bioRxiv [Preprint]. 2024 Mar 10:2024.03.07.583913. doi: 10.1101/2024.03.07.583913.

ABSTRACT

BACKGROUND: Cell migration and invasion are well-coordinated processes in development and disease but remain poorly understood. We previously showed that highly migratory neural crest (NC) cells share a 45-gene panel with other cell invasion phenomena, including cancer. To identify critical genes of the 45-gene panel, we performed a high-throughput siRNA screen and used statistical and deep learning methods to compare NC- versus non-NC-derived human cell lines.

RESULTS: We find 14 out of 45 genes significantly reduces c8161 melanoma cell migration; only 4 are shared with HT1080 fibrosarcoma cells (BMP4, ITGB1, KCNE3, RASGRP1). Deep learning attention network analysis identified distinct cell-cell interaction patterns and significant alterations after BMP4 or RASGRP1 knockdown in c8161 cells. Addition of recombinant proteins to the culture media identified 5 out of the 10 known secreted molecules stimulate c8161 cell migration, including BMP4. BMP4 siRNA knockdown inhibited c8161 cell invasion in vivo and in vitro ; however, its addition to the culture media rescued c8161 cell invasion.

CONCLUSION: A high-throughput screen and deep learning rapidly distilled a 45-gene panel to a small subset of genes that appear critical to melanoma cell invasion and warrant deeper in vivo functional analysis for their role in driving the neural crest.

PMID:38496683 | PMC:PMC10942447 | DOI:10.1101/2024.03.07.583913

Res Sq [Preprint]. 2024 Mar 4:rs.3.rs-3959596. doi: 10.21203/rs.3.rs-3959596/v1.

ABSTRACT

Autism Spectrum Disorder (ASD) is among the most prevalent neurodevelopmental disorders, yet the current diagnostic procedures rely on behavioral analyses and interviews and lack objective screening methods. This study seeks to address this gap by integrating upper limb kinematics and deep learning methods to identify potential biomarkers that could be validated in younger age groups in the future to enhance the identification of ASD. Forty-one school-age children, with and without an ASD diagnosis (Mean age ± SE = 10.3 ± 0.4; 12 Females), participated in the study. A single Inertial Measurement Unit (IMU) was affixed to the child's wrist as they engaged in a continuous reaching and placing task. Deep learning techniques were employed to classify children with and without ASD. Our findings suggest delays in motor planning and control in school-age children compared to healthy adults. Compared to TD children, children with ASD exhibited poor motor planning and control as seen by greater number of movement units, more movement overshooting, and prolonged time to peak velocity/acceleration. Compensatory movement strategies such as greater velocity and acceleration were also seen in the ASD group. More importantly, using Multilayer Perceptron (MLP) model, we demonstrated an accuracy of ~ 78.1% in classifying children with and without ASD. These findings underscore the potential use of studying upper limb movement kinematics during goal-directed arm movements and deep learning methods as valuable tools for classifying and, consequently, aiding in the diagnosis and early identification of ASD upon further validation in younger children.

PMID:38496641 | PMC:PMC10942561 | DOI:10.21203/rs.3.rs-3959596/v1

bioRxiv [Preprint]. 2024 Mar 7:2023.05.09.540044. doi: 10.1101/2023.05.09.540044.

ABSTRACT

De novo design of complex protein folds using solely computational means remains a significant challenge. Here, we use a robust deep learning pipeline to design complex folds and soluble analogues of integral membrane proteins. Unique membrane topologies, such as those from GPCRs, are not found in the soluble proteome and we demonstrate that their structural features can be recapitulated in solution. Biophysical analyses reveal high thermal stability of the designs and experimental structures show remarkable design accuracy. The soluble analogues were functionalized with native structural motifs, standing as a proof-of-concept for bringing membrane protein functions to the soluble proteome, potentially enabling new approaches in drug discovery. In summary, we designed complex protein topologies and enriched them with functionalities from membrane proteins, with high experimental success rates, leading to a de facto expansion of the functional soluble fold space.

