J Ophthalmic Vis Res. 2024 Sep 16;19(3):354-367. doi: 10.18502/jovr.v19i3.15893. eCollection 2024 Jul-Sep.

ABSTRACT

Artificial intelligence (AI) holds immense promise for transforming ophthalmic care through automated screening, precision diagnostics, and optimized treatment planning. This paper reviews recent advances and challenges in applying AI techniques such as machine learning and deep learning to major eye diseases. In diabetic retinopathy, AI algorithms analyze retinal images to accurately identify lesions, which helps clinicians in ophthalmology practice. Systems like IDx-DR (IDx Technologies Inc, USA) are FDA-approved for autonomous detection of referable diabetic retinopathy. For glaucoma, deep learning models assess optic nerve head morphology in fundus photographs to detect damage. In age-related macular degeneration, AI can quantify drusen and diagnose disease severity from both color fundus and optical coherence tomography images. AI has also been used in screening for retinopathy of prematurity, keratoconus, and dry eye disease. Beyond screening, AI can aid treatment decisions by forecasting disease progression and anti-VEGF response. However, potential limitations such as the quality and diversity of training data, lack of rigorous clinical validation, and challenges in regulatory approval and clinician trust must be addressed for the widespread adoption of AI. Two other significant hurdles include the integration of AI into existing clinical workflows and ensuring transparency in AI decision-making processes. With continued research to address these limitations, AI promises to enable earlier diagnosis, optimized resource allocation, personalized treatment, and improved patient outcomes. Besides, synergistic human-AI systems could set a new standard for evidence-based, precise ophthalmic care.

PMID:39359529 | PMC:PMC11444002 | DOI:10.18502/jovr.v19i3.15893

Neurol Educ. 2023 Nov 29;2(4):e200103. doi: 10.1212/NE9.0000000000200103. eCollection 2023 Dec 22.

ABSTRACT

Game-based learning (GBL) has emerged as a promising approach to engage students and promote deep learning in a variety of educational settings. Neurology and neuroscience are complex fields that require an understanding of intricate neural structures and their functional roles. GBL can support the acquisition and application of such knowledge. In this article, we give an overview of the current state of GBL in neuroscience education. First, we review the language of gaming, establishing conceptual definitions for game elements, gamification, serious games, and GBL. Second, we discuss a literature review of games in the educational literature for adult learners involved in neuroscience. Third, we review available games intended for neuroscience education. Finally, we share tips for educators interested in developing their own educational games. By leveraging the unique features of games, including interactivity, feedback, and immersive experiences, educators and learners can engage with complex neuroscience concepts in a fun, engaging, and effective way.

PMID:39359316 | PMC:PMC11446165 | DOI:10.1212/NE9.0000000000200103

Front Bioeng Biotechnol. 2024 Sep 18;12:1468738. doi: 10.3389/fbioe.2024.1468738. eCollection 2024.

ABSTRACT

Droplet-based microfluidics techniques coupled to microscopy allow for the characterization of cells at the single-cell scale. However, such techniques generate substantial amounts of data and microscopy images that must be analyzed. Droplets on these images usually need to be classified depending on the number of cells they contain. This verification, when visually carried out by the experimenter image-per-image, is time-consuming and impractical for analysis of many assays or when an assay yields many putative droplets of interest. Machine learning models have already been developed to classify cell-containing droplets within microscopy images, but not in the context of assays in which non-cellular structures are present inside the droplet in addition to cells. Here we develop a deep learning model using the neural network ResNet-50 that can be applied to functional droplet-based microfluidic assays to classify droplets according to the number of cells they contain with >90% accuracy in a very short time. This model performs high accuracy classification of droplets containing both cells with non-cellular structures and cells alone and can accommodate several different cell types, for generalization to a broader array of droplet-based microfluidics applications.

PMID:39359262 | PMC:PMC11445169 | DOI:10.3389/fbioe.2024.1468738

Global Spine J. 2024 Oct 2:21925682241290752. doi: 10.1177/21925682241290752. Online ahead of print.

ABSTRACT

STUDY DESIGN: Narrative review.

OBJECTIVES: Artificial intelligence (AI) is being increasingly applied to the domain of spine surgery. We present a review of AI in spine surgery, including its use across all stages of the perioperative process and applications for research. We also provide commentary regarding future ethical considerations of AI use and how it may affect surgeon-industry relations.

