Training Dataset Research Articles

Ischemic Stroke Lesion Segmentation on Multiparametric CT Perfusion Maps Using Deep Neural Network

Background: Accurate delineation of lesions in acute ischemic stroke is important for determining the extent of tissue damage and the identification of potentially salvageable brain tissues. Automatic segmentation on CT images is challenging due to the poor contrast-to-noise ratio. Quantitative CT perfusion images improve the estimation of the perfusion deficit regions; however, they are limited by a poor signal-to-noise ratio. The study aims to investigate the potential of deep learning (DL) algorithms for the improved segmentation of ischemic lesions. Methods: This study proposes a novel DL architecture, DenseResU-NetCTPSS, for stroke segmentation using multiparametric CT perfusion images. The proposed network is benchmarked against state-of-the-art DL models. Its performance is assessed using the ISLES-2018 challenge dataset, a widely recognized dataset for stroke segmentation in CT images. The proposed network was evaluated on both training and test datasets. Results: The final optimized network takes three image sequences, namely CT, cerebral blood volume (CBV), and time to max (Tmax), as input to perform segmentation. The network achieved a dice score of 0.65 ± 0.19 and 0.45 ± 0.32 on the training and testing datasets. The model demonstrated a notable improvement over existing state-of-the-art DL models. Conclusions: The optimized model combines CT, CBV, and Tmax images, enabling automatic lesion identification with reasonable accuracy and aiding radiologists in faster, more objective assessments.

Real-Time Auto Controlling of Viable Cell Density in Perfusion Cultivation Aided by In-Line Dielectric Spectroscopy With Segmented Adaptive PLS Model.

Serving as a dedicated process analytical technology (PAT) tool for biomass monitoring and control, the capacitance probe, or dielectric spectroscopy, is showing great potential in robust pharmaceutical manufacturing, especially with the growing interest in integrated continuous bioprocessing. Despite its potential, challenges still exist in terms of its accuracy and applicability, particularly when it is used to monitor cells during stationary and decline phases. In this study, data pre-processing methods were first evaluated through cross-validation, where the first-order derivative emerged as the most effective method to diminish variability in prediction accuracy across different training datasets. Subsequently, a segmented adaptive partial least squares (SA-PLS) model was developed, and its accuracy and universality were demonstrated through several validation studies using different clones and culture processes. Furthermore, a real-time viable cell density (VCD) auto-control system in perfusion culture was established, where the VCD was maintained around the target with notable precision and robustness. This model enhanced the monitoring capabilities of capacitance-based PAT tools throughout the cultivation, expanded their application in cell-specific automatic control strategies, and contributed vitally to the advancement of continuous manufacturing paradigms.

AI-Driven Signal Processing for SF6 Circuit Breaker Performance Optimization

This work presents an approach based on signal processing and artificial intelligence (AI) to identify the pre-insertion resistor (PIR) and main contact instants during the operation of high-voltage SF6 circuit breakers to help improve the settings of controlled switching and attenuate transients. For this, the current and voltage signals of a real Brazilian substation are used as AI inputs, considering the noise and interferences common in this type of environment. Thus, the proposed modeling considers the signal preprocessing steps for feature extraction, the generation of the dataset for model training, the use of different machine learning techniques to automatically find the desired points, and, finally, the identification of the best moments for controlled switching of the circuit breakers. As a result, the models evaluated obtained good performance in the identification of operation points above 93%, considering precision and accuracy. In addition, valuable statistical notes related to the controlled switching condition are obtained from the circuit breakers evaluated in this research.