PMID:38496615 | PMC:PMC10942269 | DOI:10.1101/2023.05.09.540044

bioRxiv [Preprint]. 2024 Mar 6:2024.03.04.583444. doi: 10.1101/2024.03.04.583444.

ABSTRACT

5-hydroxymethylcytosine (5hmC), a critical epigenetic mark with a significant role in regulating tissue-specific gene expression, is essential for understanding the dynamic functions of the human genome. Using tissue-specific 5hmC sequencing data, we introduce Deep5hmC, a multimodal deep learning framework that integrates both the DNA sequence and the histone modification information to predict genome-wide 5hmC modification. The multimodal design of Deep5hmC demonstrates remarkable improvement in predicting both qualitative and quantitative 5hmC modification compared to unimodal versions of Deep5hmC and state-of-the-art machine learning methods. This improvement is demonstrated through benchmarking on a comprehensive set of 5hmC sequencing data collected at four time points during forebrain organoid development and across 17 human tissues. Notably, Deep5hmC showcases its practical utility by accurately predicting gene expression and identifying differentially hydroxymethylated regions in a case-control study of Alzheimer's disease.

PMID:38496575 | PMC:PMC10942288 | DOI:10.1101/2024.03.04.583444

Res Sq [Preprint]. 2024 Mar 7:rs.3.rs-4001384. doi: 10.21203/rs.3.rs-4001384/v1.

ABSTRACT

Recent GWASs have demonstrated that comorbid disorders share genetic liabilities. But whether and how these shared liabilities can be used for the classification and differentiation of comorbid disorders remains unclear. In this study, we use polygenic risk scores (PRSs) estimated from 42 comorbid traits and the deep neural networks (DNN) architecture to classify and differentiate schizophrenia (SCZ), bipolar disorder (BIP) and major depressive disorder (MDD). Multiple PRSs were obtained for individuals from the schizophrenia (SCZ) (cases = 6,317, controls = 7,240), bipolar disorder (BIP) (cases = 2,634, controls 4,425) and major depressive disorder (MDD) (cases = 1,704, controls = 3,357) datasets, and classification models were constructed with and without the inclusion of PRSs of the target (SCZ, BIP or MDD). Models with the inclusion of target PRSs performed well as expected. Surprisingly, we found that SCZ could be classified with only the PRSs from 35 comorbid traits (not including the target SCZ and directly related traits) (accuracy 0.760 ± 0.007, AUC 0.843 ± 0.005). Similar results were obtained for BIP (33 traits, accuracy 0.768 ± 0.007, AUC 0.848 ± 0.009), and MDD (36 traits, accuracy 0.794 ± 0.010, AUC 0.869 ± 0.004). Furthermore, these PRSs from comorbid traits alone could effectively differentiate unaffected controls, SCZ, BIP, and MDD patients (average categorical accuracy 0.861 ± 0.003, average AUC 0.961 ± 0.041). These results suggest that the shared liabilities from comorbid traits alone may be sufficient to classify SCZ, BIP and MDD. More importantly, these results imply that a data-driven and objective diagnosis and differentiation of SCZ, BIP and MDD may be feasible.

PMID:38496574 | PMC:PMC10942564 | DOI:10.21203/rs.3.rs-4001384/v1

Clin Breast Cancer. 2024 Feb 20:S1526-8209(24)00041-7. doi: 10.1016/j.clbc.2024.02.003. Online ahead of print.

ABSTRACT

OBJECTIVES: To develop predictive nomograms based on clinical and ultrasound features and to improve the clinical strategy for US BI-RADS 4A lesions.

METHODS: Patients with US BI-RADS 4A lesions from 3 hospitals between January 2016 and June 2020 were retrospectively included. Clinical and ultrasound features were extracted to establish nomograms CE (based on clinical experience) and DL (based on deep-learning algorithm). The performances of nomograms were evaluated by receiver operator characteristic curves, calibration curves and decision curves. Diagnostic performances with DL of radiologists were analyzed.