METHODS: We conducted a comprehensive literature review of peer-reviewed articles that examined applications of AI during the pre-, intra-, or postoperative spine surgery process. We also discussed the relationship among AI, spine industry partners, and surgeons.

RESULTS: Preoperatively, AI has been mainly applied to image analysis, patient diagnosis and stratification, decision-making. Intraoperatively, AI has been used to aid image guidance and navigation. Postoperatively, AI has been used for outcomes prediction and analysis. AI can enable curation and analysis of huge datasets that can enhance research efforts. Large amounts of data are being accrued by industry sources for use by their AI platforms, though the inner workings of these datasets or algorithms are not well known.

CONCLUSIONS: AI has found numerous uses in the pre-, intra-, or postoperative spine surgery process, and the applications of AI continue to grow. The clinical applications and benefits of AI will continue to be more fully realized, but so will certain ethical considerations. Making industry-sponsored databases open source, or at least somehow available to the public, will help alleviate potential biases and obscurities between surgeons and industry and will benefit patient care.

PMID:39359113 | DOI:10.1177/21925682241290752

NMR Biomed. 2024 Oct 2:e5266. doi: 10.1002/nbm.5266. Online ahead of print.

ABSTRACT

Efficient abdominal coverage with T1-mapping methods currently available in the clinic is limited by the breath hold period (BHP) and the time needed for T1 recovery. This work develops a T1-mapping framework for efficient abdominal coverage based on rapid T1 recovery curve (T1RC) sampling, slice-selective inversion, optimized slice interleaving, and a convolutional neural network (CNN)-based T1 estimation. The effect of reducing the T1RC sampling was evaluated by comparing T1 estimates from T1RC ranging from 0.63 to 2.0 s with reference T1 values obtained from T1RC = 2.5-5 s. Slice interleaving methodologies were evaluated by comparing the T1 variation in abdominal organs across slices. The repeatability of the proposed framework was demonstrated by performing acquisition on test subjects across imaging sessions. Analysis of in vivo data based on retrospectively shortening the T1RC showed that with the CNN framework, a T1RC = 0.84 s yielded T1 estimates without significant changes in mean T1 (p > 0.05) or significant increase in T1 variability (p > 0.48) compared to the reference. Prospectively acquired data using T1RC = 0.84 s, an optimized slice interleaving scheme, and the CNN framework enabled 21 slices in a 20 s BHP. Analyses across abdominal organs produced T1 values within 2% of the reference. Repeatability experiments yielded Pearson's correlation, repeatability coefficient, and coefficient of variation of 0.99, 2.5%, and 0.12%, respectively. The proposed T1 mapping framework provides full abdominal coverage within a single BHP.

PMID:39358992 | DOI:10.1002/nbm.5266

J Pathol Clin Res. 2024 Nov;10(6):e70004. doi: 10.1002/2056-4538.70004.

ABSTRACT

EGFR mutations are a major prognostic factor in lung adenocarcinoma. However, current detection methods require sufficient samples and are costly. Deep learning is promising for mutation prediction in histopathological image analysis but has limitations in that it does not sufficiently reflect tumor heterogeneity and lacks interpretability. In this study, we developed a deep learning model to predict the presence of EGFR mutations by analyzing histopathological patterns in whole slide images (WSIs). We also introduced the EGFR mutation prevalence (EMP) score, which quantifies EGFR prevalence in WSIs based on patch-level predictions, and evaluated its interpretability and utility. Our model estimates the probability of EGFR prevalence in each patch by partitioning the WSI based on multiple-instance learning and predicts the presence of EGFR mutations at the slide level. We utilized a patch-masking scheduler training strategy to enable the model to learn various histopathological patterns of EGFR. This study included 868 WSI samples from lung adenocarcinoma patients collected from three medical institutions: Hallym University Medical Center, Inha University Hospital, and Chungnam National University Hospital. For the test dataset, 197 WSIs were collected from Ajou University Medical Center to evaluate the presence of EGFR mutations. Our model demonstrated prediction performance with an area under the receiver operating characteristic curve of 0.7680 (0.7607-0.7720) and an area under the precision-recall curve of 0.8391 (0.8326-0.8430). The EMP score showed Spearman correlation coefficients of 0.4705 (p = 0.0087) for p.L858R and 0.5918 (p = 0.0037) for exon 19 deletions in 64 samples subjected to next-generation sequencing analysis. Additionally, high EMP scores were associated with papillary and acinar patterns (p = 0.0038 and p = 0.0255, respectively), whereas low EMP scores were associated with solid patterns (p = 0.0001). These results validate the reliability of our model and suggest that it can provide crucial information for rapid screening and treatment plans.