Exploration of linear and interpretable models for quantification of cell parameters via contactless short-wave infrared hyperspectral sensing

The development of optical sensors for label-free quantification of cell parameters has numerous uses in the biomedical arena. However, using current optical probes requires the laborious collection of sufficiently large datasets that can be used to calibrate optical probe signals to true metabolite concentrations. Further, most practitioners find it difficult to confidently adapt black box chemometric models that are difficult to troubleshoot in high-stakes applications such as biopharmaceutical manufacturing. Replacing optical probes with contactless short-wave infrared (SWIR) hyperspectral cameras allows efficient collection of thousands of absorption signals in a handful of images. This high repetition allows for effective denoising of each spectrum, so interpretable linear models can quantify metabolites. To illustrate, an interpretable linear model called L-SLR is trained using small datasets obtained with a SWIR HSI camera to quantify fructose, viable cell density (VCD), glucose, and lactate. The performance of this model is also compared to other existing linear models, namely Partial Least Squares (PLS) and Non-negative Matrix Factorization (NMF). Using only 50% of the dataset for training, reasonable test performance of mean absolute error (MAE) and correlations (r2) are achieved for glucose (r2 = 0.88, MAE = 37 mg/dL), lactate (r2 = 0.93, MAE = 15.08 mg/dL), and VCD (r2 = 0.81, MAE = 8.6 × 105 cells/mL). Further, these models are also able to handle quantification of a metabolite like fructose in the presence of high background concentration of similar metabolite with almost identical chemical interactions in water like glucose. The model achieves reasonable quantification performance for large fructose level (100–1000 mg/dL) quantification (r2 = 0.92, MAE = 25.1 mg/dL) and small fructose level (< 60 mg/dL) concentrations (r2 = 0.85, MAE = 4.97 mg/dL) in complex media like Fetal Bovine Serum (FBS). Finally, the model provides sparse interpretable weight matrices that hint at the underlying solution changes that correlate to each cell parameter prediction.

Predicting rifampicin resistance inM. tuberculosisusing machine learning informed by protein structural and chemical features

BackgroundRifampicin remains a key antibiotic in the treatment of tuberculosis. Despite advances in cataloguing resistance-associated variants (RAVs), novel and rare mutations in the relevent gene,rpoB, will be encountered in clinical samples, complicating the task of using genetics to predict whether a sample is resistant or not to rifampicin. We have trained a series of machine learning models with the aim of complementing genetics-based drug susceptibility testing.MethodsWe built a Test+Train dataset comprising 219 susceptible mutations and 46 RAVs. Features derived from the structure of the RNA polymerase or the change in chemistry introduced by the mutation were considered, however, only a few, notably the distance from the rifampicin binding site, were found to be predictive on their own. Due to the paucity of RAVs we used Monte Carlo cross-validation with 50 repeats to train four different machine learning models.ResultsAll four models behaved similarly with sensitivities and specificities in the range 0.84–0.88 and 0.94–0.97 although we preferred the ensemble of Decision Tree models as they are easy to inspect and understand. We showed that measuring distances from molecular dynamics simulations did not improve performance.ConclusionsIt is possible to predict whether a mutation inrpoBconfers resistance to rifampicin using a machine learning model trained on a combination of structural, chemical and evolutionary features, however performance is moderate and training is complicated by the lack of data.

Mitochondrial segmentation and function prediction in live-cell images with deep learning

Mitochondrial morphology and function are intrinsically linked, indicating the opportunity to predict functions by analyzing morphological features in live-cell imaging. Herein, we introduce MoDL, a deep learning algorithm for mitochondrial image segmentation and function prediction. Trained on a dataset of 20,000 manually labeled mitochondria from super-resolution (SR) images, MoDL achieves superior segmentation accuracy, enabling comprehensive morphological analysis. Furthermore, MoDL predicts mitochondrial functions by employing an ensemble learning strategy, powered by an extended training dataset of over 100,000 SR images, each annotated with functional data from biochemical assays. By leveraging this large dataset alongside data fine-tuning and retraining, MoDL demonstrates the ability to precisely predict functions of heterogeneous mitochondria from unseen cell types through small sample size training. Our results highlight the MoDL’s potential to significantly impact mitochondrial research and drug discovery, illustrating its utility in exploring the complex relationship between mitochondrial form and function within a wide range of biological contexts.

