Deep Learning Algorithms Research Articles

Evaluating Explainable Artificial Intelligence (XAI) techniques in chest radiology imaging through a human-centered Lens.

The field of radiology imaging has experienced a remarkable increase in using of deep learning (DL) algorithms to support diagnostic and treatment decisions. This rise has led to the development of Explainable AI (XAI) system to improve the transparency and trust of complex DL methods. However, XAI systems face challenges in gaining acceptance within the healthcare sector, mainly due to technical hurdles in utilizing these systems in practice and the lack of human-centered evaluation/validation. In this study, we focus on visual XAI systems applied to DL-enabled diagnostic system in chest radiography. In particular, we conduct a user study to evaluate two prominent visual XAI techniques from the human perspective. To this end, we created two clinical scenarios for diagnosing pneumonia and COVID-19 using DL techniques applied to chest X-ray and CT scans. The achieved accuracy rates were 90% for pneumonia and 98% for COVID-19. Subsequently, we employed two well-known XAI methods, Grad-CAM (Gradient-weighted Class Activation Mapping) and LIME (Local Interpretable Model-agnostic Explanations), to generate visual explanations elucidating the AI decision-making process. The visual explainability results were shared through a user study, undergoing evaluation by medical professionals in terms of clinical relevance, coherency, and user trust. In general, participants expressed a positive perception of the use of XAI systems in chest radiography. However, there was a noticeable lack of awareness regarding their value and practical aspects. Regarding preferences, Grad-CAM showed superior performance over LIME in terms of coherency and trust, although concerns were raised about its clinical usability. Our findings highlight key user-driven explainability requirements, emphasizing the importance of multi-modal explainability and the necessity to increase awareness of XAI systems among medical practitioners. Inclusive design was also identified as a crucial need to ensure better alignment of these systems with user needs.

Federated learning and deep learning framework for MRI image and speech signal-based multi-modal depression detection

Adolescence is a significant period for developing skills and knowledge and learning about managing relationships and emotions by gathering attributes for maturity. Recently, Depression arises as a common mental health issue in adolescents and this affects the daily life of the person. This leads to educational and social impairments and this acts as a major risk for suicide. As a result, the identification and treatment for this disorder are essential. By applying Deep learning (DL) algorithms to medical data, the mental condition of a person can be predicted. However, the traditional deep learning models face the challenge in processing the huge sized data. Hence, FL has emerged as an efficient solution for addressing the data size issue of DL. Here, Depression detection in adolescents is carried out by considering the FL framework, which comprises two modules, namely the local module and the Global module. The detection process is done in the local module using the proposed Exponential African Pelican Optimization based Deep Convolutional Neural Network (ExpAPO-DCNN), whereas the Global module produces the aggregated output of the local module. In this research, FL utilizes the DL model in producing the output, where the DL model considered two modalities of inputs, such as speech signal and Magnetic Resonance Imaging (MRI) image. The processing steps used for this research are pre-processing, feature extraction and detection. For MRI and speech signals, all the above processes are carried out individually. Finally, both the outputs are fused utilizing the overlap coefficient. The ExpAPO-DCNN obtained accuracy, Loss, Root mean Squared error (RMSE), Mean Squared error (MSE), True Negative rate (TNR), and True Positive rate (TPR) of 98.00 %, 0.023, 0.058, 0.240, 97.90 %, and 96.30 %, respectively.

A bidirectional gated recurrent unit based novel stacking ensemble regressor for foretelling the global horizontal irradiance

The rapid expansion of solar power generation has led to new challenges in solar intermittency, requiring precise forecasts of Global Horizontal Irradiance (GHI). Accurate GHI predictions are crucial for integrating sustainable energy sources into traditional electrical grid management. The article proposes an innovative solution, the novel Enhanced Stack Ensemble with a Bi-directional Gated Recurrent Unit (ESE-Bi-GRU), which uses machine learning (ML) boosting regressors such as Ada Boost, Cat Boost, Extreme Gradient Boost, and Gradient Boost, and Light Gradient Boost Machine acts as a base learner and the deep learning (DL) algorithms such as Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) for both directions are taken as a meta-learner. The predictive performance of the proposed ESE-Bi-GRU model is evaluated against individual models, showing significant reductions in mean absolute error (MAE) by 86.03 % and root mean squared error (RMSE) by 66.43 %. The model's ability to minimize prediction errors, such as MAE and RMSE holds promise for more effective planning and utilization of sporadic solar resources. By improving GHI forecast accuracy, the ESE-Bi-GRU model contributes to optimizing the integration of sustainable energy sources within the broader energy grid, fostering a more sustainable and environmentally conscious approach to energy management.

