J Med Imaging (Bellingham). 2025 Jan;12(1):014501. doi: 10.1117/1.JMI.12.1.014501. Epub 2025 Jan 6.

ABSTRACT

PURPOSE: The prevalence of type 2 diabetes mellitus (T2DM) has been steadily increasing over the years. We aim to predict the occurrence of T2DM using mammography images within 5 years using two different methods and compare their performance.

APPROACH: We examined 312 samples, including 110 positive cases (developed T2DM after 5 years) and 202 negative cases (did not develop T2DM) using two different methods. In the first method, a radiomics-based approach, we utilized radiomics features and machine learning (ML) algorithms. The entire breast region was chosen as the region of interest for extracting radiomics features. Then, a binary breast image was created from which we extracted 668 features and analyzed them using various ML algorithms. In the second method, a complex convolutional neural network (CNN) with a modified ResNet architecture and various kernel sizes was applied to raw mammography images for the prediction task. A nested, stratified five-fold cross-validation was done for both parts A and B to compute accuracy, sensitivity, specificity, and area under the receiver operating curve (AUROC). Hyperparameter tuning was also done to enhance the model's performance and reliability.

RESULTS: The radiomics approach's light gradient boosting model gave 68.9% accuracy, 30.7% sensitivity, 89.5% specificity, and 0.63 AUROC. The CNN method achieved an AUROC of 0.58 over 20 epochs.

CONCLUSION: Radiomics outperformed CNN by 0.05 in terms of AUROC. This may be due to the more straightforward interpretability and clinical relevance of predefined radiomics features compared with the complex, abstract features learned by CNNs.

PMID:39776665 | PMC:PMC11702674 | DOI:10.1117/1.JMI.12.1.014501

Biol Imaging. 2024 Nov 15;4:e13. doi: 10.1017/S2633903X24000096. eCollection 2024.

ABSTRACT

We propose a neural network architecture and a training procedure to estimate blurring operators and deblur images from a single degraded image. Our key assumption is that the forward operators can be parameterized by a low-dimensional vector. The models we consider include a description of the point spread function with Zernike polynomials in the pupil plane or product-convolution expansions, which incorporate space-varying operators. Numerical experiments show that the proposed method can accurately and robustly recover the blur parameters even for large noise levels. For a convolution model, the average signal-to-noise ratio of the recovered point spread function ranges from 13 dB in the noiseless regime to 8 dB in the high-noise regime. In comparison, the tested alternatives yield negative values. This operator estimate can then be used as an input for an unrolled neural network to deblur the image. Quantitative experiments on synthetic data demonstrate that this method outperforms other commonly used methods both perceptually and in terms of SSIM. The algorithm can process a 512 512 image under a second on a consumer graphics card and does not require any human interaction once the operator parameterization has been set up.1.

PMID:39776610 | PMC:PMC11704139 | DOI:10.1017/S2633903X24000096

Biol Imaging. 2024 Dec 20;4:e16. doi: 10.1017/S2633903X24000151. eCollection 2024.

ABSTRACT

With the fast development of modern microscopes and bioimaging techniques, an unprecedentedly large amount of imaging data is being generated, stored, analyzed, and shared through networks. The size of the data poses great challenges for current data infrastructure. One common way to reduce the data size is by image compression. This study analyzes multiple classic and deep-learning-based image compression methods, as well as an empirical study on their impact on downstream deep-learning-based image processing models. We used deep-learning-based label-free prediction models (i.e., predicting fluorescent images from bright-field images) as an example downstream task for the comparison and analysis of the impact of image compression. Different compression techniques are compared in compression ratio, image similarity, and, most importantly, the prediction accuracy of label-free models on original and compressed images. We found that artificial intelligence (AI)-based compression techniques largely outperform the classic ones with minimal influence on the downstream 2D label-free tasks. In the end, we hope this study could shed light on the potential of deep-learning-based image compression and raise the awareness of the potential impacts of image compression on downstream deep-learning models for analysis.