PMID:39358807 | DOI:10.1002/2056-4538.70004

BioData Min. 2024 Oct 2;17(1):38. doi: 10.1186/s13040-024-00391-z.

ABSTRACT

BACKGROUND: The rapid growth of deep learning, as well as the vast and ever-growing amount of available data, have provided ample opportunity for advances in fusion and analysis of complex and heterogeneous data types. Different data modalities provide complementary information that can be leveraged to gain a more complete understanding of each subject. In the biomedical domain, multi-omics data includes molecular (genomics, transcriptomics, proteomics, epigenomics, metabolomics, etc.) and imaging (radiomics, pathomics) modalities which, when combined, have the potential to improve performance on prediction, classification, clustering and other tasks. Deep learning encompasses a wide variety of methods, each of which have certain strengths and weaknesses for multi-omics integration.

METHOD: In this review, we categorize recent deep learning-based approaches by their basic architectures and discuss their unique capabilities in relation to one another. We also discuss some emerging themes advancing the field of multi-omics integration.

RESULTS: Deep learning-based multi-omics integration methods were categorized broadly into non-generative (feedforward neural networks, graph convolutional neural networks, and autoencoders) and generative (variational methods, generative adversarial models, and a generative pretrained model). Generative methods have the advantage of being able to impose constraints on the shared representations to enforce certain properties or incorporate prior knowledge. They can also be used to generate or impute missing modalities. Recent advances achieved by these methods include the ability to handle incomplete data as well as going beyond the traditional molecular omics data types to integrate other modalities such as imaging data.

CONCLUSION: We expect to see further growth in methods that can handle missingness, as this is a common challenge in working with complex and heterogeneous data. Additionally, methods that integrate more data types are expected to improve performance on downstream tasks by capturing a comprehensive view of each sample.

PMID:39358793 | DOI:10.1186/s13040-024-00391-z

Nat Rev Genet. 2024 Oct 2. doi: 10.1038/s41576-024-00774-2. Online ahead of print.

ABSTRACT

Since the discovery of RNA splicing and its role in gene expression, researchers have sought a set of rules, an algorithm or a computational model that could predict the splice isoforms, and their frequencies, produced from any transcribed gene in a specific cellular context. Over the past 30 years, these models have evolved from simple position weight matrices to deep-learning models capable of integrating sequence data across vast genomic distances. Most recently, new model architectures are moving the field closer to context-specific alternative splicing predictions, and advances in sequencing technologies are expanding the type of data that can be used to inform and interpret such models. Together, these developments are driving improved understanding of splicing regulatory mechanisms and emerging applications of the splicing code to the rational design of RNA- and splicing-based therapeutics.

PMID:39358547 | DOI:10.1038/s41576-024-00774-2

Sci Rep. 2024 Oct 3;14(1):22941. doi: 10.1038/s41598-024-74526-x.

ABSTRACT

High-sensitivity acceleration sensors have been independently developed by our research group to detect vibrations that are > 10 dB smaller than those detected by conventional commercial sensors. This study is the first to measure high-frequency micro-vibrations in muscle fibers, termed micro-mechanomyogram (MMG) in patients with Parkinson's disease (PwPD) using a high-sensitivity acceleration sensor. We specifically measured the extensor pollicis brevis muscle at the base of the thumb in PwPD and healthy controls (HC) and detected not only low-frequency MMG (< 15 Hz) but also micro-MMG (≥ 15 Hz), which was preciously undetectable using commercial acceleration sensors. Analysis revealed remarkable differences in the frequency characteristics of micro-MMG between PwPD and HC. Specifically, during muscle power output, the low-frequency MMG energy was greater in PwPD than in HC, while the micro-MMG energy was smaller in PwPD compared to HC. These results suggest that micro-MMG detected by the high-sensitivity acceleration sensor provides crucial information for distinguishing between PwPD and HC. Moreover, a deep learning model trained on both low-frequency MMG and micro-MMG achieved a high accuracy (92.19%) in classifying PwPD and HC, demonstrating the potential for a diagnostic system for PwPD using micro-MMG.