Analyzing the dynamics between crude oil spot prices and futures prices by maturity terms: Deep learning approaches to futures-based forecasting

Predicting crude oil prices is critical due to their significant impact on the global economy. While numerous studies have focused on time series forecasting and machine learning methods to predict crude oil spot prices, the use of futures prices segmented by the maturity terms as predictors is comparatively underexplored. This study explored a deep learning approach to investigate using futures prices across different maturities to predict spot prices. Specifically, this paper analyzes the predictive power of one-, two-, three-, six-, and 12-month crude oil futures contracts. This study employs multiple deep learning algorithms, including Multilayer Perceptron (MLP), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Convolutional Neural Network (CNN), and Temporal Convolutional Neural Network (TCN), to forecast crude oil spot prices. This research investigates the performance of these machine learning models when exploring the relationship between futures and spot prices of crude oil. In addition, this research incorporates extensive hyperparameter tuning to enhance the interpretability of the machine learning models when forecasting spot prices using futures prices, thereby contributing to the field of Explainable Artificial Intelligence (XAI) with an optimal set of hyperparameters. In summary, this research systematically shows results demonstrating predictive power in terms of XAI between spot prices and futures prices with different maturities, and machine learning algorithms.

Artificial Intelligence enabled self-powered sensing and wind energy harvesting system for bridges monitoring

With the advancement of society, the significance of Artificial Intelligence (AI) and big data in facilitating the intelligent development of bridges is increasing. Bridges function as crucial urban links, and ensuring their safe and stable operation is paramount. Therefore, this paper proposes an Artificial Intelligence enabled self-powered sensing wind energy harvesting system (WEHS) designed specifically for bridges. The system mainly consists of two electromagnetic generators (EMG1 and EMG2) and a rolling ball triboelectric nanogenerator (RB-TENG). WEHS achieves the recycling of ambient wind energy through the lower electromagnetic generator (EMG2). A single-channel triboelectric nanogenerator (RB-TENG) enables the sensing of wind speed within the system, while the top electromagnetic generator (EMG1) utilizes a single magnet with coils of different diameters to realize wind direction sensing. Reynolds wind tunnel that the WEHS has an average power output of 392 mW and a power density of 414.86 W/m3, which is adequate to power the sensor. The accuracy of wind speed and direction prediction based on the Artificial Intelligence GRU deep learning algorithm model is 96 % and 99.04 %, respectively. Furthermore, by implementing deep learning models for wind speed and direction, it becomes feasible to forecast the variations in wind direction and speed in the vicinity of bridges. The WEHS exhibits significant potential for bridges application and can contribute to developing AIoT-based smart bridges.

Telemedicine in Diabetic Retinal Screening: Pre- and Post-COVID-19 Challenges a New Perspective

Abstract Telemedicine in ophthalmology has been around for decades and has been successful with its use in diabetic retinal screening in countries like the UK (with the introduction of the UK National Diabetic Eye Screening Programme in 2003). However, most telemedicine, in the field of diabetic retinopathy, has largely been reliant on human graders for triage purposes. With the advent of COVID-19, patients with chronic conditions, such as diabetes, were disproportionately affected. The pandemic also caused significant rise in patients on waiting lists. Before the pandemic, there have been studies illustrating the use of artificial intelligence (AI) to analyse images obtained from patients screened for monitoring of their diabetic retinopathy. The image analysis by AI and deep-learning algorithms offers insight into the future of screening in diabetes. The transition, from the use of human graders in teleophthalmology to the use of AI-based image analysis has the potential to screen a wider cohort of patients, thereby tackling waiting lists awaiting screening which has lengthened since after COVID-19. It is therefore vital to understand the role of AI in screening diabetic retinopathy patients, from a patient-acceptability, cost-effectiveness and reliability perspective as, this offers potential answers to streamline the screening process further.