RESULTS: 1616 patients from 2 hospitals were randomly divided into training and internal validation cohorts at a ratio of 7:3. Hundred patients from another hospital made up external validation cohort. DL achieved more optimized AUCs than CE (internal validation: 0.916 vs. 0.863, P < .01; external validation: 0.884 vs. 0.776, P = .05). The sensitivities of DL were higher than CE (internal validation: 81.03% vs. 72.41%, P = .044; external validation: 93.75% vs. 81.25%, P = .4795) without losing specificity (internal validation: 84.91% vs. 86.47%, P = .353; external validation: 69.14% vs. 71.60%, P = .789). Decision curves indicated DL adds more clinical net benefit. With DL's assistance, both radiologists achieved higher AUCs (0.712 vs. 0.801; 0.547 vs. 0.800), improved specificities (70.93% vs. 74.42%, P < .001; 59.3% vs. 81.4%, P = .004), and decreased unnecessary biopsy rates by 6.7% and 24%.

CONCLUSION: DL was developed to discriminate US BI-RADS 4A lesions with a higher diagnostic power and more clinical net benefit than CE. Using DL may guide clinicians to make precise clinical decisions and avoid overtreatment of benign lesions.

PMID:38494415 | DOI:10.1016/j.clbc.2024.02.003

Ultrasound Med Biol. 2024 Mar 16:S0301-5629(24)00112-1. doi: 10.1016/j.ultrasmedbio.2024.02.013. Online ahead of print.

ABSTRACT

OBJECTIVE: Deep learning algorithms have commonly been used for the differential diagnosis between benign and malignant thyroid nodules. The aim of the study described here was to develop an integrated system that combines a deep learning model and a clinical standard Thyroid Imaging Reporting and Data System (TI-RADS) for the simultaneous segmentation and risk stratification of thyroid nodules.

METHODS: Three hundred four ultrasound images from two independent sites with TI-RADS 4 thyroid nodules were collected. The edge connection and Criminisi algorithm were used to remove manually induced markers in ultrasound images. An integrated system based on TI-RADS and a mask region-based convolution neural network (Mask R-CNN) was proposed to stratify subclasses of TI-RADS 4 thyroid nodules and to segment thyroid nodules in the ultrasound images. Accuracy and the precision-recall curve were used to evaluate stratification performance, and the Dice similarity coefficient (DSC) between the segmentation of Mask R-CNN and the radiologist's contour was used to evaluate the segmentation performance of the model.

RESULTS: The combined approach could significantly enhance the performance of the proposed integrated system. Overall stratification accuracy of TI-RADS 4 thyroid nodules, mean average precision and mean DSC of the proposed model in the independent test set was 90.79%, 0.8579 and 0.83, respectively. Specifically, stratification accuracy values for TI-RADS 4a, 4b and 4c thyroid nodules were 95.83%, 84.21% and 77.78%, respectively.

CONCLUSION: An integrated system combining TI-RADS and a deep learning model was developed. The system can provide clinicians with not only diagnostic assistance from TI-RADS but also accurate segmentation of thyroid nodules, which improves the applicability of the system in clinical practice.

PMID:38494413 | DOI:10.1016/j.ultrasmedbio.2024.02.013

J Genet Eng Biotechnol. 2024 Mar;22(1):100359. doi: 10.1016/j.jgeb.2024.100359. Epub 2024 Feb 21.

ABSTRACT

BACKGROUND: Examining functions and characteristics of DNA sequences is a highly challenging task. When it comes to the human genome, which is made up of exons and introns, this task is more challenging. Human exons and introns contain millions to billions of nucleotides, which contributes to the complexity observed in this sequences. Considering how complicated the subject of genomics is, it is obvious that using signal processing techniques and deep learning tools to build a strong predictive model can be very helpful for the development of the research of the human genome.