GWAS-Assisted and multi-trait genomic prediction for improvement of seed yield and canning quality traits in a black bean breeding panel.

In recent years, black beans (Phaseolus vulgaris L.) have gained popularity in the U.S., with improved seed yield and canning quality being critical traits for new cultivars. Achieving genetic gains in these traits is often challenging due to negative trait associations and the need for specialized equipment and trained sensory panels for evaluation. This study investigates the integration of genomics and phenomics to enhance selection accuracy for these complex traits. We evaluated the prediction accuracy of single and multi-trait genomic prediction (GP) models, incorporating near-infrared spectroscopy (NIRS) data and markers identified through genome-wide association studies (GWAS). The models demonstrated moderate prediction accuracies for yield and canning appearance, and high accuracies for color retention. No significant differences were found between single-trait and multi-trait models within the same breeding cycle. However, across breeding cycles, multi-trait models outperformed single-trait models by up to 45% and 63% for canning appearance and seed yield, respectively. Interestingly, incorporating significant SNP markers identified by GWAS and NIRS data into the models tended to decrease prediction accuracy both within and between breeding cycles. As genotypes from the new breeding cycle were included, the models' prediction accuracy generally increased. Our findings underscore the potential of multi-trait models to enhance the prediction of complex traits such as seed yield and canning quality in dry beans and highlight the importance of continually updating the training dataset for effective GP implementation in dry bean breeding.

PHARAOH: A collaborative crowdsourcing platform for phenotyping and regional analysis of histology

Deep learning has proven capable of automating key aspects of histopathologic analysis. However, its context-specific nature and continued reliance on large expert-annotated training datasets hinders the development of a critical mass of applications to garner widespread adoption in clinical/research workflows. Here, we present an online collaborative platform that streamlines tissue image annotation to promote the development and sharing of custom computer vision models for PHenotyping And Regional Analysis Of Histology (PHARAOH; https://www.pathologyreports.ai/). Specifically, PHARAOH uses a weakly supervised, human-in-the-loop learning framework whereby patch-level image features are leveraged to organize large swaths of tissue into morphologically-uniform clusters for batched annotation by human experts. By providing cluster-level labels on only a handful of cases, we show how custom PHARAOH models can be developed efficiently and used to guide the quantification of cellular features that correlate with molecular, pathologic and patient outcome data. Moreover, by using our PHARAOH pipeline, we showcase how correlation of cohort-level cytoarchitectural features with accompanying biological and outcome data can help systematically devise interpretable morphometric models of disease. Both the custom model design and feature extraction pipelines are amenable to crowdsourcing, positioning PHARAOH to become a fully scalable, systems-level solution for the expansion, generalization and cataloging of computational pathology applications.

System- and sample-agnostic isotropic three-dimensional microscopy by weakly physics-informed, domain-shift-resistant axial deblurring

Three-dimensional subcellular imaging is essential for biomedical research, but the diffraction limit of optical microscopy compromises axial resolution, hindering accurate three-dimensional structural analysis. This challenge is particularly pronounced in label-free imaging of thick, heterogeneous tissues, where assumptions about data distribution (e.g. sparsity, label-specific distribution, and lateral-axial similarity) and system priors (e.g. independent and identically distributed noise and linear shift-invariant point-spread functions are often invalid. Here, we introduce SSAI-3D, a weakly physics-informed, domain-shift-resistant framework for robust isotropic three-dimensional imaging. SSAI-3D enables robust axial deblurring by generating a diverse, noise-resilient, sample-informed training dataset and sparsely fine-tuning a large pre-trained blind deblurring network. SSAI-3D is applied to label-free nonlinear imaging of living organoids, freshly excised human endometrium tissue, and mouse whisker pads, and further validated in publicly available ground-truth-paired experimental datasets of three-dimensional heterogeneous biological tissues with unknown blurring and noise across different microscopy systems.

