Front Surg. 2024 Apr 19;11:1370017. doi: 10.3389/fsurg.2024.1370017. eCollection 2024.

ABSTRACT

INTRODUCTION: The utilization of artificial intelligence (AI) augments intraoperative safety and surgical training. The recognition of parathyroid glands (PGs) is difficult for inexperienced surgeons. The aim of this study was to find out whether deep learning could be used to auxiliary identification of PGs on intraoperative videos in patients undergoing thyroid surgery.

METHODS: In this retrospective study, 50 patients undergoing thyroid surgery between 2021 and 2023 were randomly assigned (7:3 ratio) to a training cohort (n = 35) and a validation cohort (n = 15). The combined datasets included 98 videos with 9,944 annotated frames. An independent test cohort included 15 videos (1,500 frames) from an additional 15 patients. We developed a deep-learning model Video-Trans-U-HRNet to segment parathyroid glands in surgical videos, comparing it with three advanced medical AI methods on the internal validation cohort. Additionally, we assessed its performance against four surgeons (2 senior surgeons and 2 junior surgeons) on the independent test cohort, calculating precision and recall metrics for the model.

RESULTS: Our model demonstrated superior performance compared to other AI models on the internal validation cohort. The DICE and accuracy achieved by our model were 0.760 and 74.7% respectively, surpassing Video-TransUnet (0.710, 70.1%), Video-SwinUnet (0.754, 73.6%), and TransUnet (0.705, 69.4%). For the external test, our method got 89.5% precision 77.3% recall and 70.8% accuracy. In the statistical analysis, our model demonstrated results comparable to those of senior surgeons (senior surgeon 1: χ2 = 0.989, p = 0.320; senior surgeon 2: χ2 = 1.373, p = 0.241) and outperformed 2 junior surgeons (junior surgeon 1: χ2 = 3.889, p = 0.048; junior surgeon 2: χ2 = 4.763, p = 0.029).

DISCUSSION: We introduce an innovative intraoperative video method for identifying PGs, highlighting the potential advancements of AI in the surgical domain. The segmentation method employed for parathyroid glands in intraoperative videos offer surgeons supplementary guidance in locating real PGs. The method developed may have utility in facilitating training and decreasing the learning curve associated with the use of this technology.

PMID:38708363 | PMC:PMC11066234 | DOI:10.3389/fsurg.2024.1370017

Data Brief. 2024 Apr 12;54:110407. doi: 10.1016/j.dib.2024.110407. eCollection 2024 Jun.

ABSTRACT

Mathematical entity recognition is essential for machines to define and illustrate mathematical substance faultlessly and to facilitate sufficient mathematical operations and reasoning. As mathematical entity recognition in the Bangla language is novel, to our best knowledge, there is no available dataset exists in any repository. In this paper, we present state of the art Bangla mathematical entity dataset containing 13,717 observations. Each record has a mathematical statement, mathematical type and mathematical entity. This dataset can be utilized to conduct research involving the recognition of mathematical operators, renowned mathematical terms (such as complex numbers, real numbers, prime numbers, etc.), and operands as numbers. The findings mentioned above, and their combination are also feasible with a modest tweak to the dataset. Furthermore, we have structured this dataset in raw format and made a CSV file, incorporating three columns: text, math entity, and label. As an outcome, researchers may easily handle the data, facilitating a variety of deep learning and machine learning explorations.

PMID:38708312 | PMC:PMC11068499 | DOI:10.1016/j.dib.2024.110407

Data Brief. 2024 Apr 16;54:110433. doi: 10.1016/j.dib.2024.110433. eCollection 2024 Jun.

ABSTRACT

This paper is a description of a bird vocalisation dataset containing electronic recordings of birds in Uganda. The data was collected from 7 locations namely Bwindi impenetrable forest, Kibale forest national park, Matheniko game reserve, Moroto district, Kidepo National Park, Lake Mburo National Park and Murchison Falls National Park. The data was collected between May and June 2023. All together there are 570 recordings from 212 unique species amounting to more than 4 hours of audio. This represents a significant addition to the publicly available electronically recorded vocalisations for birds in Africa. The research was funded by Google Africa Research Collabs for the project entitled, "BASIS: Broad Avian Species Surveillance with Intelligent Sensing".