A Comparative Study of Metaheuristic Feature Selection Algorithms for Respiratory Disease Classification.

The correct diagnosis and early treatment of respiratory diseases can significantly improve the health status of patients, reduce healthcare expenses, and enhance quality of life. Therefore, there has been extensive interest in developing automatic respiratory disease detection systems. Most recent methods for detecting respiratory disease use machine and deep learning algorithms. The success of these machine learning methods depends heavily on the selection of proper features to be used in the classifier. Although metaheuristic-based feature selection methods have been successful in addressing difficulties presented by high-dimensional medical data in various biomedical classification tasks, there is not much research on the utilization of metaheuristic methods in respiratory disease classification. This paper aims to conduct a detailed and comparative analysis of six widely used metaheuristic optimization methods using eight different transfer functions in respiratory disease classification. For this purpose, two different classification cases were examined: binary and multi-class. The findings demonstrate that metaheuristic algorithms using correct transfer functions could effectively reduce data dimensionality while enhancing classification accuracy.

A TRANSFORMAÇÃO DA INDÚSTRIA FARMACÊUTICA PELA INTELIGÊNCIA ARTIFICIAL: IMPACTOS NO DESENVOLVIMENTO DE MEDICAMENTOS

The overall objective is to investigate how AI influences everything from the discovery of new compounds to the optimization of research and development processes. The methodology was based on a literature review using Google Scholar, prioritizing publications between 2020 and 2024. The main advances, such as neural networks and deep learning algorithms, which accelerate the identification of bioactive compounds and the conduct of clinical trials were explored. In development, AI has proven to be an essential tool for optimizing processes, reducing costs and increasing the efficiency of clinical trials, standing out for its ability to predict failures and personalize treatments. The conclusion highlights that, despite the advantages of AI, there are ethical and regulatory challenges that need to be addressed to ensure its safe and effective implementation. By overcoming these obstacles, AI will continue to profoundly transform the pharmaceutical industry, promoting more effective and accessible treatments.

Unpaired Image Translation to Mitigate Domain Shift in Liquid Argon Time Projection Chamber Detector Responses

Abstract Deep learning algorithms often are developed and trained on a training dataset and deployed on test datasets. Any systematic difference between the training and a test dataset may severely degrade the final algorithm performance on the test dataset -- what is known as the domain shift problem. This issue is prevalent in many scientific domains where algorithms are trained on simulated data but applied to real-world datasets. Typically, the domain shift problem is solved through various domain adaptation methods. However, these methods are often tailored to a specific downstream task and may not easily generalize to different tasks. This work explores the feasibility of using an alternative way to solve the domain shift problem that is not specific to any downstream algorithm. The proposed approach relies on modern Unpaired Image-to-Image translation techniques, designed to find translations between different image domains in a fully unsupervised fashion. In this study, the approach is applied to a domain shift problem commonly encountered in Liquid Argon Time Projection Chamber detector research when seeking a way to translate samples between two differently distributed LArTPC detector datasets deterministically. This translation allows for mapping real-world data into the simulated data domain where the downstream algorithms can be run with much less domain-shift-related performance degradation. Conversely, using the translation from the simulated data in a real-world domain can increase the realism of the simulated dataset and reduce the magnitude of any systematic uncertainties. We adapted several popular UI2I translation algorithms to work on scientific data and demonstrated the viability of these techniques for solving the domain shift problem with LArTPC detector data. To facilitate further development of domain adaptation techniques for scientific datasets, the "Simple Liquid-Argon Track Samples" (SLATS) dataset used in this study also is published.

Development and Multinational Validation of an Ensemble Deep Learning Algorithm for Detecting and Predicting Structural Heart Disease Using Noisy Single-lead Electrocardiograms.

