Electrocardiography analysis is widely used in various clinical applications and Deep Learning models for classification tasks are currently in the focus of research. Due to their data-driven character, they bear the potential to handle signal noise efficiently, but its influence on the accuracy of these methods is still unclear. Therefore, we benchmark the influence of four types of noise on the accuracy of a Deep Learning-based method for atrial fibrillation detection in 12-lead electrocardiograms. We use a subset of a publicly available dataset (PTBXL) and use the metadata provided by human experts regarding noise for assigning a signal quality to each electrocardiogram. Furthermore, we compute a quantitative signal-to-noise ratio for each electrocardiogram. We analyze the accuracy of the Deep Learning model with respect to both metrics and observe that the method can robustly identify atrial fibrillation, even in cases signals are labelled by human experts as being noisy on multiple leads. False positive and false negative rates are slightly worse for data being labelled as noisy. Interestingly, data annotated as showing baseline drift noise results in an accuracy very similar to data without. We conclude that the issue of processing noisy electrocardiography data can be addressed successfully by Deep Learning methods that might not need preprocessing as many conventional methods do.
All fields of science depend on mathematical models. Occam's razor refers to the principle that good models should exclude parameters beyond those minimally required to describe the systems they represent. This is because redundancy can lead to incorrect estimates of model parameters from data, and thus inaccurate or ambiguous conclusions. Here, we show how deep learning can be powerfully leveraged to address Occam's razor. FixFit, our new method, uses a feedforward deep neural network with a bottleneck layer to characterize and predict the behavior of a given model from its input parameters. FixFit has three major benefits. First, it provides a metric to quantify the original model's degree of complexity. Second, it allows for the unique fitting of data. Third, it provides an unbiased way to discriminate between experimental hypotheses that add value versus those that do not. In two use cases, we demonstrate the broad applicability of this method across scientific domains. To validate the method using a known system, we apply FixFit to recover known composite parameters for the Kepler orbit model. To illustrate how the method can be applied to less well-established fields, we use it to identify parameters for a multi-scale brain model and reduce the search space for viable candidate mechanisms.
Confounding information in the form of text or markings embedded in medical images can severely affect the training of diagnostic deep learning algorithms. However, data collected for clinical purposes often have such markings embedded in them. In dermatology, known examples include drawings or rulers that are overrepresented in images of malignant lesions. In this paper, we encounter text and calipers placed on the images found in national databases containing fetal screening ultrasound scans, which correlate with standard planes to be predicted. In order to utilize the vast amounts of data available in these databases, we develop and validate a series of methods for minimizing the confounding effects of embedded text and calipers on deep learning algorithms designed for ultrasound, using standard plane classification as a test case.
In this paper, we present a review of the recent work in deep learning methods for user interface design. The survey encompasses well known deep learning techniques (deep neural networks, convolutional neural networks, recurrent neural networks, autoencoders, and generative adversarial networks) and datasets widely used to design user interface applications. We highlight important problems and emerging research frontiers in this field. We believe that the use of deep learning for user interface design automation tasks could be one of the high potential fields for the advancement of the software development industry.
Knee Osteoarthritis (KOA) is a prevalent musculoskeletal disorder that severely impacts mobility and quality of life, particularly among older adults. Its diagnosis often relies on subjective assessments using the Kellgren-Lawrence (KL) grading system, leading to variability in clinical evaluations. To address these challenges, we propose a confidence-driven deep learning framework for early KOA detection, focusing on distinguishing KL-0 and KL-2 stages. The Siamese-based framework integrates a novel multi-level feature extraction architecture with a hybrid loss strategy. Specifically, multi-level Global Average Pooling (GAP) layers are employed to extract features from varying network depths, ensuring comprehensive feature representation, while the hybrid loss strategy partitions training samples into high-, medium-, and low-confidence subsets. Tailored loss functions are applied to improve model robustness and effectively handle uncertainty in annotations. Experimental results on the Osteoarthritis Initiative (OAI) dataset demonstrate that the proposed framework achieves competitive accuracy, sensitivity, and specificity, comparable to those of expert radiologists. Cohen's kappa values (k > 0.85)) confirm substantial agreement, while McNemar's test (p > 0.05) indicates no statistically significant differences between the model and radiologists. Additionally, Confidence distribution analysis reveals that the model emulates radiologists' decision-making patterns. These findings highlight the potential of the proposed approach to serve as an auxiliary diagnostic tool, enhancing early KOA detection and reducing clinical workload.