PMID:39358456 | DOI:10.1038/s41598-024-74526-x

Sci Rep. 2024 Oct 2;14(1):22926. doi: 10.1038/s41598-024-73639-7.

ABSTRACT

The COVID-19 pandemic affected countries across the globe, demanding drastic public health policies to mitigate the spread of infection, which led to economic crises as a collateral damage. In this work, we investigate the impact of human mobility, described via international commercial flights, on COVID-19 infection dynamics on a global scale. We developed a graph neural network (GNN)-based framework called Dynamic Weighted GraphSAGE (DWSAGE), which operates over spatiotemporal graphs and is well-suited for dynamically changing flight information updated daily. This architecture is designed to be structurally sensitive, capable of learning the relationships between edge features and node features. To gain insights into the influence of air traffic on infection spread, we conducted local sensitivity analysis on our model through perturbation experiments. Our analyses identified Western Europe, the Middle East, and North America as leading regions in fueling the pandemic due to the high volume of air traffic originating or transiting through these areas. We used these observations to propose air traffic reduction strategies that can significantly impact controlling the pandemic with minimal disruption to human mobility. Our work provides a robust deep learning-based tool to study global pandemics and is of key relevance to policymakers for making informed decisions regarding air traffic restrictions during future outbreaks.

PMID:39358428 | DOI:10.1038/s41598-024-73639-7

NPJ Precis Oncol. 2024 Oct 2;8(1):218. doi: 10.1038/s41698-024-00718-3.

ABSTRACT

As part of the advancement in therapeutic decision-making for brain tumor patients at St. Jude Children's Research Hospital (SJCRH), we developed three robust classifiers, a deep learning neural network (NN), k-nearest neighbor (kNN), and random forest (RF), trained on a reference series DNA-methylation profiles to classify central nervous system (CNS) tumor types. The models' performance was rigorously validated against 2054 samples from two independent cohorts. In addition to classic metrics of model performance, we compared the robustness of the three models to reduced tumor purity, a critical consideration in the clinical utility of such classifiers. Our findings revealed that the NN model exhibited the highest accuracy and maintained a balance between precision and recall. The NN model was the most resistant to drops in performance associated with a reduction in tumor purity, showing good performance until the purity fell below 50%. Through rigorous validation, our study emphasizes the potential of DNA-methylation-based deep learning methods to improve precision medicine for brain tumor classification in the clinical setting.

PMID:39358389 | DOI:10.1038/s41698-024-00718-3

Sci Rep. 2024 Oct 2;14(1):22885. doi: 10.1038/s41598-024-72998-5.

ABSTRACT

Predicting rock tunnel squeezing in underground projects is challenging due to its intricate and unpredictable nature. This study proposes an innovative approach to enhance the accuracy and reliability of tunnel squeezing prediction. The proposed method combines ensemble learning techniques with Q-learning and online Markov chain integration. A deep learning model is trained on a comprehensive database comprising tunnel parameters including diameter (D), burial depth (H), support stiffness (K), and tunneling quality index (Q). Multiple deep learning models are trained concurrently, leveraging ensemble learning to capture diverse patterns and improve prediction performance. Integration of the Q-learning-Online Markov Chain further refines predictions. The online Markov chain analyzes historical sequences of tunnel parameters and squeezing class transitions, establishing transition probabilities between different squeezing classes. The Q-learning algorithm optimizes decision-making by learning the optimal policy for transitioning between tunnel states. The proposed model is evaluated using a dataset from various tunnel construction projects, assessing performance through metrics like accuracy, precision, recall, and F1-score. Results demonstrate the efficiency of the ensemble deep learning model combined with Q-learning-Online Markov Chain in predicting surrounding rock tunnel squeezing. This approach offers insights into parameter interrelationships and dynamic squeezing characteristics, enabling proactive planning and support measures implementation to mitigate tunnel squeezing hazards and ensure underground structure safety. Experimental results show the model achieves a prediction accuracy of 98.11%, surpassing individual CNN and RNN models, with an AUC value of 0.98.