Underwater DVL Optimization Network (UDON): A Learning-Based DVL Velocity Optimizing Method for Underwater Navigation

As the exploration of marine resources continues to deepen, the utilization of Autonomous Underwater Vehicles (AUVs) for conducting marine resource surveys and underwater environmental mapping has become a common practice. In order to successfully accomplish exploration missions, AUVs require high-precision underwater navigation information as support. A Strapdown Inertial Navigation System (SINS) can provide AUVs with accurate attitude and heading information, while a Doppler Velocity Log (DVL) is capable of measuring the velocity vector of the AUVs. Therefore, the integrated SINS/DVL navigation system can furnish the necessary navigational information required by an AUV. In response to the issue of DVL being susceptible to external environmental interference, leading to reduced measurement accuracy, this paper proposes an end-to-end deep-learning approach to enhance the accuracy of DVL velocity vector measurements. The utilization of the raw measurement data from an Inertial Measurement Unit (IMU), which includes gyroscopes and accelerometers, to assist the DVL in velocity vector estimation and to refine it towards the Global Positioning System (GPS) velocity vector, compensates for the external environmental interference affecting the DVL, therefore enhancing the navigation accuracy. To evaluate the proposed method, we conducted lake experiments using SINS and DVL equipment, from which the collected data were organized into a dataset for training and assessing the model. The research results show that the DVL vector predicted by our model can achieve a maximum improvement of 69.26% in terms of root mean square error and a maximum improvement of 78.62% in terms of relative trajectory error.

Artificial Intelligence Application in Diagnosing, Classifying, Localizing, Detecting and Estimation the Severity of Skin Condition in Aesthetic Medicine: A Review

ABSTRACTBackgroundThe integration of artificial intelligence (AI) and machine learning (ML) has revolutionized aesthetic medicine, enhancing the diagnosis, classification, and treatment of skin conditions. These technologies offer high precision, personalized care, and the potential to reduce human error. This review aimed to evaluate the current applications of AI and ML in aesthetic medicine, focusing on studies graded as Level I or II evidence by the Oxford Centre for Evidence‐Based Medicine (CEBM).MethodsA comprehensive search of MEDLINE, PubMed, and Ovid databases identified studies employing AI and ML for diagnosing and managing skin conditions. Studies were included if they demonstrated high diagnostic accuracy, improved treatment personalization, or other measurable clinical outcomes.ResultsAI and ML systems showed high accuracy in detecting and diagnosing conditions such as skin cancer, acne, psoriasis, and seborrheic dermatitis. AI‐based platforms facilitated personalized treatment plans, enhancing therapeutic outcomes while minimizing errors. The integration of AI reduced diagnostic time and lowered healthcare costs, demonstrating significant potential for improving patient care. However, challenges such as algorithmic bias, data privacy concerns, and the need for high‐quality training datasets were highlighted.ConclusionAI and ML have transformative potential in aesthetic medicine, offering improved diagnostic precision, enhanced patient outcomes, and cost reductions. Addressing limitations related to algorithm bias, regulatory oversight, and data quality is essential to fully realize the benefits of AI in clinical practice. Future research should focus on developing robust, ethical, and regulatory‐compliant AI solutions.

Cultural Sensitivity in AI Language Learning: Using NLP to Enhance Language Understanding Across Cultural Contexts

The aim of this paper is to demonstrate the integration of cultural awareness into AI language learning systems using Natural Language Processing (NLP) models. Recognising that the traditional methods of language learning tend not to capture cultural variations, the studies focus on how culturally relevant NLP models (GPT-3, BERT, RNNs) promote language acquisition and cross-cultural integration. Data preprocessing, cultural data augmentation and model training was used to encode cultural variables such as greetings, politeness, and contextually relevant expressions into training datasets. Experimental results suggest that culturally sensitive models outperform the generic model on tasks that involve cultural awareness, such as politeness detection, cultural phrase recognition, and tone sensitivity. GPT-3 for example improved the accuracy of politeness detection tasks from 72.4% to 85.7%. These results highlight the need to bring cultural awareness into AI to develop engaging, context-sensitive language learning. It concludes that culturally based NLP models enhance learner capacity for real-world communication and foster global cultural awareness. Future work should focus on expanding culturally diverse datasets and refining AI models to address subtle cultural complexities