AI-enhanced 12-lead ECG can detect a range of structural heart diseases (SHDs) but has a limited role in community-based screening. We developed and externally validated a noise-resilient single-lead AI-ECG algorithm that can detect SHD and predict the risk of their development using wearable/portable devices. Using 266,740 ECGs from 99,205 patients with paired echocardiographic data at Yale New Haven Hospital, we developed ADAPT-HEART, a noise-resilient, deep-learning algorithm, to detect SHD using lead I ECG. SHD was defined as a composite of LVEF<40%, moderate or severe left-sided valvular disease, and severe LVH. ADAPT-HEART was validated in four community hospitals in the US, and the population-based cohort of ELSA-Brasil. We assessed the model's performance as a predictive biomarker among those without baseline SHD across hospital-based sites and the UK Biobank. The development population had a median age of 66 [IQR, 54-77] years and included 49,947 (50.3%) women, with 18,896 (19.0%) having any SHD. ADAPT-HEART had an AUROC of 0.879 (95% CI, 0.870-0.888) with good calibration for detecting SHD in the test set, and consistent performance in hospital-based external sites (AUROC: 0.852-0.891) and ELSA-Brasil (AUROC: 0.859). Among those without baseline SHD, high vs. low ADAPT-HEART probability conferred a 2.8- to 5.7-fold increase in the risk of future SHD across data sources (all P<0.05). We propose a novel model that detects and predicts a range of SHDs from noisy single-lead ECGs obtainable on portable/wearable devices, providing a scalable strategy for community-based screening and risk stratification for SHD.

Real-time Deep Learning Based Mobile Application for Detecting Edible Fungi: Mushapp

Mushroom consumption and wild mushroom gathering are increasing in our country and in the world. Mushroom poisoning has an important place in food poisoning cases. Mushroom poisoning accounts for approximately 7% of poisoning cases in adults. Mushroom collection and consumption is common in many regions of our country. In this study, a deep learning based mobile application was developed to reduce the incidence of mushroom poisoning by taking a photo of a mushroom and determining the type and toxicity of the mushroom from the photo. This mobile application is called MushAPP. In the first phase of the study, 5150 mushroom images of 20 mushroom species were used to create the dataset. The dataset was then pre-processed and converted into a format that can be used by the deep learning algorithm. The mobile application side of the project was developed in Android Studio IDE environment. An artificial intelligence model was integrated into the designed mobile application. In the application, the type and toxicity status of the mushroom viewed from the mobile device camera are determined and presented to the user. The research findings were analyzed and it was determined that the accuracy rate of the application in detecting the mushroom species was 99.8%.

Optical Image Processing using Eulerian Amplification

Recent advancements in digital photography, particularly through mobile phone cameras, microscopes, and satellite imaging, have made high-resolution image capture increasingly accessible. However, these captured images often contain subtle changes in color resolution and density that are challenging to observe with the naked eye. This study explores the application of Eulerian amplification as a method to detect and enhance such small-amplitude variations, making them visible for improved analysis. Unlike traditional approaches, Eulerian methods are particularly effective for processing smooth structures and high-quality printed materials where minor amplifications in color can reveal significant grading details. Using a pixel-by-pixel Eulerian path analysis, this method allows for a detailed comparison of color intensity values on a fine 8-bit scale, which ranges from 0 to 255. Through this analysis, color distribution patterns are visualized across the spatial structure of the image, which is particularly beneficial for identifying specific variations in quality in printed materials, such as highly graded cards. To manage large image structures, spatial decomposition using a multi-level Gaussian pyramid technique is employed. By applying temporal filtering within select frequency bands (0.4–4 Hz) at a coarse pyramid level, the algorithm effectively pools spatial data, enabling the detection of color density shifts that may indicate image defects. The amplified image data is further processed using a classifier neural network, which provides high sensitivity and specificity in detecting edge defects, image centering issues, scuffs, and breakages. The system demonstrates promising results in identifying true positives while maintaining low false-negative rates, thus confirming the reliability of deep machine learning algorithms in high-sensitivity defect detection. This paper presents the significance of Eulerian amplification in optical image processing, offering a robust tool for precise, automated defect detection across a variety of high-resolution image types.

Local classification of crystalline structures in complex plasmas using a PointNet

Abstract In complex plasmas, microparticles can form ordered crystalline structures under specific conditions. Accurately identifying these structures, such as face-centered cubic, hexagonal close-packed, and body-centered cubic, is a common task in physics. Previous methods rely on detecting symmetries in the spatial arrangement of particles, often requiring extensive calculations. This study presents a novel approach by utilizing a PointNet-based deep learning algorithm, called WignerNet, to classify these structures directly from three-dimensional reconstructions of their Voronoi cells. The model was trained exclusively on artificial and labeled data, incorporating various noise levels, and subsequently tested on real experimental data. The results demonstrate that our method effectively classifies structures, reducing computational complexity and improving accuracy compared to conventional techniques. This advancement opens up new possibilities for real-time analysis of complex plasma systems in various research.