PMID:38708308 | PMC:PMC11067472 | DOI:10.1016/j.dib.2024.110433

Data Brief. 2024 Apr 15;54:110426. doi: 10.1016/j.dib.2024.110426. eCollection 2024 Jun.

ABSTRACT

Artificial Intelligence (AI) allows computers to self-develop decision-making algorithms through huge data analysis. In medical investigations, using computers to automatically diagnose diseases is a promising area of research that could change healthcare strategies worldwide. However, it can be challenging to reproduce or/and compare various approaches due to the often-limited datasets comprising medical images. Since there is no open access dataset for the Gallbladder (GB) organ, we introduce, in this study, a large dataset that includes 10,692 GB Ultrasound Images (UI) acquired at high resolution from 1,782 individuals. These UI include many disease types related to the GB, and they are organized around nine important anatomical landmarks. The data in this collection can be used to train machine learning (ML) and deep learning (DL) models for computer-aided detection of GB diseases. It can also help academics conduct comparative studies and test out novel techniques for analyzing UI to explore the medical domain of GB diseases. The objective is then to help move medical imaging forward so that patients get better treatment.

PMID:38708300 | PMC:PMC11068544 | DOI:10.1016/j.dib.2024.110426

J Clin Orthop Trauma. 2024 Apr 24;52:102421. doi: 10.1016/j.jcot.2024.102421. eCollection 2024 May.

ABSTRACT

BACKGROUND AND OBJECTIVES: Machine Learning (ML) and Deep Learning (DL) are novel technologies that can facilitate early diagnosis of Ankylosing Spondylitis (AS) and predict better patient-specific treatments. We aim to provide the current update on their use at different stages of AS diagnosis and treatment, describe different types of techniques used, dataset descriptions, contributions and limitations of existing work and ed to identify gaps in current knowledge for future works.

METHODS: We curated the data of this review from the PubMed database. We searched the full-text articles related to the use of ML/DL in the diagnosis and treatment of AS, for the period 2013-2023. Each article was manually scrutinized to be included or excluded for this review as per its relevance.

RESULTS: This review revealed that ML/DL technology is useful to assist and promote early diagnosis through AS patient characteristic profile creation, and identification of new AS-related biomarkers. They can help in forecasting the progression of AS and predict treatment responses to aid patient-specific treatment planning. However, there was a lack of sufficient-sized datasets sourced from multi-centres containing different types of diagnostic parameters. Also, there is less research on ML/DL-based AS treatment as compared to ML/DL-based AS diagnosis.

CONCLUSION: ML/DL can facilitate an early diagnosis and patient-tailored treatment for effective handling of AS. Benefits are especially higher in places with a lack of diagnostic resources and human experts. The use of ML/DL-trained models for AS diagnosis and treatment can provide the necessary support to the otherwise overwhelming healthcare systems in a cost-effective and timely way.

PMID:38708092 | PMC:PMC11063901 | DOI:10.1016/j.jcot.2024.102421

Phys Imaging Radiat Oncol. 2024 Apr 15;30:100577. doi: 10.1016/j.phro.2024.100577. eCollection 2024 Apr.

ABSTRACT

BACKGROUND AND PURPOSE: Radiation-induced erectile dysfunction (RiED) commonly affects prostate cancer patients, prompting clinical trials across institutions to explore dose-sparing to internal-pudendal-arteries (IPA) for preserving sexual potency. IPA, challenging to segment, isn't conventionally considered an organ-at-risk (OAR). This study proposes a deep learning (DL) auto-segmentation model for IPA, using Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) or CT alone to accommodate varied clinical practices.

MATERIALS AND METHODS: A total of 86 patients with CT and MRI images and noisy IPA labels were recruited in this study. We split the data into 42/14/30 for model training, testing, and a clinical observer study, respectively. There were three major innovations in this model: 1) we designed an architecture with squeeze-and-excite blocks and modality attention for effective feature extraction and production of accurate segmentation, 2) a novel loss function was used for training the model effectively with noisy labels, and 3) modality dropout strategy was used for making the model capable of segmentation in the absence of MRI.