RESULTS: After representing human exons and introns with color images using Frequency Chaos Game Representation, two pre-trained convolutional neural network models (Resnet-50 and GoogleNet) and a proposed CNN model having 13 hidden layers were used to classify our obtained images. We have reached a value of 92% for the accuracy rate for Resnet-50 model in about 7 h for the execution time, a value of 91.5% for the accuracy rate for the GoogleNet model in 2 h and a half for the execution time. For our proposed CNN model, we have reached 91.6% for the accuracy rate in 2 h and 37 min.

CONCLUSIONS: Our proposed CNN model is faster than the Resnet-50 model in terms of execution time. It was able to slightly exceed the GoogleNet model for the accuracy rate value.

PMID:38494268 | DOI:10.1016/j.jgeb.2024.100359

Comput Biol Med. 2024 Mar 7;172:108250. doi: 10.1016/j.compbiomed.2024.108250. Online ahead of print.

ABSTRACT

Lung tumor segmentation in medical imaging is a critical step in the diagnosis and treatment planning for lung cancer. Accurate segmentation, however, is challenging due to the variability in tumor size, shape, and contrast against surrounding tissues. In this work, we present MSMV-Net, a novel deep learning architecture that integrates multi-scale multi-view (MSMV) learning modules and multi-scale uncertainty-based deep supervision (MUDS) for enhanced segmentation of lung tumors in computed tomography images. MSMV-Net capitalizes on the strengths of multi-view analysis and multi-scale feature extraction to address the limitations posed by small 3D lung tumors. The results indicate that MSMV-Net achieves state-of-the-art performance in lung tumor segmentation, recording a global Dice score of 55.60% on the LUNA dataset and 59.94% on the MSD dataset. Ablation studies conducted on the MSD dataset further validate that our method enhances segmentation accuracy.

PMID:38493603 | DOI:10.1016/j.compbiomed.2024.108250

J Biomech. 2024 Mar 13;166:112049. doi: 10.1016/j.jbiomech.2024.112049. Online ahead of print.

ABSTRACT

Markerless motion capture has recently attracted significant interest in clinical gait analysis and human movement science. Its ease of use and potential to streamline motion capture recordings bear great potential for out-of-the-laboratory measurements in large cohorts. While previous studies have shown that markerless systems can achieve acceptable accuracy and reliability for kinematic parameters of gait, they also noted higher inter-trial variability of markerless data. Since increased inter-trial variability can have important implications for data post-processing and analysis, this study compared the inter-trial variability of simultaneously recorded markerless and marker-based data. For this purpose, the data of 18 healthy volunteers were used who were instructed to simulate four different gait patterns: physiological, crouch, circumduction, and equinus gait. Gait analysis was performed using the smartphone-based markerless system OpenCap and a marker-based motion capture system. We compared the inter-trial variability of both systems and also evaluated if changes in inter-trial variability may depend on the analyzed gait pattern. Compared to the marker-based data, we observed an increase of inter-trial variability for the markerless system ranging from 6.6% to 22.0% for the different gait patterns. Our findings demonstrate that the markerless pose estimation pipelines can introduce additionally variability in the kinematic data across different gait patterns and levels of natural variability. We recommend using averaged waveforms rather than single ones to mitigate this problem. Further, caution is advised when using variability-based metrics in gait and human movement analysis based on markerless data as increased inter-trial variability can lead to misleading results.

PMID:38493576 | DOI:10.1016/j.jbiomech.2024.112049

Brief Bioinform. 2024 Jan 22;25(2):bbae102. doi: 10.1093/bib/bbae102.