PMID:39358373 | DOI:10.1038/s41598-024-72998-5

Asian J Surg. 2024 Oct 1:S1015-9584(24)02173-0. doi: 10.1016/j.asjsur.2024.09.128. Online ahead of print.

NO ABSTRACT

PMID:39358142 | DOI:10.1016/j.asjsur.2024.09.128

Brief Bioinform. 2024 Sep 23;25(6):bbae488. doi: 10.1093/bib/bbae488.

ABSTRACT

High affinity is crucial for the efficacy and specificity of antibody. Due to involving high-throughput screens, biological experiments for antibody affinity maturation are time-consuming and have a low success rate. Precise computational-assisted antibody design promises to accelerate this process, but there is still a lack of effective computational methods capable of pinpointing beneficial mutations within the complementarity-determining region (CDR) of antibodies. Moreover, random mutations often lead to challenges in antibody expression and immunogenicity. In this study, to enhance the affinity of a human antibody against avian influenza virus, a CDR library was constructed and evolutionary information was acquired through sequence alignment to restrict the mutation positions and types. Concurrently, a statistical potential methodology was developed based on amino acid interactions between antibodies and antigens to calculate potential affinity-enhanced antibodies, which were further subjected to molecular dynamics simulations. Subsequently, experimental validation confirmed that a point mutation enhancing 2.5-fold affinity was obtained from 10 designs, resulting in the antibody affinity of 2 nM. A predictive model for antibody-antigen interactions based on the binding interface was also developed, achieving an Area Under the Curve (AUC) of 0.83 and a precision of 0.89 on the test set. Lastly, a novel approach involving combinations of affinity-enhancing mutations and an iterative mutation optimization scheme similar to the Monte Carlo method were proposed. This study presents computational methods that rapidly and accurately enhance antibody affinity, addressing issues related to antibody expression and immunogenicity.

PMID:39358035 | DOI:10.1093/bib/bbae488

Ann Rheum Dis. 2024 Oct 2:ard-2024-225862. doi: 10.1136/ard-2024-225862. Online ahead of print.

ABSTRACT

OBJECTIVES: To assess the ability of a previously trained deep-learning algorithm to identify the presence of inflammation on MRI of sacroiliac joints (SIJ) in a large external validation set of patients with axial spondyloarthritis (axSpA).

METHODS: Baseline SIJ MRI scans were collected from two prospective randomised controlled trials in patients with non-radiographic (nr-) and radiographic (r-) axSpA (RAPID-axSpA: NCT01087762 and C-OPTIMISE: NCT02505542) and were centrally evaluated by two expert readers (and adjudicator in case of disagreement) for the presence of inflammation by the 2009 Assessment of SpondyloArthritis International Society (ASAS) definition. Scans were processed by the deep-learning algorithm, blinded to clinical information and central expert readings.

RESULTS: Pooling the patients from RAPID-axSpA (n=152) and C-OPTIMISE (n=579) yielded a validation set of 731 patients (mean age: 34.2 years, SD: 8.6; 505/731 (69.1%) male), of which 326/731 (44.6%) had nr-axSpA and 436/731 (59.6%) had inflammation on MRI per central readings. Scans were obtained from over 30 scanners from 5 manufacturers across over 100 clinical sites. Comparing the trained algorithm with the human central readings yielded a sensitivity of 70% (95% CI 66% to 73%), specificity of 81% (95% CI 78% to 84%), positive predictive value of 84% (95% CI 82% to 87%), negative predictive value of 64% (95% CI 61% to 68%), Cohen's kappa of 0.49 (95% CI 0.43 to 0.55) and absolute agreement of 74% (95% CI 72% to 77%).

CONCLUSION: The algorithm enabled acceptable detection of inflammation according to the 2009 ASAS MRI definition in a large external validation cohort.

PMID:39357994 | DOI:10.1136/ard-2024-225862

Comput Methods Programs Biomed. 2024 Sep 19;257:108435. doi: 10.1016/j.cmpb.2024.108435. Online ahead of print.