RESULTS: Test dataset metrics were DSC 61.71 ± 7.7 %, ASD 2.5 ± .87 mm, and HD95 7.0 ± 2.3 mm. AI segmented contours showed dosimetric similarity to expert physician's contours. Observer study indicated higher scores for AI contours (mean = 3.7) compared to inexperienced physicians' contours (mean = 3.1). Inexperienced physicians improved scores to 3.7 when starting with AI contours.

CONCLUSION: The proposed model achieved good quality IPA contours to improve uniformity of segmentation and to facilitate introduction of standardized IPA segmentation into clinical trials and practice.

PMID:38707629 | PMC:PMC11068618 | DOI:10.1016/j.phro.2024.100577

Comput Struct Biotechnol J. 2024 Apr 15;23:1796-1807. doi: 10.1016/j.csbj.2024.04.032. eCollection 2024 Dec.

ABSTRACT

Protein subcellular localization prediction is of great significance in bioinformatics and biological research. Most of the proteins do not have experimentally determined localization information, computational prediction methods and tools have been acting as an active research area for more than two decades now. Knowledge of the subcellular location of a protein provides valuable information about its functionalities, the functioning of the cell, and other possible interactions with proteins. Fast, reliable, and accurate predictors provides platforms to harness the abundance of sequence data to predict subcellular locations accordingly. During the last decade, there has been a considerable amount of research effort aimed at developing subcellular localization predictors. This paper reviews recent subcellular localization prediction tools in the Eukaryotic, Prokaryotic, and Virus-based categories followed by a detailed analysis. Each predictor is discussed based on its main features, strengths, weaknesses, algorithms used, prediction techniques, and analysis. This review is supported by prediction tools taxonomies that highlight their rele- vant area and examples for uncomplicated categorization and ease of understandability. These taxonomies help users find suitable tools according to their needs. Furthermore, recent research gaps and challenges are discussed to cover areas that need the utmost attention. This survey provides an in-depth analysis of the most recent prediction tools to facilitate readers and can be considered a quick guide for researchers to identify and explore the recent literature advancements.

PMID:38707539 | PMC:PMC11066471 | DOI:10.1016/j.csbj.2024.04.032

Comput Struct Biotechnol J. 2024 Apr 17;23:1786-1795. doi: 10.1016/j.csbj.2024.04.039. eCollection 2024 Dec.

ABSTRACT

The rapid growth of spatially resolved transcriptomics technology provides new perspectives on spatial tissue architecture. Deep learning has been widely applied to derive useful representations for spatial transcriptome analysis. However, effectively integrating spatial multi-modal data remains challenging. Here, we present ConGcR, a contrastive learning-based model for integrating gene expression, spatial location, and tissue morphology for data representation and spatial tissue architecture identification. Graph convolution and ResNet were used as encoders for gene expression with spatial location and histological image inputs, respectively. We further enhanced ConGcR with a graph auto-encoder as ConGaR to better model spatially embedded representations. We validated our models using 16 human brains, four chicken hearts, eight breast tumors, and 30 human lung spatial transcriptomics samples. The results showed that our models generated more effective embeddings for obtaining tissue architectures closer to the ground truth than other methods. Overall, our models not only can improve tissue architecture identification's accuracy but also may provide valuable insights and effective data representation for other tasks in spatial transcriptome analyses.

PMID:38707535 | PMC:PMC11068546 | DOI:10.1016/j.csbj.2024.04.039

Heliyon. 2024 Apr 26;10(9):e30308. doi: 10.1016/j.heliyon.2024.e30308. eCollection 2024 May 15.