ABSTRACT

In recent years, there has been a growing trend in the realm of parallel clustering analysis for single-cell RNA-seq (scRNA) and single-cell Assay of Transposase Accessible Chromatin (scATAC) data. However, prevailing methods often treat these two data modalities as equals, neglecting the fact that the scRNA mode holds significantly richer information compared to the scATAC. This disregard hinders the model benefits from the insights derived from multiple modalities, compromising the overall clustering performance. To this end, we propose an effective multi-modal clustering model scEMC for parallel scRNA and Assay of Transposase Accessible Chromatin data. Concretely, we have devised a skip aggregation network to simultaneously learn global structural information among cells and integrate data from diverse modalities. To safeguard the quality of integrated cell representation against the influence stemming from sparse scATAC data, we connect the scRNA data with the aggregated representation via skip connection. Moreover, to effectively fit the real distribution of cells, we introduced a Zero Inflated Negative Binomial-based denoising autoencoder that accommodates corrupted data containing synthetic noise, concurrently integrating a joint optimization module that employs multiple losses. Extensive experiments serve to underscore the effectiveness of our model. This work contributes significantly to the ongoing exploration of cell subpopulations and tumor microenvironments, and the code of our work will be public at https://github.com/DayuHuu/scEMC.

PMID:38493338 | DOI:10.1093/bib/bbae102

Sci Rep. 2024 Mar 16;14(1):6366. doi: 10.1038/s41598-024-56867-9.

ABSTRACT

This study aimed to develop a deep learning (DL) model for predicting the recurrence risk of lung adenocarcinoma (LUAD) based on its histopathological features. Clinicopathological data and whole slide images from 164 LUAD cases were collected and used to train DL models with an ImageNet pre-trained efficientnet-b2 architecture, densenet201, and resnet152. The models were trained to classify each image patch into high-risk or low-risk groups, and the case-level result was determined by multiple instance learning with final FC layer's features from a model from all patches. Analysis of the clinicopathological and genetic characteristics of the model-based risk group was performed. For predicting recurrence, the model had an area under the curve score of 0.763 with 0.750, 0.633 and 0.680 of sensitivity, specificity, and accuracy in the test set, respectively. High-risk cases for recurrence predicted by the model (HR group) were significantly associated with shorter recurrence-free survival and a higher stage (both, p < 0.001). The HR group was associated with specific histopathological features such as poorly differentiated components, complex glandular pattern components, tumor spread through air spaces, and a higher grade. In the HR group, pleural invasion, necrosis, and lymphatic invasion were more frequent, and the size of the invasion was larger (all, p < 0.001). Several genetic mutations, including TP53 (p = 0.007) mutations, were more frequently found in the HR group. The results of stages I-II were similar to those of the general cohort. DL-based model can predict the recurrence risk of LUAD and identify the presence of the TP53 gene mutation by analyzing histopathologic features.

PMID:38493247 | DOI:10.1038/s41598-024-56867-9

Commun Med (Lond). 2024 Mar 16;4(1):51. doi: 10.1038/s43856-024-00476-0.

ABSTRACT

BACKGROUND: Type 1 diabetes (T1D) simulators, crucial for advancing diabetes treatments, often fall short of capturing the entire complexity of the glucose-insulin system due to the imprecise approximation of the physiological models. This study introduces a simulation approach employing a conditional deep generative model. The aim is to overcome the limitations of existing T1D simulators by synthesizing virtual patients that more accurately represent the entire glucose-insulin system physiology.

METHODS: Our methodology utilizes a sequence-to-sequence generative adversarial network to simulate virtual T1D patients causally. Causality is embedded in the model by introducing shifted input-output pairs during training, with a 90-min shift capturing the impact of input insulin and carbohydrates on blood glucose. To validate our approach, we train and evaluate the model using three distinct datasets, each consisting of 27, 12, and 10 T1D patients, respectively. In addition, we subject the trained model to further validation for closed-loop therapy, employing a state-of-the-art controller.

RESULTS: The generated patients display statistical similarity to real patients when evaluated on the time-in-range results for each of the standard blood glucose ranges in T1D management along with means and variability outcomes. When tested for causality, authentic causal links are identified between the insulin, carbohydrates, and blood glucose levels of the virtual patients. The trained generative model demonstrates behaviours that are closer to reality compared to conventional T1D simulators when subjected to closed-loop insulin therapy using a state-of-the-art controller.