PMID:39776609 | PMC:PMC11704128 | DOI:10.1017/S2633903X24000151

Biol Imaging. 2024 Nov 22;4:e14. doi: 10.1017/S2633903X24000114. eCollection 2024.

ABSTRACT

This manuscript showcases the latest advancements in deepImageJ, a pivotal Fiji/ImageJ plugin for bioimage analysis in life sciences. The plugin, known for its user-friendly interface, facilitates the application of diverse pre-trained convolutional neural networks to custom data. The manuscript demonstrates several deepImageJ capabilities, particularly in deploying complex pipelines, three-dimensional (3D) image analysis, and processing large images. A key development is the integration of the Java Deep Learning Library, expanding deepImageJ's compatibility with various deep learning (DL) frameworks, including TensorFlow, PyTorch, and ONNX. This allows for running multiple engines within a single Fiji/ImageJ instance, streamlining complex bioimage analysis workflows. The manuscript details three case studies to demonstrate these capabilities. The first case study explores integrated image-to-image translation followed by nuclei segmentation. The second case study focuses on 3D nuclei segmentation. The third case study showcases large image volume segmentation and compatibility with the BioImage Model Zoo. These use cases underscore deepImageJ's versatility and power to make advanced DLmore accessible and efficient for bioimage analysis. The new developments within deepImageJ seek to provide a more flexible and enriched user-friendly framework to enable next-generation image processing in life science.

PMID:39776608 | PMC:PMC11704127 | DOI:10.1017/S2633903X24000114

Mach Learn Clin Neuroimaging (2024). 2025;15266:13-23. doi: 10.1007/978-3-031-78761-4_2. Epub 2024 Dec 6.

ABSTRACT

Quantitative Susceptibility Mapping (QSM) is a technique that derives tissue magnetic susceptibility distributions from phase measurements obtained through Magnetic Resonance (MR) imaging. This involves solving an ill-posed dipole inversion problem, however, and thus time-consuming and cumbersome data acquisition from several distinct head orientations becomes necessary to obtain an accurate solution. Most recent (supervised) deep learning methods for single-phase QSM require training data obtained via multiple orientations. In this work, we present an alternative unsupervised learning approach that can efficiently train on single-orientation measurement data alone, named ProxiMO (Proximal Multi-Operator), combining Learned Proximal Convolutional Neural Networks (LP-CNN) with multi-operator imaging (MOI). This integration enables LP-CNN training for QSM on single-phase data without ground truth reconstructions. We further introduce a semi-supervised variant, which further boosts the reconstruction performance, compared to the traditional supervised fashions. Extensive experiments on multicenter datasets illustrate the advantage of unsupervised training and the superiority of the proposed approach for QSM reconstruction. Code is available at https://github.com/shmuelor/ProxiMO.

PMID:39776602 | PMC:PMC11705005 | DOI:10.1007/978-3-031-78761-4_2

Viruses. 2024 Nov 29;16(12):1864. doi: 10.3390/v16121864.

ABSTRACT

In this study, we introduce a novel approach that integrates interpretability techniques from both traditional machine learning (ML) and deep neural networks (DNN) to quantify feature importance using global and local interpretation methods. Our method bridges the gap between interpretable ML models and powerful deep learning (DL) architectures, providing comprehensive insights into the key drivers behind model predictions, especially in detecting outliers within medical data. We applied this method to analyze COVID-19 pandemic data from 2020, yielding intriguing insights. We used a dataset consisting of individuals who were tested for COVID-19 during the early stages of the pandemic in 2020. The dataset included self-reported symptoms and test results from a wide demographic, and our goal was to identify the most important symptoms that could help predict COVID-19 infection accurately. By applying interpretability techniques to both machine learning and deep learning models, we aimed to improve understanding of symptomatology and enhance early detection of COVID-19 cases. Notably, even though less than 1% of our cohort reported having a sore throat, this symptom emerged as a significant indicator of active COVID-19 infection, appearing 7 out of 9 times in the top four most important features across all methodologies. This suggests its potential as an early symptom marker. Studies have shown that individuals reporting sore throat may have a compromised immune system, where antibody generation is not functioning correctly. This aligns with our data, which indicates that 5% of patients with sore throats required hospitalization. Our analysis also revealed a concerning trend of diminished immune response post-COVID infection, increasing the likelihood of severe cases requiring hospitalization. This finding underscores the importance of monitoring patients post-recovery for potential complications and tailoring medical interventions accordingly. Our study also raises critical questions about the efficacy of COVID-19 vaccines in individuals presenting with sore throat as a symptom. The results suggest that booster shots might be necessary for this population to ensure adequate immunity, given the observed immune response patterns. The proposed method not only enhances our understanding of COVID-19 symptomatology but also demonstrates its broader utility in medical outlier detection. This research contributes valuable insights to ongoing efforts in creating interpretable models for COVID-19 management and vaccine optimization strategies. By leveraging feature importance and interpretability, these models empower physicians, healthcare workers, and researchers to understand complex relationships within medical data, facilitating more informed decision-making for patient care and public health initiatives.