Inferring the dust emission at submillimeter and millimeter wavelengths using neural networks

The Planck mission provided all-sky dust emission maps in the submillimeter (submm) to millimeter (mm) range at an angular resolution of 5$^ prime $. In addition, some specific sources can be observed at long wavelengths and higher resolution using ground-based telescopes. These observations are limited to small scales and are sometimes not delivered to the community. These ground-based observations require extensive data processing before they become available for scientific analysis, and suffer from extended emission filtering. At present, we are still unable to fully understand the emissivity variations observed in different astrophysical environments at long (submm and mm) wavelengths. Several models have been developed to reproduce the diffuse Galactic medium, and each distinct environment requires an adjustment of the models. It is therefore challenging to estimate any dust emission in the submm-mm at a better resolution than the 5$^ prime $ from Planck . In this analysis, based on supervised deep learning algorithms, we produced dust emission predictions in the two Planck bands centered at 850 mic (353 GHz) and 1.38 mm (217 GHz) at the Herschel resolution (37$^ prime $). Prediction or forecasting is a frequently used term in machine learning or neural network research that refers to the output of an algorithm that has been trained on a given dataset and that is being used for modeling purposes. Herschel data of Galactic environments, ranging from 160 mic to 500 mic and smoothed to an angular resolution of 5$^ prime $, were used to train the neural network. This training aimed to provide the most accurate model for reproducing Planck maps of dust emission at 850 mic and 1.38 mm. Then, using Herschel data only, the model was applied to predict dust emission maps at 37$^ prime The neural network is capable of reproducing dust emission maps of various Galactic environments with a difference of only a few percent at the Planck resolution. Remarkably, it also performs well for nearby extragalactic environments. This could indicate that large dust grains, probed by submm or mm observations, have similar properties in both our Galaxy and nearby galaxies, or at least that their spectral behaviors are comparable in Galactic and extragalactic environments. For the first time, we provide to the community dust emission prediction maps at 850 mic and 1.38 mm at the 37$^ prime $ of several surveys: Hi-GAL, Gould Belt, Cold Cores, HERITAGE, Helga, HerM33es, KINGFISH, and Very Nearby Galaxies. The ratio of these two wavelength brightness bands reveals a derived emissivity spectral index statistically close to 1 for all the surveys, which favors the hypothesis of a flattened dust emission spectrum for wavelengths larger than 850 mic . Neural networks appear to be powerful algorithms that are highly efficient at learning from large datasets and achieving accurate reproductions with a deviation of only a few percent. However, to fully recover the input data during the training, it is essential to sample a sufficiently large range of datasets and physical conditions.

Live feature transformation scale extraction of face images based on deep learning algorithm

In view of the low extraction accuracy and efficiency of the current feature extraction methods of face images, a method for extracting the scale of live feature transformation of face images based on deep learning algorithm is proposed. First, the deep learning method is used to complete the denoising of face images; then the Gabor wave filter is used to decompose the face signal, and it is input into the deep subspace model to complete the extraction of feature transformation scale; finally, the live feature transformation scale extraction of face images is completed based on (particle PSO swarm optimization algorithm). Experimental results show that the proposed method for extracting face image features has higher accuracy, faster recognition speed and better overall application effect.

Deep learning and genetic algorithm driven accelerated design for frequency-multiplexed complex-amplitude coding meta-hologram