ABSTRACT

BACKGROUND AND OBJECTIVE: Hepatocellular carcinoma (HCC) ranks fourth in cancer mortality, underscoring the importance of accurate prognostic predictions to improve postoperative survival rates in patients. Although micronecrosis has been shown to have high prognostic value in HCC, its application in clinical prognosis prediction requires specialized knowledge and complex calculations, which poses challenges for clinicians. It would be of interest to develop a model to help clinicians make full use of micronecrosis to assess patient survival.

METHODS: To address these challenges, we propose a HCC prognosis prediction model that integrates pathological micronecrosis information through Graph Convolutional Neural Networks (GCN). This approach enables GCN to utilize micronecrosis, which has been shown to be highly correlated with prognosis, thereby significantly enhancing prognostic stratification quality. We developed our model using 3622 slides from 752 patients with primary HCC from the FAH-ZJUMS dataset and conducted internal and external validations on the FAH-ZJUMS and TCGA-LIHC datasets, respectively.

RESULTS: Our method outperformed the baseline by 8.18% in internal validation and 9.02% in external validations. Overall, this paper presents a deep learning research paradigm that integrates HCC micronecrosis, enhancing both the accuracy and interpretability of prognostic predictions, with potential applicability to other pathological prognostic markers.

CONCLUSIONS: This study proposes a composite GCN prognostic model that integrates information on HCC micronecrosis, collecting large dataset of HCC histopathological images. This approach could assist clinicians in analyzing HCC patient survival and precisely locating and visualizing necrotic tissues that affect prognosis. Following the research paradigm outlined in this paper, other prognostic biomarker integration models with GCN could be developed, significantly enhancing the predictive performance and interpretability of prognostic model.

PMID:39357091 | DOI:10.1016/j.cmpb.2024.108435

J Am Chem Soc. 2024 Oct 2. doi: 10.1021/jacs.4c08866. Online ahead of print.

ABSTRACT

Deep learning (DL) methods have gained notable prominence in predictive and generative tasks in molecular space. However, their application in chemical reactions remains grossly underutilized. Chemical reactions are intrinsically complex: typically involving multiple molecules besides bond-breaking/forming events. In reaction discovery, one aims to maximize yield and/or selectivity that depends on a number of factors, mostly centered on reacting partners and reaction conditions. Herein, we introduce RE-EXPLORE, a novel approach that integrates deep reinforcement learning (RL) with an RNN-based deep generative model to identify prospective new reactants/catalysts, whose yield/selectivity is estimated using a pretrained regressor. Three chemical databases (ChEMBL, ZINC, and COCONUT containing half a million to one million unlabeled molecules) are independently used for pretraining the generators to enrich them with valuable information from diverse chemical space. Standard RL methods are found to be insufficient, as learners tend to prioritize exploitation for immediate gains, resulting in repetitive generation of same/similar molecules. Our engineered reward function includes a Tanimoto-based uniqueness factor within the RL loop that improved the exploration of the environment and has helped accrue larger returns. Integration of a user-defined core fragment into the generated molecules facilitated learning of specific reaction types. Together, RE-EXPLORE can navigate the reaction space toward practically meaningful regions and offers notable improvements across the three distinct reaction types considered in this study. It identifies high-yielding substrates and highly enantioselective chiral catalysts. This RL-based approach has the potential to expedite reaction discovery and aid in the synthesis planning of important compounds, including drugs and pharmaceuticals.

PMID:39356950 | DOI:10.1021/jacs.4c08866

Sci Adv. 2024 Oct 4;10(40):eadp5491. doi: 10.1126/sciadv.adp5491. Epub 2024 Oct 2.

ABSTRACT

The PP2A-B55 phosphatase regulates a plethora of signaling pathways throughout eukaryotes. How PP2A-B55 selects its substrates presents a severe knowledge gap. By integrating AlphaFold modeling with comprehensive high-resolution mutational scanning, we show that α helices in substrates bind B55 through an evolutionary conserved mechanism. Despite a large diversity in sequence and composition, these α helices share key amino acid determinants that engage discrete hydrophobic and electrostatic patches. Using deep learning protein design, we generate a specific and potent competitive peptide inhibitor of PP2A-B55 substrate interactions. With this inhibitor, we uncover that PP2A-B55 regulates the nuclear exosome targeting (NEXT) complex by binding to an α-helical recruitment module in the RNA binding protein 7 (RBM7), a component of the NEXT complex. Collectively, our findings provide a framework for the understanding and interrogation of PP2A-B55 function in health and disease.