Artificial intelligence as a transforming factor in motility disorders–automatic detection of motility patterns in high-resolution anorectal manometry

High-resolution anorectal manometry (HR-ARM) is the gold standard for anorectal functional disorders’ evaluation, despite being limited by its accessibility and complex data analysis. The London Protocol and Classification were developed to standardize anorectal motility patterns classification. This proof-of-concept study aims to develop and validate an artificial intelligence model for identification and differentiation of disorders of anal tone and contractility in HR-ARM. A dataset of 701 HR-ARM exams from a tertiary center, classified according to London Classification, was used to develop and test multiple machine learning (ML) algorithms. The exams were divided in a training and testing dataset with a 80/20% ratio. The testing dataset was used for models’ evaluation through its accuracy, sensitivity, specificity, positive and negative predictive values and area under the receiving-operating characteristic curve. LGBM Classifier had the best performance, with an accuracy of 87.0% for identifying disorders of anal tone and contractility. Different ML models excelled in distinguishing specific disorders of anal tone and contractility, with accuracy over 90.0%. This is the first worldwide study proving the accuracy of a ML model for differentiation of motility patterns in HR-ARM, demonstrating the value of artificial intelligence models in optimizing HR-ARM availability while reducing interobserver variability and increasing accuracy.

Behavior modeling for a new flexure-based mechanism by Hunger Game Search and physics-guided artificial neural network

Compliant mechanism has some advantages and has been widely applied in many accurate positioning systems. However, modeling the compliant mechanism behavior has suffered from many challenges, such as unstable results, and the limitation of training data set. In the field of compliant mechanism modeling, there has been no research interested in applying meta-heuristics optimization algorithms to optimize the weights and biases of the neural network globally. Additionally, the Physics-Guided Artificial Neural Network, a research direction that has received much attention recently, has not been considered in problems related to compliant mechanisms. In order to surmount those drawbacks, this paper pioneers a new approach to model behaviors of a compliant mechanism using the Hunger Game Search and the Physics-Guided Artificial Neural Network. The Hunger Game Search can directly search the model’s weights and biases so that the target function that takes advantages of both physical and data information can be minimized. The investigations on diverse training set ratios and the ANOVA tests at the significance level reveal that using the Hunger Game Search can result in smaller errors than using the backpropagation method. Furthermore, applying the Hunger Game Search to a Physics-Guided Artificial Neural Network not only can reduce error but also can increase convergence speed compared to applying Hunger Game Search to conventional neural networks. Those results demonstrate the potential of the proposed method in modeling the behavior of compliant mechanisms.

Deep guided wave convolution neural network committee-based multi-path fusion diagnosis method for fatigue corner crack

Accurate diagnosis of crack size is a critical task for guided wave (GW)-based structural health monitoring (SHM). However, fatigue cracks would have complex morphology due to complex structural geometries and loading conditions, in which multiple dimension characteristics, like crack length, depth, and angle are involved. It is challenging to quantitatively evaluate these characteristics with GW signals from a single excitation-sensing path. This paper proposes a novel deep guided wave convolution neural network (CNN) committee-based multi-path GW fusion diagnosis method, aiming at quantitative evaluation of dimension characteristics of the complex fatigue damage. GW signals from multiple excitation-sensing paths are synthesized as a high-dimension input image to enhance the effects of the fatigue crack. Besides, the deep GW-CNN committee is developed for damage quantification, in which each GW-CNN is trained with a portion of the training dataset to reduce the impact of small sample size. The proposed method is validated on fatigue tests of landing gear beam specimens under variable amplitude loading, which is designed referring to the critical region of a real aircraft and its fatigue crack presents as a corner crack. The leave-one-out validation results show the effectiveness of the proposed method, especially improvements in the diagnosis of small cracks.