When capturing and storing images, devices inevitably introduce noise. Reducing this noise is a critical task called image denoising. Deep learning has become the de facto method for image denoising, especially with the emergence of Transformer-based models that have achieved notable state-of-the-art results on various image tasks. However, deep learning-based methods often suffer from a lack of generalization ability. For example, deep models trained on Gaussian noise may perform poorly when tested on other noise distributions. To address this issue, we present a novel approach to enhance the generalization performance of denoising networks, known as masked training. Our method involves masking random pixels of the input image and reconstructing the missing information during training. We also mask out the features in the self-attention layers to avoid the impact of training-testing inconsistency. Our approach exhibits better generalization ability than other deep learning models and is directly applicable to real-world scenarios. Additionally, our interpretability analysis demonstrates the superiority of our method.
Stereo vision is essential for many applications. Currently, the synchronization of the streams coming from two cameras is done using mostly hardware. A software-based synchronization method would reduce the cost, weight and size of the entire system and allow for more flexibility when building such systems. With this goal in mind, we present here a comparison of different deep learning-based systems and prove that some are efficient and generalizable enough for such a task. This study paves the way to a production ready software-based video synchronization system.
Deep Learning (DL) developers come from different backgrounds, e.g., medicine, genomics, finance, and computer science. To create a DL model, they must learn and use high-level programming languages (e.g., Python), thus needing to handle related setups and solve programming errors. This paper presents DeepBlocks, a visual programming tool that allows DL developers to design, train, and evaluate models without relying on specific programming languages. DeepBlocks works by building on the typical model structure: a sequence of learnable functions whose arrangement defines the specific characteristics of the model. We derived DeepBlocks' design goals from a 5-participants formative interview, and we validated the first implementation of the tool through a typical use case. Results are promising and show that developers could visually design complex DL architectures.
Autism Spectrum Disorder (ASD) is one neuro developmental disorder that is now widespread in the world. ASD persists throughout the life of an individual, impacting the way they behave and communicate, resulting to notable deficits consisting of social life retardation, repeated behavioural traits and a restriction in their interests. Early detection of the disorder helps in the onset treatment and helps one to lead a normal life. There are clinical approaches used in detection of autism, relying on behavioural data and in worst cases, neuroimaging. Quantitative methods involving machine learning have been studied and developed to overcome issues with clinical approaches. These quantitative methods rely on machine learning, with some complex methods based on deep learning developed to accelerate detection and diagnosis of ASD. These literature is aimed at exploring most state-of-the-art probabilistic methods in use today, characterizing them with the type of dataset they're most applied on, their accuracy according to their novel research and how well they are suited in ASD classification. The findings will purposely serve as a benchmark in selection of the model to use when performing ASD detection.
Beamforming with large-scale antenna arrays has been widely used in recent years, which is acknowledged as an important part in 5G and incoming 6G. Thus, various techniques are leveraged to improve its performance, e.g., deep learning, advanced optimization algorithms, etc. Although its performance in many previous research scenarios with deep learning is quite attractive, usually it drops rapidly when the environment or dataset is changed. Therefore, designing effective beamforming network with strong robustness is an open issue for the intelligent wireless communications. In this paper, we propose a robust beamforming self-supervised network, and verify it in two kinds of different datasets with various scenarios. Simulation results show that the proposed self-supervised network with hybrid learning performs well in both classic DeepMIMO and new WAIR-D dataset with the strong robustness under the various environments. Also, we present the principle to explain the rationality of this kind of hybrid learning, which is instructive to apply with more kinds of datasets.
Imaging through scattering is a pervasive and difficult problem in many biological applications. The high background and the exponentially attenuated target signals due to scattering fundamentally limits the imaging depth of fluorescence microscopy. Light-field systems are favorable for high-speed volumetric imaging, but the 2D-to-3D reconstruction is fundamentally ill-posed, and scattering exacerbates the condition of the inverse problem. Here, we develop a scattering simulator that models low-contrast target signals buried in heterogeneous strong background. We then train a deep neural network solely on synthetic data to descatter and reconstruct a 3D volume from a single-shot light-field measurement with low signal-to-background ratio (SBR). We apply this network to our previously developed Computational Miniature Mesoscope and demonstrate the robustness of our deep learning algorithm on scattering phantoms with different scattering conditions. The network can robustly reconstruct emitters in 3D with a 2D measurement of SBR as low as 1.05 and as deep as a scattering length. We analyze fundamental tradeoffs based on network design factors and out-of-distribution data that affect the deep learning model's generalizability to real experimental data. Broadly, we believe that our simulator-based deep learning approach can be applied to a wide range of imaging through scattering techniques where experimental paired training data is lacking.