PMID:39356758 | DOI:10.1126/sciadv.adp5491

PLOS Digit Health. 2024 Oct 2;3(10):e0000588. doi: 10.1371/journal.pdig.0000588. eCollection 2024 Oct.

ABSTRACT

This paper introduces a novel approach to address the challenges associated with transferring blood pressure (BP) data obtained from oscillometric devices used in self-measured BP monitoring systems to integrate this data into medical health records or a proxy database accessible by clinicians, particularly in low literacy populations. To this end, we developed an automated image transcription technique to effectively transcribe readings from BP devices, ultimately enhancing the accessibility and usability of BP data for monitoring and managing BP during pregnancy and the postpartum period, particularly in low-resource settings and low-literate populations. In the designed study, the photos of the BP devices were captured as part of perinatal mobile health (mHealth) monitoring programs, conducted in four studies across two countries. The Guatemala Set 1 and Guatemala Set 2 datasets include the data captured by a cohort of 49 lay midwives from 1697 and 584 pregnant women carrying singletons in the second and third trimesters in rural Guatemala during routine screening. Additionally, we designed an mHealth system in Georgia for postpartum women to monitor and report their BP at home with 23 and 49 African American participants contributing to the Georgia I3 and Georgia IMPROVE projects, respectively. We developed a deep learning-based model which operates in two steps: LCD localization using the You Only Look Once (YOLO) object detection model and digit recognition using a convolutional neural network-based model capable of recognizing multiple digits. We applied color correction and thresholding techniques to minimize the impact of reflection and artifacts. Three experiments were conducted based on the devices used for training the digit recognition model. Overall, our results demonstrate that the device-specific model with transfer learning and the device independent model outperformed the device-specific model without transfer learning. The mean absolute error (MAE) of image transcription on held-out test datasets using the device-independent digit recognition were 1.2 and 0.8 mmHg for systolic and diastolic BP in the Georgia IMPROVE and 0.9 and 0.5 mmHg in Guatemala Set 2 datasets. The MAE, far below the FDA recommendation of 5 mmHg, makes the proposed automatic image transcription model suitable for general use when used with appropriate low-error BP devices.

PMID:39356720 | DOI:10.1371/journal.pdig.0000588

Plast Reconstr Surg. 2024 Oct 1. doi: 10.1097/PRS.0000000000011792. Online ahead of print.

ABSTRACT

BACKGROUND: Early detection of rare genetic diseases, including velocardiofacial syndrome (VCFS), is essential for patient well-being. However, their rarity and limited clinical experience of physicians make diagnosis challenging. Deep learning algorithms have emerged as promising tools for efficient and accurate diagnosis. This study investigates the use of a deep learning algorithm to develop a face recognition model for diagnosing VCFS.

METHODS: The study employed publicly available labeled face datasets to train the multitask cascaded convolutional neural networks (MTCNN) model. Subsequently, we examined the binary classification performance for diagnosing VCFS using the most efficient face recognition model. A total of 98 VCFS patients (920 facial photographs) and 91 non-VCFS controls (463 facial photographs) were randomly divided into training and test sets. Additionally, we analyzed whether the classification results matched the known facial phenotype of VCFS.

RESULTS: The face recognition model demonstrated high accuracy, ranging from 94% to 99%, depending on the training dataset. The accuracy of the binary classification diagnostic model varied from 81% to 88% when evaluating with photographs taken at various angles, but reached 95% evaluating with frontal photographs only. Gradient-weighted class activation mapping heatmap revealed the high importance level of perinasal and periorbital areas, exhibiting consistency with the conventional facial phenotypes of VCFS.

CONCLUSION: This study shows the feasibility and effectiveness of MTCNN-based model for detecting VCFS solely from facial photographs. The high accuracy underscores the potential of deep learning in aiding early diagnosis of rare genetic diseases, facilitating timely interventions for patient care.

PMID:39356705 | DOI:10.1097/PRS.0000000000011792