CONCLUSIONS: These results highlight our approach's capability to accurately capture physiological dynamics and establish genuine causal relationships, holding promise for enhancing the development and evaluation of therapies in diabetes.

PMID:38493243 | DOI:10.1038/s43856-024-00476-0

Ann Biomed Eng. 2024 Mar 16. doi: 10.1007/s10439-024-03486-0. Online ahead of print.

ABSTRACT

In recent years, research on automated sleep analysis has witnessed significant growth, reflecting advancements in understanding sleep patterns and their impact on overall health. This review synthesizes findings from an exhaustive analysis of 87 papers, systematically retrieved from prominent databases such as Google Scholar, PubMed, IEEE Xplore, and ScienceDirect. The selection criteria prioritized studies focusing on methods employed, signal modalities utilized, and machine learning algorithms applied in automated sleep analysis. The overarching goal was to critically evaluate the strengths and weaknesses of the proposed methods, shedding light on the current landscape and future directions in sleep research. An in-depth exploration of the reviewed literature revealed a diverse range of methodologies and machine learning approaches employed in automated sleep studies. Notably, K-Nearest Neighbors (KNN), Ensemble Learning Methods, and Support Vector Machine (SVM) emerged as versatile and potent classifiers, exhibiting high accuracies in various applications. However, challenges such as performance variability and computational demands were observed, necessitating judicious classifier selection based on dataset intricacies. In addition, the integration of traditional feature extraction methods with deep structures and the combination of different deep neural networks were identified as promising strategies to enhance diagnostic accuracy in sleep-related studies. The reviewed literature emphasized the need for adaptive classifiers, cross-modality integration, and collaborative efforts to drive the field toward more accurate, robust, and accessible sleep-related diagnostic solutions. This comprehensive review serves as a solid foundation for researchers and practitioners, providing an organized synthesis of the current state of knowledge in automated sleep analysis. By highlighting the strengths and challenges of various methodologies, this review aims to guide future research toward more effective and nuanced approaches to sleep diagnostics.

PMID:38493234 | DOI:10.1007/s10439-024-03486-0

Sci Rep. 2024 Mar 16;14(1):6394. doi: 10.1038/s41598-024-56439-x.

ABSTRACT

The underwater environment is filled with various sounds, with its soundscape composed of biological, geographical, and anthropological sounds. Our work focused on developing a novel method to observe and classify these sounds, enriching our understanding of the underwater ecosystem. We constructed a biologging system allowing near-real-time observation of underwater soundscapes. Utilizing deep-learning-based edge processing, this system classifies the sources of sounds, and upon the tagged animal surfacing, it transmits positional data, results of sound source classification, and sensor readings such as depth and temperature. To test the system, we attached the logger to sea turtles (Chelonia mydas) and collected data through a cellular network. The data provided information on the location-specific sounds detected by the sea turtles, suggesting the possibility to infer the distribution of specific species of organisms over time. The data showed that not only biological sounds but also geographical and anthropological sounds can be classified, highlighting the potential for conducting multi-point and long-term observations to monitor the distribution patterns of various sound sources. This system, which can be considered an autonomous mobile platform for oceanographic observations, including soundscapes, has significant potential to enhance our understanding of acoustic diversity.

PMID:38493174 | DOI:10.1038/s41598-024-56439-x

Plant Methods. 2024 Mar 16;20(1):44. doi: 10.1186/s13007-024-01171-w.