Next-generation integrated nanophotonic device designs leverage advanced optimization techniques such as inverse design and topology optimization which achieve high performance and extreme miniaturization by optimizing a massively complex design space enabled by small feature sizes. However, unless the optimization is heavily constrained, the generated small features are not reliably fabricated, leading to optical performance degradation. Even for simpler, conventional designs, fabrication-induced performance degradation still occurs. The degree of deviation from the original design not only depends on the size and shape of its features, but also on the distribution of features and the surrounding environment, presenting complex, proximity-dependent behavior. Without proprietary fabrication process specifications, design corrections can only be made after calibrating fabrication runs take place. In this work, we introduce a general deep machine learning model that automatically corrects photonic device design layouts prior to first fabrication. Only a small set of scanning electron microscopy images of engineered training features are required to create the deep learning model. With correction, the outcome of the fabricated layout is closer to what is intended, and thus so too is the performance of the design. Without modifying the nanofabrication process, adding significant computation in design, or requiring proprietary process specifications, we believe our model opens the door to new levels of reliability and performance in next-generation photonic circuits.
With the development of deep learning processors and accelerators, deep learning models have been widely deployed on edge devices as part of the Internet of Things. Edge device models are generally considered as valuable intellectual properties that are worth for careful protection. Unfortunately, these models have a great risk of being stolen or illegally copied. The existing model protections using encryption algorithms are suffered from high computation overhead which is not practical due to the limited computing capacity on edge devices. In this work, we propose a light-weight, practical, and general Edge device model Pro tection method at neuron level, denoted as EdgePro. Specifically, we select several neurons as authorization neurons and set their activation values to locking values and scale the neuron outputs as the "asswords" during training. EdgePro protects the model by ensuring it can only work correctly when the "passwords" are met, at the cost of encrypting and storing the information of the "passwords" instead of the whole model. Extensive experimental results indicate that EdgePro can work well on the task of protecting on datasets with different modes. The inference time increase of EdgePro is only 60% of state-of-the-art methods, and the accuracy loss is less than 1%. Additionally, EdgePro is robust against adaptive attacks including fine-tuning and pruning, which makes it more practical in real-world applications. EdgePro is also open sourced to facilitate future research: https://github.com/Leon022/Edg
Most of the work on chatter detection is based on laboratory machining tests, thus without the constraints of noise, the variety of situations to be managed in the industry, and the uncertainties on the parameters (sensor position, tool engagement, and sometimes even spindle rotation frequency). This work presents an approach first based on mechanical skills first to identify the optimal signal processing, then based on deep learning to automatically detect the phenomenon of chatter in machining. To do so, we use input images from the Fast Fourier Transform analysis of a vibration signal from the machining of an Alstom industrial railway system. The FFT images are first cropped in height to perform a renormalization of the signal scale. This renormalization step eliminates a bias regularly observed in the literature, which did not eliminate a simple correlation with signal amplitude because chatter is often, but not always, associated with stronger vibrations. Deep learning extracted features from the image data by exploiting pre-trained deep neural networks such as VGG16 and ResNet50. The data is derived from an industrial vibration signal (with noise, a not ideally placed sensors, and uncertainties on process parameters), which was very poorly represented in the literature. It is shown that by depleting the signal (renormalized amplitude and absence of information on the rotation speed), the model obtains a good result (an accuracy of 99% in the training and validation phases), which had never been done in the literature. In addition, since a part of the test dataset is similar to the training dataset, the proposed model obtains an accuracy of 98.82%. The robustness is proved by adding another test dataset containing noisy and more or less ambiguous images to be classified. The model still manages to detect these cases from its training of unambiguous cases, showing that the obtained generalization is efficient with an accuracy of 73.71%.
In this work we present a deep learning approach to conduct hypothesis-free, transcriptomics-based matching of drugs for diseases. Our proposed neural network architecture is trained on approved drug-disease indications, taking as input the relevant disease and drug differential gene expression profiles, and learns to identify novel indications. We assemble an evaluation dataset of disease-drug indications spanning 68 diseases and evaluate in silico our approach against the most widely used transcriptomics-based matching baselines, CMap and the Characteristic Direction. Our results show a more than 200% improvement over both baselines in terms of standard retrieval metrics. We further showcase our model's ability to capture different genes' expressions interactions among drugs and diseases. We provide our trained models, data and code to predict with them at https://github.com/healx/dgem-nn-public.
The discovery of novel drug target (DT) interactions is an important step in the drug development process. The majority of computer techniques for predicting DT interactions have focused on binary classification, with the goal of determining whether or not a DT pair interacts. Protein ligand interactions, on the other hand, assume a continuous range of binding strength values, also known as binding affinity, and forecasting this value remains a difficulty. As the amount of affinity data in DT knowledge-bases grows, advanced learning techniques such as deep learning architectures can be used to predict binding affinities. In this paper, we present a deep-learning-based methodology for predicting DT binding affinities using just sequencing information from both targets and drugs. The results show that the proposed deep learning-based model that uses the 1D representations of targets and drugs is an effective approach for drug target binding affinity prediction and it does not require additional chemical domain knowledge to work with. The model in which high-level representations of a drug and a target are constructed via CNNs that uses residual skip connections and also with an additional stream to create a high-level combined representation of the drug-target pair achieved the best Concordance Index (CI) performance in one of the largest benchmark datasets, outperforming the recent state-of-the-art method AttentionDTA and many other machine-learning and deep-learning based baseline methods for DT binding affinity prediction that uses the 1D representations of targets and drugs.