Leveraging descriptor learning and functional map-based shape matching for automated anatomical Landmarking in mouse mandibles.

Geometric morphometrics is used in the biological sciences to quantify morphological traits. However, the need for manual landmark placement hampers scalability, which is both time-consuming, labor-intensive, and open to human error. The selected landmarks embody a specific hypothesis regarding the critical geometry relevant to the biological question. Any adjustment to this hypothesis necessitates acquiring a new set of landmarks or revising them significantly, which can be impractical for large datasets. There is a pressing need for more efficient and flexible methods for landmark placement that can adapt to different hypotheses without requiring extensive human effort. This study investigates the precision and accuracy of landmarks derived from functional correspondences obtained through the functional map framework of geometry processing. We utilize a deep functional map network to learn shape descriptors, which enable us to achieve functional map-based and point-to-point correspondences between specimens in our dataset. Our methodology involves automating the landmarking process by interrogating these maps to identify corresponding landmarks, using manually placed landmarks from the entire dataset as a reference. We apply our method to a dataset of rodent mandibles and compare its performance to MALPACA's, a standard tool for automatic landmark placement. Our model demonstrates a speed improvement compared to MALPACA while maintaining a competitive level of accuracy. Although MALPACA typically shows the lowest RMSE, our models perform comparably well, particularly with smaller training datasets, indicating strong generalizability. Visual assessments confirm the precision of our automated landmark placements, with deviations consistently falling within an acceptable range for MALPACA estimates. Our results underscore the potential of unsupervised learning models in anatomical landmark placement, presenting a practical and efficient alternative to traditional methods. Our approach saves significant time and effort and provides the flexibility to adapt to different hypotheses about critical geometrical features without the need for manual re-acquisition of landmarks. This advancement can significantly enhance the scalability and applicability of geometric morphometrics, making it more feasible for large datasets and diverse biological studies.

NNFit: A Self-Supervised Deep Learning Method for Accelerated Quantification of High- Resolution Short Echo Time MR Spectroscopy Datasets.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To develop and evaluate the performance of NNFit, a self-supervised deep-learning method for quantification of high-resolution short echo-time (TE) echo-planar spectroscopic imaging (EPSI) datasets, with the goal of addressing the computational bottleneck of conventional spectral quantification methods in the clinical workflow. Materials and Methods This retrospective study included 89 short-TE whole-brain EPSI/GRAPPA scans from clinical trials for glioblastoma (Trial 1, May 2014-October 2018) and major-depressive-disorder (Trial 2, 2022- 2023). The training dataset included 685k spectra from 20 participants (60 scans) in Trial 1. The testing-dataset included 115k spectra from 5 participants (13 scans) in Trial 1 and 145k spectra from 7 participants (16 scans) in Trial 2. A comparative analysis was performed between NNFit and a widely used parametric-modeling spectral quantitation method (FITT). Metabolite maps generated by each method were compared using the structural- similarity-index-measure (SSIM) and linear-correlation-coefficient (R2). Radiation treatment volumes for glioblastoma based on the metabolite maps were compared with the Dice-coefficient and a two-tailed t test. Results Average SSIM and R2 scores for Trial 1 test set data were 0.91/0.90 (choline), 0.93/0.93 (creatine), 0.93/0.93 (n-acetylaspartate), 0.80/0.72 (myo-inositol), and 0.59/0.47 (glutamate + glutamine). Average scores for Trial 2 test set data were 0.95/0.95, 0.98/0.97, 0.98/0.98, 0.92/0.92, and 0.79/0.81 respectively. The treatment volumes had average Dice coefficient of 0.92. NNFit's average processing time was 90.1 seconds, whereas FITT took 52.9 minutes on average. Conclusion This study demonstrates that a deep learning approach to spectral quantitation offers comparable performance to conventional quantification methods for EPSI data, but with faster processing at short-TE. ©RSNA, 2025.