PMID:39772174 | DOI:10.3390/v16121864

Sensors (Basel). 2024 Dec 23;24(24):8220. doi: 10.3390/s24248220.

ABSTRACT

To address the limitations of existing deep learning-based algorithms in detecting surface defects on brake pipe ends, a novel lightweight detection algorithm, FP-YOLOv8, is proposed. This algorithm is developed based on the YOLOv8n framework with the aim of improving accuracy and model lightweight design. First, the C2f_GhostV2 module has been designed to replace the original C2f module. It reduces the model's parameter count through its unique design. It achieves improved feature representation by adopting specific technique within its structure. Additionally, it incorporates the decoupled fully connected (DFC) attention mechanism, which minimizes information loss during long-range feature transmission by separately capturing pixel information along horizontal and vertical axes via convolution. Second, the Dynamic ATSS label allocation strategy is applied, which dynamically adjusts label assignments by integrating Anchor IoUs and predicted IoUs, effectively reducing the misclassification of high-quality prediction samples as negative samples. Thus, it improves the detection accuracy of the model. Lastly, an asymmetric small-target detection head, FADH, is proposed to utilize depth-separable convolution to accomplish classification and regression tasks, enabling more precise capture of detailed information across scales and improving the detection of small-target defects. The experimental results show that FP-YOLOv8 achieves a mAP50 of 89.5% and an F1-score of 87% on the ends surface defects dataset, representing improvements of 3.3% and 6.0%, respectively, over the YOLOv8n algorithm, Meanwhile, it reduces model parameters and computational costs by 14.3% and 21.0%. Additionally, compared to the baseline model, the AP50 values for cracks, scratches, and flash defects rise by 5.5%, 5.6%, and 2.3%, respectively. These results validate the efficacy of FP-YOLOv8 in enhancing defect detection accuracy, reducing missed detection rates, and decreasing model parameter counts and computational demands, thus meeting the requirements of online defect detection for brake pipe ends surfaces.

PMID:39771953 | DOI:10.3390/s24248220

Sensors (Basel). 2024 Dec 23;24(24):8217. doi: 10.3390/s24248217.

ABSTRACT

Multi-modal systems extract information about the environment using specialized sensors that are optimized based on the wavelength of the phenomenology and material interactions. To maximize the entropy, complementary systems operating in regions of non-overlapping wavelengths are optimal. VIS-IR (Visible-Infrared) systems have been at the forefront of multi-modal fusion research and are used extensively to represent information in all-day all-weather applications. Prior to image fusion, the image pairs have to be properly registered and mapped to a common resolution palette. However, due to differences in the device physics of image capture, information from VIS-IR sensors cannot be directly correlated, which is a major bottleneck for this area of research. In the absence of camera metadata, image registration is performed manually, which is not practical for large datasets. Most of the work published in this area assumes calibrated sensors and the availability of camera metadata providing registered image pairs, which limits the generalization capability of these systems. In this work, we propose a novel end-to-end pipeline termed DeepFusion for image registration and fusion. Firstly, we design a recursive crop and scale wavelet spectral decomposition (WSD) algorithm for automatically extracting the patch of visible data representing the thermal information. After data extraction, both the images are registered to a common resolution palette and forwarded to the DNN for image fusion. The fusion performance of the proposed pipeline is compared and quantified with state-of-the-art classical and DNN architectures for open-source and custom datasets demonstrating the efficacy of the pipeline. Furthermore, we also propose a novel keypoint-based metric for quantifying the quality of fused output.