Frequency-multiplexed metasurfaces represent a significant innovation in breaking the functional limitations of traditional metasurfaces, showing immense potential in multi-channel communication. However, existing frequency-multiplexed metasurfaces primarily focus on pure phase and linear polarization modulation, neglecting the modulation for complex amplitude and circularly polarized waves. Additionally, crosstalk suppression between dual-frequency channels often requires meticulous tuning of the meta-atom structure. Therefore, manually designing a set of meta-atoms that satisfies both complex amplitude modulation and low crosstalk at dual frequencies is extremely challenging and time-consuming. Here, we utilize the method of deep learning and genetic algorithm to design a kind of meta-atom capable of bi-spectral 2-bit amplitude and arbitrary phase modulation, which greatly reduces the design difficulty and achieves excellent low-crosstalk performance. This method can be easily generalized to the design of other complex meta-atoms to improve the design efficiency. Furthermore, we propose a frequency-multiplexed complex-amplitude coding meta-hologram for modulating left-handed circularly polarized (LCP) waves. When illuminated with LCP light, it can reconstruct two distinct holographic images at two different frequencies in the near field with high quality. The independent modulation capability of the metasurface for multiple degrees of freedom of frequency, amplitude and phase gives it broad application prospects in multi-channel communication, data storage and perfect holography.

An integrated method for detecting lung cancer via CT scanning via optimization, deep learning, and IoT data transmission.

With its increasing global prevalence, lung cancer remains a critical health concern. Despite the advancement of screening programs, patient selection and risk stratification pose significant challenges. This study addresses the pressing need for early detection through a novel diagnostic approach that leverages innovative image processing techniques. The urgency of early lung cancer detection is emphasized by its alarming growth worldwide. While computed tomography (CT) surpasses traditional X-ray methods, a comprehensive diagnosis requires a combination of imaging techniques. This research introduces an advanced diagnostic tool implemented through image processing methodologies. The methodology commences with histogram equalization, a crucial step in artifact removal from CT images sourced from a medical database. Accurate lung CT image segmentation, which is vital for cancer diagnosis, follows. The Otsu thresholding method and optimization, employing Colliding Bodies Optimization (CBO), enhance the precision of the segmentation process. A local binary pattern (LBP) is deployed for feature extraction, enabling the identification of nodule sizes and precise locations. The resulting image underwent classification using the densely connected CNN (DenseNet) deep learning algorithm, which effectively distinguished between benign and malignant tumors. The proposed CBO+DenseNet CNN exhibits remarkable performance improvements over traditional methods. Notable enhancements in accuracy (98.17%), specificity (97.32%), precision (97.46%), and recall (97.89%) are observed, as evidenced by the results from the fractional randomized voting model (FRVM). These findings highlight the potential of the proposed model as an advanced diagnostic tool. Its improved metrics promise heightened accuracy in tumor classification and localization. The proposed model uniquely combines Colliding Bodies Optimization (CBO) with DenseNet CNN, enhancing segmentation and classification accuracy for lung cancer detection, setting it apart from traditional methods with superior performance metrics.

Harnessing AI for Next-Generation Financial Fraud Detection: A Data-Driven Revolution

Artificial Intelligence in Financial Fraud Detection: A Comprehensive Approach to Enhancing Financial Security The rise of artificial intelligence (AI) offers an opportunity to significantly strengthen financial security by combating financial fraud, which has become increasingly complex and widespread. Traditional detection methods are often insufficient in identifying and preventing fraudulent activities, prompting a shift towards AI-based solutions. This study explores the application of AI, particularly machine learning algorithms, in improving the accuracy and efficiency of fraud detection. By analyzing large financial datasets, AI can detect anomalies that may indicate fraudulent behavior more effectively than traditional approaches. This research adopts a two-phase methodology. The first phase involves a thorough review of existing financial fraud detection methods, comparing traditional techniques with AI-based models to identify gaps. Various machine learning approaches, including supervised, unsupervised, and deep learning algorithms, are reviewed for their effectiveness in detecting fraud. The second phase involves developing and testing an AI model to identify fraudulent patterns within transactional data. The model uses machine learning algorithms to process vast datasets and detect deviations from typical financial behaviors, flagging potentially fraudulent activities. The expected results indicate that AI systems can outperform traditional fraud detection methods by significantly reducing false positives and improving the detection rate of genuine fraud. This reduction in false positives is vital for financial institutions, as it reduces unnecessary investigations and saves valuable resources. Additionally, enhanced fraud detection protects both institutions and consumers from financial losses.The findings of this study aim to provide financial institutions with practical insights into the implementation of AI-driven fraud detection systems. Furthermore, the research highlights the need for continuous refinement of AI models to adapt to the evolving nature of financial fraud. By leveraging AI technologies, financial institutions can revolutionize their approach to fraud detection, making financial systems more secure and responsive to emerging threats.