A prognostic survival model based on endocrine-related gene expression in acute myelogenous leukemia.

Accurate prediction of survival in patients with acute myelogenous leukemia (AML) is challenging. Therefore, we developed a predictive survival model using endocrine-related gene expression to identify an endocrine signature for accurate stratification of AML prognosis. RNA matrices and clinical data for AML were downloaded from a training dataset (GEO) and two validation datasets (TCGA and TARGET). In relation to the survival outcome, a risk model was constructed by incorporating seven endocrine-related genes. The model exhibited favorable predictive efficacy in estimating 5-year survival rates, as demonstrated by both the training and validation cohorts. Multivariable analysis revealed that the endocrine signature demonstrated autonomous prognostic significance in the aforementioned cohorts. Prediction accuracy for 5-year overall survival increased using a nomogram combining endocrine risk score and classical prognostic factors compared with using classical prognostic factors alone. The model predictions were confirmed using AML cell lines. The endocrine-related prognostic model established in this study improves AML survival prediction accuracy.

Analysis of borehole strain anomalies before the 2017 Jiuzhaigou Ms 7.0 earthquake based on a graph neural network

Abstract. On 8 August 2017, a strong earthquake of magnitude 7.0 occurred in Jiuzhaigou, Sichuan Province, China. To assess pre-earthquake anomalies, we utilized variational mode decomposition to preprocess borehole strain observation data and combined them with a graph WaveNet neural network model to process data from multiple stations. We obtained 1-year data from four stations near the epicenter as the training dataset and data from 1 January to 10 August 2017 as the test dataset. For the prediction results of the variational mode decomposition–graph WaveNet model, the anomalous days were extracted using statistical methods, and the results of anomalous-day accumulation at multiple stations showed that an increase in the number of anomalous days occurred 15–32 d before the earthquake. The acceleration effect of anomalous accumulation was most obvious 20 d before the earthquake, and an increase in the number of anomalous days also occurred in the 1 to 3 d post-earthquake. We tentatively deduce that the pre-earthquake anomalies are caused by the diffusion of strain energy near the epicenter during the accumulation process, which can be used as a signal of pre-seismic anomalies, whereas the post-earthquake anomalies are caused by the frequent occurrence of aftershocks.

Application of Artificial Intelligence in Radiological Image Analysis for Pulmonary Disease Diagnosis: A Review of Current Methods and Challenges

Introduction and purposeArtificial intelligence (AI), particularly machine learning (ML) and deep learning (DL), is revolutionizing radiology by improving diagnostic accuracy and efficiency. This paper examines AI applications, especially convolutional neural networks (CNNs), in diagnosing pulmonary diseases, such as pneumonia, tuberculosis, and lung cancer. The goal is to explore the impact of these technologies and assess challenges in their integration into clinical practice. Material and methodsThis review is based on articles from the PubMed database, published between 2015 and 2024, using keywords such as artificial intelligence in radiology, AI in medicine, AI in chest X-ray, and AI in chest-CT. ResultsAI, driven by ML and DL, has significantly enhanced medical imaging analysis, automating tasks that require expert interpretation. CNNs excel in processing raw image data and identifying hierarchical features, surpassing traditional methods in diagnosing lung diseases from radiographs and CT scans. AI systems demonstrate exceptional accuracy in detecting pneumonia, tuberculosis, and lung cancer, providing rapid, consistent results, particularly valuable in resource-limited settings. However, challenges persist, including the need for diverse training datasets, model interpretability, and integration into existing workflows. ConclusionsAI, especially CNN-based DL models, is reshaping radiology by advancing diagnostic capabilities. While it often outperforms traditional methods, AI is best used to complement human expertise. Overcoming challenges in data quality, system integration, and training is essential for broader clinical adoption. Continued research will enhance AI’s reliability and utility, ultimately improving patient outcomes.