Video-based person re-identification (video re-ID) has lately fascinated growing attention due to its broad practical applications in various areas, such as surveillance, smart city, and public safety. Nevertheless, video re-ID is quite difficult and is an ongoing stage due to numerous uncertain challenges such as viewpoint, occlusion, pose variation, and uncertain video sequence, etc. In the last couple of years, deep learning on video re-ID has continuously achieved surprising results on public datasets, with various approaches being developed to handle diverse problems in video re-ID. Compared to image-based re-ID, video re-ID is much more challenging and complex. To encourage future research and challenges, this first comprehensive paper introduces a review of up-to-date advancements in deep learning approaches for video re-ID. It broadly covers three important aspects, including brief video re-ID methods with their limitations, major milestones with technical challenges, and architectural design. It offers comparative performance analysis on various available datasets, guidance to improve video re-ID with valuable thoughts, and exciting research directions.
Surface meshes are a favoured domain for representing structural and functional information on the human cortex, but their complex topology and geometry pose significant challenges for deep learning analysis. While Transformers have excelled as domain-agnostic architectures for sequence-to-sequence learning, the quadratic cost of the self-attention operation remains an obstacle for many dense prediction tasks. Inspired by some of the latest advances in hierarchical modelling with vision transformers, we introduce the Multiscale Surface Vision Transformer (MS-SiT) as a backbone architecture for surface deep learning. The self-attention mechanism is applied within local-mesh-windows to allow for high-resolution sampling of the underlying data, while a shifted-window strategy improves the sharing of information between windows. Neighbouring patches are successively merged, allowing the MS-SiT to learn hierarchical representations suitable for any prediction task. Results demonstrate that the MS-SiT outperforms existing surface deep learning methods for neonatal phenotyping prediction tasks using the Developing Human Connectome Project (dHCP) dataset. Furthermore, building the MS-SiT backbone into a U-shaped architecture for surface segmentation demonstrates competitive results on cortical parcellation using the UK Biobank (UKB) and manually-annotated MindBoggle datasets. Code and trained models are publicly available at https://github.com/metrics-lab/surface-vision-transformers.
Malaria is usually diagnosed by a microbiologist by examining a small sample of blood smear. Reducing mortality from malaria infection is possible if it is diagnosed early and followed with appropriate treatment. While the WHO has set audacious goals of reducing malaria incidence and mortality rates by 90% in 2030 and eliminating malaria in 35 countries by that time, it still remains a difficult challenge. Computer-assisted diagnostics are on the rise these days as they can be used effectively as a primary test in the absence of or providing assistance to a physician or pathologist. The purpose of this paper is to describe an approach to detecting, localizing and counting parasitic cells in blood sample images towards easing the burden on healthcare workers.
Purpose: To develop and evaluate an end-to-end deep learning pipeline for segmentation and analysis of cardiac magnetic resonance images to provide core-lab processing for a multi-centre registry of Fontan patients. Materials and Methods: This retrospective study used training (n = 175), validation (n = 25) and testing (n = 50) cardiac magnetic resonance image exams collected from 13 institutions in the UK, US and Canada. The data was used to train and evaluate a pipeline containing three deep-learning models. The pipeline's performance was assessed on the Dice and IoU score between the automated and reference standard manual segmentation. Cardiac function values were calculated from both the automated and manual segmentation and evaluated using Bland-Altman analysis and paired t-tests. The overall pipeline was further evaluated qualitatively on 475 unseen patient exams. Results: For the 50 testing dataset, the pipeline achieved a median Dice score of 0.91 (0.89-0.94) for end-diastolic volume, 0.86 (0.82-0.89) for end-systolic volume, and 0.74 (0.70-0.77) for myocardial mass. The deep learning-derived end-diastolic volume, end-systolic volume, myocardial mass, stroke volume and ejection fraction had no statistical difference compared to the same values derived from manual segmentation with p values all greater than 0.05. For the 475 unseen patient exams, the pipeline achieved 68% adequate segmentation in both systole and diastole, 26% needed minor adjustments in either systole or diastole, 5% needed major adjustments, and the cropping model only failed in 0.4%. Conclusion: Deep learning pipeline can provide standardised 'core-lab' segmentation for Fontan patients. This pipeline can now be applied to the >4500 cardiac magnetic resonance exams currently in the FORCE registry as well as any new patients that are recruited.