PMID:39771950 | DOI:10.3390/s24248217

Sensors (Basel). 2024 Dec 23;24(24):8207. doi: 10.3390/s24248207.

ABSTRACT

Aiming at the severe occlusion problem and the tiny-scale object problem in the multi-fitting detection task, the Scene Knowledge Integrating Network (SKIN), including the scene filter module (SFM) and scene structure information module (SSIM) is proposed. Firstly, the particularity of the scene in the multi-fitting detection task is analyzed. Hence, the aggregation of the fittings is defined as the scene according to the professional knowledge of the power field and the habit of the operators in identifying the fittings. So, the scene knowledge will include global context information, fitting fine-grained visual information and scene structure information. Then, a scene filter module is designed to learn the global context information and fitting fine-grained visual information, and a scene structure module is designed to learn the scene structure information. Finally, the scene semantic features are used as the carrier to integrate three categories of information into the relative scene features, which can assist in the recognition of the occluded fittings and the tiny-scale fittings after feature mining and feature integration. The experiments show that the proposed network can effectively improve the performance of the multi-fitting detection task compared with the Faster R-CNN and other state-of-the-art models. In particular, the detection performances of the occluded and tiny-scale fittings are significantly improved.

PMID:39771941 | DOI:10.3390/s24248207

Sensors (Basel). 2024 Dec 21;24(24):8172. doi: 10.3390/s24248172.

ABSTRACT

Grapevines (Vitis vinifera L.) are one of the most economically relevant crops worldwide, yet they are highly vulnerable to various diseases, causing substantial economic losses for winegrowers. This systematic review evaluates the application of remote sensing and proximal tools for vineyard disease detection, addressing current capabilities, gaps, and future directions in sensor-based field monitoring of grapevine diseases. The review covers 104 studies published between 2008 and October 2024, identified through searches in Scopus and Web of Science, conducted on 25 January 2024, and updated on 10 October 2024. The included studies focused exclusively on the sensor-based detection of grapevine diseases, while excluded studies were not related to grapevine diseases, did not use remote or proximal sensing, or were not conducted in field conditions. The most studied diseases include downy mildew, powdery mildew, Flavescence dorée, esca complex, rots, and viral diseases. The main sensors identified for disease detection are RGB, multispectral, hyperspectral sensors, and field spectroscopy. A trend identified in recent published research is the integration of artificial intelligence techniques, such as machine learning and deep learning, to improve disease detection accuracy. The results demonstrate progress in sensor-based disease monitoring, with most studies concentrating on specific diseases, sensor platforms, or methodological improvements. Future research should focus on standardizing methodologies, integrating multi-sensor data, and validating approaches across diverse vineyard contexts to improve commercial applicability and sustainability, addressing both economic and environmental challenges.

PMID:39771913 | DOI:10.3390/s24248172

Sensors (Basel). 2024 Dec 21;24(24):8168. doi: 10.3390/s24248168.

ABSTRACT

This systematic review examines EEG-based imagined speech classification, emphasizing directional words essential for development in the brain-computer interface (BCI). This study employed a structured methodology to analyze approaches using public datasets, ensuring systematic evaluation and validation of results. This review highlights the feature extraction techniques that are pivotal to classification performance. These include deep learning, adaptive optimization, and frequency-specific decomposition, which enhance accuracy and robustness. Classification methods were explored by comparing traditional machine learning with deep learning and emphasizing the role of brain lateralization in imagined speech for effective recognition and classification. This study discusses the challenges of generalizability and scalability in imagined speech recognition, focusing on subject-independent approaches and multiclass scalability. Performance benchmarking across various datasets and methodologies revealed varied classification accuracies, reflecting the complexity and variability of EEG signals. This review concludes that challenges remain despite progress, particularly in classifying directional words. Future research directions include improved signal processing techniques, advanced neural network architectures, and more personalized, adaptive BCI systems. This review is critical for future efforts to develop practical communication tools for individuals with speech and motor impairments using EEG-based BCIs.