ABSTRACT

Pulmonary disease identification and characterization are among the most intriguing research topics of recent years since they require an accurate and prompt diagnosis. Although pulmonary radiography has helped in lung disease diagnosis, the interpretation of the radiographic image has always been a major concern for doctors and radiologists to reduce diagnosis errors. Due to their success in image classification and segmentation tasks, cutting-edge artificial intelligence techniques like machine learning (ML) and deep learning (DL) are widely encouraged to be applied in the field of diagnosing lung disorders and identifying them using medical images, particularly radiographic ones. For this end, the researchers are concurring to build systems based on these techniques in particular deep learning ones. In this paper, we proposed three deep-learning models that were trained to identify the presence of certain lung diseases using thoracic radiography. The first model, named "CovCXR-Net", identifies the COVID-19 disease (two cases: COVID-19 or normal). The second model, named "MDCXR3-Net", identifies the COVID-19 and pneumonia diseases (three cases: COVID-19, pneumonia, or normal), and the last model, named "MDCXR4-Net", is destined to identify the COVID-19, pneumonia and the pulmonary opacity diseases (4 cases: COVID-19, pneumonia, pulmonary opacity or normal). These models have proven their superiority in comparison with the state-of-the-art models and reached an accuracy of 99,09 %, 97.74 %, and 90,37 % respectively with three benchmarks.

PMID:38707425 | PMC:PMC11068804 | DOI:10.1016/j.heliyon.2024.e30308

Heliyon. 2024 Apr 23;10(9):e30107. doi: 10.1016/j.heliyon.2024.e30107. eCollection 2024 May 15.

ABSTRACT

Landslide susceptibility assessment (LSA) is fundamental for managing landslide geological disasters. This study presents a deep learning approach (DNN-MSFM) designed to enhance LSA modeling, particularly addressing limitations caused by the unbalanced distribution of data samples in applied datasets. DNN-MSFM approach combines a deep neural network (DNN) and a mean squared false misclassification loss function (MSFM) to handle unbalanced samples from the algorithmic perspective. The model's performance was evaluated on an unbalanced dataset containing mapping units' records of 293 landslide samples and 653 non-landslide samples from the Baota District, China. Its effectiveness was assessed through statistical metrics and compared against DNN and Support Vector Machine (SVM) basic models. The results demonstrated a significant performance enhancement from the DNN-MSFM (OverallAccuracy = 0.889 and area under the receiver operating characteristic curve (AUC) = 0.84), indicating its effectiveness in learning the underlying landslide susceptibility features and demonstrating its ability to provide improved predictions even in areas with unbalanced landslide samples. Moreover, the study emphasizes the importance of considering balanced loss functions in training DNN under various imbalance degrees and contributes to expanding the applicability of DNN in LSA modeling. Also, this study builds a foundation for further enhancements of deep learning methods for geological disaster assessments.

PMID:38707366 | PMC:PMC11068606 | DOI:10.1016/j.heliyon.2024.e30107

Biomed Eng Online. 2024 May 5;23(1):45. doi: 10.1186/s12938-024-01240-0.

ABSTRACT

BACKGROUND: Sleep-disordered breathing (SDB) affects a significant portion of the population. As such, there is a need for accessible and affordable assessment methods for diagnosis but also case-finding and long-term follow-up. Research has focused on exploiting cardiac and respiratory signals to extract proxy measures for sleep combined with SDB event detection. We introduce a novel multi-task model combining cardiac activity and respiratory effort to perform sleep-wake classification and SDB event detection in order to automatically estimate the apnea-hypopnea index (AHI) as severity indicator.

METHODS: The proposed multi-task model utilized both convolutional and recurrent neural networks and was formed by a shared part for common feature extraction, a task-specific part for sleep-wake classification, and a task-specific part for SDB event detection. The model was trained with RR intervals derived from electrocardiogram and respiratory effort signals. To assess performance, overnight polysomnography (PSG) recordings from 198 patients with varying degree of SDB were included, with manually annotated sleep stages and SDB events.

RESULTS: We achieved a Cohen's kappa of 0.70 in the sleep-wake classification task, corresponding to a Spearman's correlation coefficient (R) of 0.830 between the estimated total sleep time (TST) and the TST obtained from PSG-based sleep scoring. Combining the sleep-wake classification and SDB detection results of the multi-task model, we obtained an R of 0.891 between the estimated and the reference AHI. For severity classification of SBD groups based on AHI, a Cohen's kappa of 0.58 was achieved. The multi-task model performed better than a single-task model proposed in a previous study for AHI estimation, in particular for patients with a lower sleep efficiency (R of 0.861 with the multi-task model and R of 0.746 with single-task model with subjects having sleep efficiency < 60%).