ABSTRACT

BACKGROUND: Plant scientists have largely relied on pathogen growth assays and/or transcript analysis of stress-responsive genes for quantification of disease severity and susceptibility. These methods are destructive to plants, labor-intensive, and time-consuming, thereby limiting their application in real-time, large-scale studies. Image-based plant phenotyping is an alternative approach that enables automated measurement of various symptoms. However, most of the currently available plant image analysis tools require specific hardware platform and vendor specific software packages, and thus, are not suited for researchers who are not primarily focused on plant phenotyping. In this study, we aimed to develop a digital phenotyping tool to enhance the speed, accuracy, and reliability of disease quantification in Arabidopsis.

RESULTS: Here, we present the Arabidopsis Disease Quantification (AraDQ) image analysis tool for examination of flood-inoculated Arabidopsis seedlings grown on plates containing plant growth media. It is a cross-platform application program with a user-friendly graphical interface that contains highly accurate deep neural networks for object detection and segmentation. The only prerequisite is that the input image should contain a fixed-sized 24-color balance card placed next to the objects of interest on a white background to ensure reliable and reproducible results, regardless of the image acquisition method. The image processing pipeline automatically calculates 10 different colors and morphological parameters for individual seedlings in the given image, and disease-associated phenotypic changes can be easily assessed by comparing plant images captured before and after infection. We conducted two case studies involving bacterial and plant mutants with reduced virulence and disease resistance capabilities, respectively, and thereby demonstrated that AraDQ can capture subtle changes in plant color and morphology with a high level of sensitivity.

CONCLUSIONS: AraDQ offers a simple, fast, and accurate approach for image-based quantification of plant disease symptoms using various parameters. Its fully automated pipeline neither requires prior image processing nor costly hardware setups, allowing easy implementation of the software by researchers interested in digital phenotyping of diseased plants.

PMID:38493119 | DOI:10.1186/s13007-024-01171-w

Int J Pharm. 2024 Mar 14:124001. doi: 10.1016/j.ijpharm.2024.124001. Online ahead of print.

ABSTRACT

Monitoring the particle size distribution (PSD) is crucial for controlling product quality during fluidized bed granulation. This paper proposed a rapid analytical method that quantifies the D10, D50, and D90 values using a Convolutional Block Attention Module-Convolutional Neural Network (CBAM-CNN) framework tailored for deep learning with near-infrared (NIR) spectroscopy. This innovative framework, which fuses CBAM with CNN, excels at extracting intricate features while prioritizing crucial ones, thereby facilitating the creation of a robust multi-output regression model. To expand the training dataset, we incorporated the C-Mixup algorithm, ensuring that the deep learning model was trained comprehensively. Additionally, the Bayesian optimization algorithm was introduced to optimize the hyperparameters, improving the prediction performance of the deep learning model. Compared with the commonly used Partial Least Squares (PLS), Support Vector Machine (SVM), and Artificial Neural Network (ANN) models, the CBAM-CNN model yielded higher prediction accuracy. Furthermore, the CBAM-CNN model avoided spectral preprocessing, preserved the spectral information to the maximum extent, and returned multiple predicted values at one time without degrading the prediction accuracy. Therefore, the CBAM-CNN model showed better prediction performance and modeling convenience for analyzing PSD values in fluidized bed granulation.

PMID:38492896 | DOI:10.1016/j.ijpharm.2024.124001

Neuroscience. 2024 Mar 14:S0306-4522(24)00112-X. doi: 10.1016/j.neuroscience.2024.03.007. Online ahead of print.