PMID:39771903 | DOI:10.3390/s24248168

Sensors (Basel). 2024 Dec 21;24(24):8167. doi: 10.3390/s24248167.

ABSTRACT

Freezing of wind turbines causes loss of wind-generated power. Forecasting or prediction of icing on wind turbine blades based on SCADA sensor data allows taking appropriate actions before icing occurs. This paper presents a newly developed deep learning network model named PCTG (Parallel CNN-TCN GRU) for the purpose of high-accuracy and long-term prediction of icing on wind turbine blades. This model combines three networks, the CNN, TCN, and GRU, in order to incorporate both the temporal aspect of SCADA time-series data as well as the dependencies of SCADA variables. The experimentations conducted by using this model and SCADA data from three wind turbines in a wind farm have generated average prediction accuracies of about 97% for prediction horizons of up to 2 days ahead. The developed model is shown to maintain at least 95% prediction accuracy for long prediction horizons of up to 22 days ahead. Furthermore, for another wind farm SCADA dataset, it is shown that the developed PCTG model achieves over 99% icing prediction accuracy 10 days ahead.

PMID:39771901 | DOI:10.3390/s24248167

Sensors (Basel). 2024 Dec 21;24(24):8164. doi: 10.3390/s24248164.

ABSTRACT

Simultaneous localization and mapping (SLAM) techniques can be used to navigate the visually impaired, but the development of robust SLAM solutions for crowded spaces is limited by the lack of realistic datasets. To address this, we introduce InCrowd-VI, a novel visual-inertial dataset specifically designed for human navigation in indoor pedestrian-rich environments. Recorded using Meta Aria Project glasses, it captures realistic scenarios without environmental control. InCrowd-VI features 58 sequences totaling a 5 km trajectory length and 1.5 h of recording time, including RGB, stereo images, and IMU measurements. The dataset captures important challenges such as pedestrian occlusions, varying crowd densities, complex layouts, and lighting changes. Ground-truth trajectories, accurate to approximately 2 cm, are provided in the dataset, originating from the Meta Aria project machine perception SLAM service. In addition, a semi-dense 3D point cloud of scenes is provided for each sequence. The evaluation of state-of-the-art visual odometry (VO) and SLAM algorithms on InCrowd-VI revealed severe performance limitations in these realistic scenarios. Under challenging conditions, systems exceeded the required localization accuracy of 0.5 m and the 1% drift threshold, with classical methods showing drift up to 5-10%. While deep learning-based approaches maintained high pose estimation coverage (>90%), they failed to achieve real-time processing speeds necessary for walking pace navigation. These results demonstrate the need and value of a new dataset to advance SLAM research for visually impaired navigation in complex indoor environments.

PMID:39771900 | DOI:10.3390/s24248164

Sensors (Basel). 2024 Dec 20;24(24):8157. doi: 10.3390/s24248157.

ABSTRACT

Rolling bearings play a crucial role in industrial equipment, and their failure is highly likely to cause a series of serious consequences. Traditional deep learning-based bearing fault diagnosis algorithms rely on large amounts of training data; training and inference processes consume significant computational resources. Thus, developing a lightweight and suitable fault diagnosis algorithm for small samples is particularly crucial. In this paper, we propose a bearing fault diagnosis algorithm based on probabilistic decoupling knowledge distillation and meta-learning (MIX-MPDKD). This algorithm is lightweight and deployable, performing well in small sample scenarios and effectively solving the deployment problem of large networks in resource-constrained environments. Firstly, our model utilizes the Model-Agnostic Meta-Learning algorithm to initialize the parameters of the teacher model and conduct efficient training. Subsequently, by employing the proposed probability-based decoupled knowledge distillation approach, the outstanding performance of the teacher model was imparted to the student model, enabling the student model to converge rapidly in the context of a small sample size. Finally, the Paderborn University dataset was used for meta-training, while the bearing dataset from Case Western Reserve University, along with our laboratory dataset, was used to validate the results. The experimental results demonstrate that the algorithm achieved satisfactory accuracy performance.