CONCLUSION: Assisted with automatic sleep-wake classification, our multi-task model demonstrated proficiency in estimating AHI and assessing SDB severity based on AHI in a fully automatic manner using RR intervals and respiratory effort. This shows the potential for improving SDB screening with unobtrusive sensors also for subjects with low sleep efficiency without adding additional sensors for sleep-wake detection.

PMID:38705982 | DOI:10.1186/s12938-024-01240-0

Med Biol Eng Comput. 2024 May 6. doi: 10.1007/s11517-024-03116-w. Online ahead of print.

ABSTRACT

Among the various physiological signals, electrocardiogram (ECG) is a valid criterion for the classification of various exercise fatigue. In this study, we combine features extracted by deep neural networks with linear features from ECG and heart rate variability (HRV) for exercise fatigue classification. First, the ECG signals are converted into 2-D images by using the short-term Fourier transform (STFT), and image features are extracted by the visual geometry group (VGG) . The extracted image and linear features of ECG and HRV are sent to the different types of classifiers to distinguish distinct exercise fatigue level. To validate performance, the proposed methods are tested on (i) an open-source EPHNOGRAM dataset and (ii) a self-collected dataset (n = 51). The results reveal that the classification based on the concatenated features has the highest accuracy, and the calculation time of the system is also significantly reduced. This demonstrates that the proposed novel hybrid approach can be used to assist in improving the accuracy and timeliness of exercise fatigue classification in a real-time exercise environment. The experimental results show that the proposed method outperforms other recent state-of-the-art methods in terms of accuracy 96.90%, sensitivity 96.90%, F1-score of 0.9687 in EPHNOGRAM and accuracy 92.17%, sensitivity 92.63%, F1-score of 0.9213 in self-collected dataset.

PMID:38705958 | DOI:10.1007/s11517-024-03116-w

Sci Rep. 2024 May 5;14(1):10306. doi: 10.1038/s41598-024-60839-4.

ABSTRACT

Multiple ophthalmic diseases lead to decreased capillary perfusion that can be visualized using optical coherence tomography angiography images. To quantify the decrease in perfusion, past studies have often used the vessel density, which is the percentage of vessel pixels in the image. However, this method is often not sensitive enough to detect subtle changes in early pathology. More recent methods are based on quantifying non-perfused or intercapillary areas between the vessels. These methods rely upon the accuracy of vessel segmentation, which is a challenging task and therefore a limiting factor for reliability. Intercapillary areas computed from perfusion-distance measures are less sensitive to errors in the vessel segmentation since the distance to the next vessel is only slightly changing if gaps are present in the segmentation. We present a novel method for distinguishing between glaucoma patients and healthy controls based on features computed from the probability density function of these perfusion-distance areas. The proposed approach is evaluated on different capillary plexuses and outperforms previously proposed methods that use handcrafted features for classification. Moreover the results of the proposed method are in the same range as the ones of convolutional neural networks trained on the raw input images and is therefore a computationally efficient, simple to implement and explainable alternative to deep learning-based approaches.

PMID:38705883 | DOI:10.1038/s41598-024-60839-4

Sci Rep. 2024 May 6;14(1):10319. doi: 10.1038/s41598-024-61251-8.

ABSTRACT

Seismic imaging techniques play a crucial role in interpreting subsurface geological structures by analyzing the propagation and reflection of seismic waves. However, traditional methods face challenges in achieving high resolution due to theoretical constraints and computational costs. Leveraging recent advancements in deep learning, this study introduces a neural network framework that integrates Transformer and Convolutional Neural Network (CNN) architectures, enhanced through Adaptive Spatial Feature Fusion (ASFF), to achieve high-resolution seismic imaging. Our approach directly maps seismic data to reflection models, eliminating the need for post-processing low-resolution results. Through extensive numerical experiments, we demonstrate the outstanding ability of this method to accurately infer subsurface structures. Evaluation metrics including Root Mean Square Error (RMSE), Correlation Coefficient (CC), and Structural Similarity Index (SSIM) emphasize the model's capacity to faithfully reconstruct subsurface features. Furthermore, noise injection experiments showcase the reliability of this efficient seismic imaging method, further underscoring the potential of deep learning in seismic imaging.

