proposed by other researchers, the results demonstrate the superior performance of CapsNet-TIS.

PMID:38782224 | DOI:10.1016/j.gene.2024.148598