PMID:39771892 | DOI:10.3390/s24248157

Sensors (Basel). 2024 Dec 20;24(24):8139. doi: 10.3390/s24248139.

ABSTRACT

We demonstrate high-resolution single-pixel imaging (SPI) in the visible and near-infrared wavelength ranges using an SPI framework that incorporates a novel, dedicated sampling scheme and a reconstruction algorithm optimized for the rapid imaging of highly sparse scenes at the native digital micromirror device (DMD) resolution of 1024 × 768. The reconstruction algorithm consists of two stages. In the first stage, the vector of SPI measurements is multiplied by the generalized inverse of the measurement matrix. In the second stage, we compare two reconstruction approaches: one based on an iterative algorithm and the other on a trained neural network. The neural network outperforms the iterative method when the object resembles the training set, though it lacks the generality of the iterative approach. For images captured at a compression of 0.41 percent, corresponding to a measurement rate of 6.8 Hz with a DMD operating at 22 kHz, the typical reconstruction time on a desktop with a medium-performance GPU is comparable to the image acquisition rate. This allows the proposed SPI method to support high-resolution dynamic SPI in a variety of applications, using a standard SPI architecture with a DMD modulator operating at its native resolution and bandwidth, and enabling the real-time processing of the measured data with no additional delay on a standard desktop PC.

PMID:39771884 | DOI:10.3390/s24248139

Clin Imaging. 2025 Jan 2;119:110400. doi: 10.1016/j.clinimag.2024.110400. Online ahead of print.

ABSTRACT

PURPOSE: To develop an educational, interactive, ultra-high resolution, in vivo magnetic resonance (MR) neurography atlas for direct visualization of the brachial plexus and upper extremity.

METHODS: A total of 16 adult volunteers without known peripheral neuropathy underwent magnetic resonance (MR) neurography of the brachial plexus and upper extremity. To improve vascular suppression, subjects received an intravenous infusion of ferumoxytol. To improve image quality, MR neurography datasets were reconstructed using a deep learning algorithm. The atlas was then developed using a web-based user-interface software, which allowed for labeling of peripheral nerves and muscles, and mapping of muscles to their respective innervation. The user interface was optimized to maximize interactivity and user-friendliness.

RESULTS: Fifteen subjects completed at least one scan with no reported adverse reactions from the ferumoxytol infusions. Adequate vascular suppression was observed in all MR neurography datasets. The images of the brachial plexus and upper extremity included in this atlas allowed for identification and labeling of 177 unique anatomical structures from the neck to the wrist. The atlas was made freely accessible on the internet.

CONCLUSION: A detailed and interactive MR neurography atlas of the brachial plexus and upper extremity was successfully developed to depict small nerves and fascicular detail with unprecedented spatial and contrast resolution. This freely available online resource (https://www.hss.edu/MRNatlas) can be used as an educational tool and clinical reference. The techniques utilized in this project serve as a framework for continued work in expanding the atlas to cover other peripheral nerve territories.

PMID:39765207 | DOI:10.1016/j.clinimag.2024.110400

J Environ Manage. 2025 Jan 6;374:124011. doi: 10.1016/j.jenvman.2024.124011. Online ahead of print.

ABSTRACT

Accurate meteorological observation data is of great importance to human production activities. Meteorological observation systems have been advancing toward automation, intelligence, and informatization. Yet, instrumental malfunctions and unstable sensor node resources could cause significant deviations of data from the actual characteristics it should reflect. To achieve greater data accuracy, early detections of data anomalies, continuous collections and timely transmissions of data are essential. While obvious anomalies can be readily identified, the detection of systematic and gradually emerging anomalies requires further analyses. This study develops an interpretable deep learning method based on an autoencoder (AE), SHapley Additive exPlanations (SHAP) and Bayesian optimization (BO), in order to facilitate prompt and accurate anomaly detections of meteorological observational data. The proposed method can be unfolded into four parts. Firstly, the AE performs anomaly detections based on multidimensional meteorological datasets by marking the data that shows significant reconstruction errors. Secondly, the model evaluates the importance of each meteorological element of the flagged data via SHapley Additive exPlanation (SHAP). Thirdly, a K-sigma based threshold automatic delineation method is employed to obtain reasonable anomaly thresholds that are subject to the data characteristics of different observation sites. Finally, the BO algorithm is adopted to fine-tune difficult hyperparameters, enhancing the model's structure and thus the accuracy of anomaly detection. The practical implication of the proposed model is to inform agricultural production, climate observation, and disaster prevention.