PMID:38705877 | DOI:10.1038/s41598-024-61251-8

Methods. 2024 May 3:S1046-2023(24)00108-7. doi: 10.1016/j.ymeth.2024.04.018. Online ahead of print.

ABSTRACT

Messenger RNA (mRNA) is vital for post-transcriptional gene regulation, acting as the direct template for protein synthesis. However, the methods available for predicting mRNA subcellular localization need to be improved and enhanced. Notably, few existing algorithms can annotate mRNA sequences with multiple localizations. In this work, we propose the mRNA-CLA, an innovative multi-label subcellular localization prediction framework for mRNA, leveraging a deep learning approach with a multi-head self-attention mechanism. The framework employs a multi-scale convolutional layer to extract sequence features across different regions and uses a self-attention mechanism explicitly designed for each sequence. Paired with Position Weight Matrices (PWMs) derived from the convolutional neural network layers, our model offers interpretability in the analysis. In particular, we perform a base-level analysis of mRNA sequences from diverse subcellular localizations to determine the nucleotide specificity corresponding to each site. Our evaluations demonstrate that the mRNA-CLA model substantially outperforms existing methods and tools.

PMID:38705502 | DOI:10.1016/j.ymeth.2024.04.018

J Neurosci Methods. 2024 May 3:110157. doi: 10.1016/j.jneumeth.2024.110157. Online ahead of print.

ABSTRACT

BACKGROUND: Autism classification work on fNIRS data using dynamic graph networks. Explore the impact of the dynamic connection relationship between brain channels on ASD, and compare the brain channel connection diagrams of ASD and TD to explore potential factors that influence the development of autism.

METHOD: Using dynamic graph construction to mine the dynamic relationships of fNIRS data, obtain spatio-temporal correlations through dynamic feature extraction, and improve the information extraction capabilities of the network through spatio-temporal graph pooling to achieve classification of ASD.

RESULT: A classification effect with an accuracy of 97.2% was achieved using a short sequence of 1.75s. The results showed that the dynamic connections of channel 5 and 19, channel 12 and 25, and channel 7 and 34 have a greater impact on the classification of autism. Comparison with previously used method(s): Compared with previous deep learning models, our model achieves efficient classification using short-term fNIRS data of 1.75s, and analyzes the impact of dynamic connections on classification through dynamic graphs.

CONCLUSION: Using Dynamic Spatio-Temporal Graph Pooled Neural Networks (DSTGPN), dynamic connectivity between brain channels was found to have an impact on the classification of autism. By modelling the brain channel relationship maps of ASD and TD, hyperlink clusters were found to exist on the brain channel connections of ASD.

PMID:38705284 | DOI:10.1016/j.jneumeth.2024.110157

Int J Pharm. 2024 May 3:124188. doi: 10.1016/j.ijpharm.2024.124188. Online ahead of print.

ABSTRACT

Orodispersible films (ODFs) have emerged as innovative pharmaceutical dosage forms, offering patient-specific treatment through adjustable dosing and the combination of diverse active ingredients. This expanding field generates vast datasets, requiring advanced analytical techniques for deeper understanding of data itself. Machine learning is becoming an important tool in the rapidly changing field of pharmaceutical research, particularly in drug preformulation studies. This work aims to explore into the application of machine learning methods for the analysis of experimental data obtained by ODF characterization in order to obtain an insight into the factors governing ODF performance and use it as guidance in pharmaceutical development. Using a dataset derived from extensive experimental studies, various machine learning algorithms were employed to cluster and predict critical properties of ODFs. Our results demonstrate that machine learning models, including Support vector machine, Random forest and Deep learning, exhibit high accuracy in predicting the mechanical properties of ODFs, such as flexibility and rigidity. The predictive models offered insights into the complex interaction of formulation variables. This research is a pilot study that highlights the potential of machine learning as a transformative approach in the pharmaceutical field, paving the way for more efficient and informed drug development processes.