Deep learning models for the prediction of acute postoperative pain in PACU for video‐assisted thoracoscopic surgery

BackgroundPostoperative pain is a prevalent symptom experienced by patients undergoing surgical procedures. This study aims to develop deep learning algorithms for predicting acute postoperative pain using both essential patient details and real-time vital sign data during surgery.MethodsThrough a retrospective observational approach, we utilized Graph Attention Networks (GAT) and graph Transformer Networks (GTN) deep learning algorithms to construct the DoseFormer model while incorporating an attention mechanism. This model employed patient information and intraoperative vital signs obtained during Video-assisted thoracoscopic surgery (VATS) surgery to anticipate postoperative pain. By categorizing the static and dynamic data, the DoseFormer model performed binary classification to predict the likelihood of postoperative acute pain.ResultsA total of 1758 patients were initially included, with 1552 patients after data cleaning. These patients were then divided into training set (n = 931) and testing set (n = 621). In the testing set, the DoseFormer model exhibited significantly higher AUROC (0.98) compared to classical machine learning algorithms. Furthermore, the DoseFormer model displayed a significantly higher F1 value (0.85) in comparison to other classical machine learning algorithms. Notably, the attending anesthesiologists' F1 values (attending: 0.49, fellow: 0.43, Resident: 0.16) were significantly lower than those of the DoseFormer model in predicting acute postoperative pain.ConclusionsDeep learning model can predict postoperative acute pain events based on patients' basic information and intraoperative vital signs.

Equipping computational pathology systems with artifact processing pipelines: a showcase for computation and performance trade-offs

BackgroundHistopathology is a gold standard for cancer diagnosis. It involves extracting tissue specimens from suspicious areas to prepare a glass slide for a microscopic examination. However, histological tissue processing procedures result in the introduction of artifacts, which are ultimately transferred to the digitized version of glass slides, known as whole slide images (WSIs). Artifacts are diagnostically irrelevant areas and may result in wrong predictions from deep learning (DL) algorithms. Therefore, detecting and excluding artifacts in the computational pathology (CPATH) system is essential for reliable automated diagnosis.MethodsIn this paper, we propose a mixture of experts (MoE) scheme for detecting five notable artifacts, including damaged tissue, blur, folded tissue, air bubbles, and histologically irrelevant blood from WSIs. First, we train independent binary DL models as experts to capture particular artifact morphology. Then, we ensemble their predictions using a fusion mechanism. We apply probabilistic thresholding over the final probability distribution to improve the sensitivity of the MoE. We developed four DL pipelines to evaluate computational and performance trade-offs. These include two MoEs and two multiclass models of state-of-the-art deep convolutional neural networks (DCNNs) and vision transformers (ViTs). These DL pipelines are quantitatively and qualitatively evaluated on external and out-of-distribution (OoD) data to assess generalizability and robustness for artifact detection application.ResultsWe extensively evaluated the proposed MoE and multiclass models. DCNNs-based MoE and ViTs-based MoE schemes outperformed simpler multiclass models and were tested on datasets from different hospitals and cancer types, where MoE using (MobileNet) DCNNs yielded the best results. The proposed MoE yields 86.15 % F1 and 97.93% sensitivity scores on unseen data, retaining less computational cost for inference than MoE using ViTs. This best performance of MoEs comes with relatively higher computational trade-offs than multiclass models. Furthermore, we apply post-processing to create an artifact segmentation mask, a potential artifact-free RoI map, a quality report, and an artifact-refined WSI for further computational analysis. During the qualitative evaluation, field experts assessed the predictive performance of MoEs over OoD WSIs. They rated artifact detection and artifact-free area preservation, where the highest agreement translated to a Cohen Kappa of 0.82, indicating substantial agreement for the overall diagnostic usability of the DCNN-based MoE scheme.ConclusionsThe proposed artifact detection pipeline will not only ensure reliable CPATH predictions but may also provide quality control. In this work, the best-performing pipeline for artifact detection is MoE with DCNNs. Our detailed experiments show that there is always a trade-off between performance and computational complexity, and no straightforward DL solution equally suits all types of data and applications. The code and HistoArtifacts dataset can be found online at Github and Zenodo, respectively.