PMID:39765064 | DOI:10.1016/j.jenvman.2024.124011

ACS Sens. 2025 Jan 7. doi: 10.1021/acssensors.4c02616. Online ahead of print.

ABSTRACT

Rapidly detecting hydrogen leaks is critical for the safe large-scale implementation of hydrogen technologies. However, to date, no technically viable sensor solution exists that meets the corresponding response time targets under technically relevant conditions. Here, we demonstrate how a tailored long short-term transformer ensemble model for accelerated sensing (LEMAS) speeds up the response of an optical plasmonic hydrogen sensor by up to a factor of 40 and eliminates its intrinsic pressure dependence in an environment emulating the inert gas encapsulation of large-scale hydrogen installations by accurately predicting its response value to a hydrogen concentration change before it is physically reached by the sensor hardware. Moreover, LEMAS provides a measure for the uncertainty of the predictions that are pivotal for safety-critical sensor applications. Our results advertise the use of deep learning for the acceleration of sensor response, also beyond the realm of plasmonic hydrogen detection.

PMID:39764741 | DOI:10.1021/acssensors.4c02616

Brief Bioinform. 2024 Nov 22;26(1):bbae685. doi: 10.1093/bib/bbae685.

ABSTRACT

Spatial transcriptomics (ST) technologies enable dissecting the tissue architecture in spatial context. To perceive the global contextual information of gene expression patterns in tissue, the spatial dependence of cells must be fully considered by integrating both local and non-local features by means of spatial-context-aware. However, the current ST integration algorithm ignores for ST dropouts, which impedes the spatial-aware of ST features, resulting in challenges in the accuracy and robustness of microenvironmental heterogeneity detecting, spatial domain clustering, and batch-effects correction. Here, we developed an STMGraph, a universal dual-view dynamic deep learning framework that combines dual-remask (MASK-REMASK) with dynamic graph attention model (DGAT) to exploit ST data outperforming pre-existing tools. The dual-remask mechanism masks the embeddings before encoding and decoding, establishing dual-decoding-view to share features mutually. DGAT leverages self-supervision to update graph linkage relationships from two distinct perspectives, thereby generating a comprehensive representation for each node. Systematic benchmarking against 10 state-of-the-art tools revealed that the STMGraph has the optimal performance with high accuracy and robustness on spatial domain clustering for the datasets of diverse ST platforms from multi- to sub-cellular resolutions. Furthermore, STMGraph aggregates ST information cross regions by dual-remask to realize the batch-effects correction implicitly, allowing for spatial domain clustering of ST multi-slices. STMGraph is platform independent and superior in spatial-context-aware to achieve microenvironmental heterogeneity detection, spatial domain clustering, batch-effects correction, and new biological discovery, and is therefore a desirable novel tool for diverse ST studies.

PMID:39764614 | DOI:10.1093/bib/bbae685

Int J Surg. 2025 Jan 7. doi: 10.1097/JS9.0000000000002220. Online ahead of print.

ABSTRACT

Detection of biomarkers of breast cancer incurs additional costs and tissue burden. We propose a deep learning-based algorithm (BBMIL) to predict classical biomarkers, immunotherapy-associated gene signatures, and prognosis-associated subtypes directly from hematoxylin and eosin stained histopathology images. BBMIL showed the best performance among comparative algorithms on the prediction of classical biomarkers, immunotherapy related gene signatures, and subtypes.

PMID:39764584 | DOI:10.1097/JS9.0000000000002220