PMID:38705248 | DOI:10.1016/j.ijpharm.2024.124188

Philos Trans R Soc Lond B Biol Sci. 2024 Jun 24;379(1904):20230444. doi: 10.1098/rstb.2023.0444. Epub 2024 May 6.

ABSTRACT

Passive acoustic monitoring (PAM) is a powerful tool for studying ecosystems. However, its effective application in tropical environments, particularly for insects, poses distinct challenges. Neotropical katydids produce complex species-specific calls, spanning mere milliseconds to seconds and spread across broad audible and ultrasonic frequencies. However, subtle differences in inter-pulse intervals or central frequencies are often the only discriminatory traits. These extremities, coupled with low source levels and susceptibility to masking by ambient noise, challenge species identification in PAM recordings. This study aimed to develop a deep learning-based solution to automate the recognition of 31 katydid species of interest in a biodiverse Panamanian forest with over 80 katydid species. Besides the innate challenges, our efforts were also encumbered by a limited and imbalanced initial training dataset comprising domain-mismatched recordings. To overcome these, we applied rigorous data engineering, improving input variance through controlled playback re-recordings and by employing physics-based data augmentation techniques, and tuning signal-processing, model and training parameters to produce a custom well-fit solution. Methods developed here are incorporated into Koogu, an open-source Python-based toolbox for developing deep learning-based bioacoustic analysis solutions. The parametric implementations offer a valuable resource, enhancing the capabilities of PAM for studying insects in tropical ecosystems. This article is part of the theme issue 'Towards a toolkit for global insect biodiversity monitoring'.

PMID:38705172 | DOI:10.1098/rstb.2023.0444

Curr Opin Struct Biol. 2024 May 4;86:102827. doi: 10.1016/j.sbi.2024.102827. Online ahead of print.

ABSTRACT

In this mini-review, we explore the new prediction methods for drug combination synergy relying on high-throughput combinatorial screens. The fast progress of the field is witnessed in the more than thirty original machine learning methods published since 2021, a clear majority of them based on deep learning techniques. We aim to put these articles under a unifying lens by highlighting the core technologies, the data sources, the input data types and synergy scores used in the methods, as well as the prediction scenarios and evaluation protocols that the articles deal with. Our finding is that the best methods accurately solve the synergy prediction scenarios involving known drugs or cell lines while the scenarios involving new drugs or cell lines still fall short of an accurate prediction level.

PMID:38705070 | DOI:10.1016/j.sbi.2024.102827

Med Image Anal. 2024 Apr 25;95:103187. doi: 10.1016/j.media.2024.103187. Online ahead of print.

ABSTRACT

Domain shift problem is commonplace for ultrasound image analysis due to difference imaging setting and diverse medical centers, which lead to poor generalizability of deep learning-based methods. Multi-Source Domain Transformation (MSDT) provides a promising way to tackle the performance degeneration caused by the domain shift, which is more practical and challenging compared to conventional single-source transformation tasks. An effective unsupervised domain combination strategy is highly required to handle multiple domains without annotations. Fidelity and quality of generated images are also important to ensure the accuracy of computer-aided diagnosis. However, existing MSDT approaches underperform in above two areas. In this paper, an efficient domain transformation model named M2O-DiffGAN is introduced to achieve a unified mapping from multiple unlabeled source domains to the target domain. A cycle-consistent "many-to-one" adversarial learning architecture is introduced to model various unlabeled domains jointly. A condition adversarial diffusion process is employed to generate images with high-fidelity, combining an adversarial projector to capture reverse transition probabilities over large step sizes for accelerating sampling. Considering the limited perceptual information of ultrasound images, an ultrasound-specific content loss helps to capture more perceptual features for synthesizing high-quality ultrasound images. Massive comparisons on six clinical datasets covering thyroid, carotid and breast demonstrate the superiority of the M2O-DiffGAN in the performance of bridging the domain gaps and enlarging the generalization of downstream analysis methods compared to state-of-the-art algorithms. It improves the mean MI, Bhattacharyya Coefficient, dice and IoU assessments by 0.390, 0.120, 0.245 and 0.250, presenting promising clinical applications.

PMID:38705056 | DOI:10.1016/j.media.2024.103187