ABSTRACT

Alzheimer's disease (AD) is the general form of dementia, leading to a progressive neurological disorder characterized by memory loss due to brain cell damage. Artificial Intelligence (AI) assists in the early identification and prediction of AD patients, determining future risks and benefits for radiologists and doctors to save time and cost. Since deep learning (DL) approaches work well with massive datasets and have recently become helpful for AD detection, there remains an area for improvement in automating detection performance. Present approaches somehow addressed the challenges of limited annotated data samples for binary classification. This contrasts with prior state-of-the-art techniques, which were constrained by their incapacity to capture abstract-level information. In this paper, we proposed a Siamese 4D-AlzNet model comprised of four parallel CNN streams (Five CNN layer blocks) and customized transfer learning models (Frozen VGG-19, Frozen VGG-16, and customized AlexNet). Siamese 4D-AlzNet was vertically and horizontally stored, and the spatial features were passed to the final layer for classification. For experiments, T1-weighted MRI images comprised of four distinct subject classes, normal control (NC), mild cognitive impairment (MCI), late mild cognitive impairment (LMCI), and AD, have been employed. Our proposed models achieved outstanding accuracy, with a remarkable 95.05% accuracy distinguishing between normal and AD subjects. The performance across remaining binary class pairs consistently exceeded 90%. We thoroughly compared our model with the latest methods using the same dataset as our reference. Our proposed model improved NC-AD and MCI-AD classification accuracy by 2% to 7%.

PMID:38492797 | DOI:10.1016/j.neuroscience.2024.03.007

Med. 2024 Mar 12:S2666-6340(24)00078-3. doi: 10.1016/j.medj.2024.02.006. Online ahead of print.

ABSTRACT

BACKGROUND: Early diagnosis of atrial fibrillation (AF) is important for preventing stroke and other complications. Predicting AF risk in advance can improve early diagnostic efficiency. Deep learning has been used for disease risk prediction; however, it lacks adherence to evidence-based medicine standards. Identifying the underlying mechanisms behind disease risk prediction is important and required.

METHODS: We developed an explainable deep learning model called HBBI-AI to predict AF risk using only heart beat-to-beat intervals (HBBIs) during sinus rhythm. We proposed a possible AF mechanism based on the model's explainability and verified this conjecture using confirmed AF risk factors while also examining new AF risk factors. Finally, we investigated the changes in clinicians' ability to predict AF risk using only HBBIs before and after learning the model's explainability.

FINDINGS: HBBI-AI consistently performed well across large in-house and external public datasets. HBBIs with large changes or extreme stability were critical predictors for increased AF risk, and the underlying cause was autonomic imbalance. We verified various AF risk factors and discovered that autonomic imbalance was associated with all these factors. Finally, cardiologists effectively understood and learned from these findings to improve their abilities in AF risk prediction.

CONCLUSIONS: HBBI-AI effectively predicted AF risk using only HBBI information through evaluating autonomic imbalance. Autonomic imbalance may play an important role in many risk factors of AF rather than in a limited number of risk factors.

FUNDING: This study was supported in part by the National Key R&D Program and the National Natural Science Foundation of China.

PMID:38492571 | DOI:10.1016/j.medj.2024.02.006

Comput Biol Med. 2024 Mar 15;172:108317. doi: 10.1016/j.compbiomed.2024.108317. Online ahead of print.

ABSTRACT

Crafting effective deep learning models for medical image analysis is a complex task, particularly in cases where the medical image dataset lacks significant inter-class variation. This challenge is further aggravated when employing such datasets to generate synthetic images using generative adversarial networks (GANs), as the output of GANs heavily relies on the input data. In this research, we propose a novel filtering algorithm called Cosine Similarity-based Image Filtering (CosSIF). We leverage CosSIF to develop two distinct filtering methods: Filtering Before GAN Training (FBGT) and Filtering After GAN Training (FAGT). FBGT involves the removal of real images that exhibit similarities to images of other classes before utilizing them as the training dataset for a GAN. On the other hand, FAGT focuses on eliminating synthetic images with less discriminative features compared to real images used for training the GAN. The experimental results reveal that the utilization of either the FAGT or FBGT method reduces low inter-class variation in clinical image classification datasets and enables GANs to generate synthetic images with greater discriminative features. Moreover, modern transformer and convolutional-based models, trained with datasets that utilize these filtering methods, lead to less bias toward the majority class, more accurate predictions of samples in the minority class, and overall better generalization capabilities. Code and implementation details are available at: https://github.com/mominul-ssv/cossif.

PMID:38492455 | DOI:10.1016/j.compbiomed.2024.108317