Large Image Datasets Research Articles

Prediction of Dry Mouth Condition Using Radiomics Features from Tongue Diagnosis Image

Xerostomia, commonly known as dry mouth, is characterized by reduced salivary secretion, which can lead to various oral health issues and discomfort. In this paper, we propose a novel, non-invasive method for predicting xerostomia through the analysis of tongue images. To predict salivary gland secretion from tongue images, we collected images from patients who visited the hospital with complaints of dry mouth and measured their saliva secretion. Features were extracted from these tongue images, and correlation analysis was performed using machine learning techniques to assess the relationship between the extracted features and measured saliva secretion. We obtained tongue images and saliva secretion measurements from 176 patients. Images were cropped to 100 × 100 pixels, resulting in 462 features. The dataset was divided into training and test sets, consisting of 160 and 16 samples, respectively. The correlation coefficients for the training and test datasets were 0.9496 and 0.9415, respectively, while the correlation coefficient for the entire dataset was 0.9482. The estimated linear equation was y = 0.9244x + 2.1664. This study aimed to predict salivary gland secretion based on tongue images. By extracting features from color images and employing a neural network machine learning model, we estimated salivary gland secretion. With a sufficiently large dataset of tongue images, further advancements in regression analysis using deep learning techniques could enhance the accuracy of these predictions.

Estimating TYLCV resistance level using RGBD sensors in production greenhouse conditions

Automated phenotyping is the task of automatically measuring plant attributes to help farmers and breeders in developing and growing strong robust plants. An automated tool for early illness detection can accelerate the process of identifying plant resistance and quickly pinpoint problematic breeding. Many such phenotyping tasks can be achieved by analyzing images from simple, low cost, RGB-D sensors. In this paper we focused on a particular case study — identifying the resistance level of tomato hybrids to the tomato yellow leaf curl virus (TYLCV) in production greenhouses. This is a difficult task, as separating between resistance levels based on images is difficult even for expert breeders. We collected a large dataset of images from an experiment containing many tomato hybrids with varying resistance levels. We used the depth information to identify the topmost part of the tomato plant. We then used deep learning models to classify the various resistance levels. For identifying plants with visual symptoms, our methods achieved an accuracy of 0.928, a precision of 0.934, and a recall of 0.95. In the multi-class case we achieved an accuracy of 0.76 in identifying the correct level, and an error of 0.278. Our methods are not particularly tailored for the specific task, and can be extended to other tasks that identify various plant diseases with visual symptoms such as ToBRFV,mildew, ToMV and others.

HOLISMOKES. XI. Evaluation of supervised neural networks for strong-lens searches in ground-based imaging surveys

While supervised neural networks have become state of the art for identifying the rare strong gravitational lenses from large imaging data sets, their selection remains significantly affected by the large number and diversity of non-lens contaminants. This work evaluates and compares systematically the performance of neural networks in order to move towards a rapid selection of galaxy-scale strong lenses with minimal human input in the era of deep, wide-scale surveys. We used multiband images from PDR2 of the Hyper-Suprime Cam (HSC) Wide survey to build test sets mimicking an actual classification experiment, with 189 securely-identified strong lenses from the literature over the HSC footprint and 70\,910 non-lens galaxies in COSMOS covering representative lens-like morphologies. Multiple networks were trained on different sets of realistic strong-lens simulations and non-lens galaxies, with various architectures and data preprocessing, mainly using the deepest $gri$-bands. Most networks reached excellent area under the Receiver Operating Characteristic (ROC) curves on the test set of 71,099 objects, and we determined the ingredients to optimize the true positive rate for a total number of false positives equal to zero or 10 $ and TPR$_ $). The overall performances strongly depend on the construction of the ground-truth training data and they typically, but not systematically, improve using our baseline residual network architecture presented in HOLISMOKES VI. TPR$_ $ tends to be higher for ResNets (simeq 10--40<!PCT!>) compared to AlexNet-like networks or G-CNNs. Improvements are found when (1) applying random shifts to the image centroids, (2) using square-root scaled images to enhance faint arcs, (3) adding $z$-band to the otherwise used $gri$-bands, or (4) using random viewpoints of the original images. In contrast, we find no improvement when adding $g - i$ difference images (where alpha is a tuned constant) to subtract emission from the central galaxy. The most significant gain is obtained with committees of networks trained on different data sets, with a moderate overlap between populations of false positives. Nearly-perfect invariance to image quality can be achieved by using realistic PSF models in our lens simulation pipeline, and by training networks either with large number of bands, or jointly with the PSF and science frames. Overall, we show the possibility to reach a $ as high as 60<!PCT!> for the test sets under consideration, which opens promising perspectives for pure selection of strong lenses without human input using the Rubin Observatory and other forthcoming ground-based surveys.

Machine learning techniques for diagnosis of rare diseases from medical images

Introduction/Background: Rare diseases are life-threatening or chronically debilitating conditions that affect a small percentage of the population. Early and accurate diagnosis of rare diseases is challenging due to their complexity and limited understanding. Conventional diagnostic methods are often inadequate, time-consuming, and expensive. Machine learning (ML) has emerged as a promising technique for the analysis of medical images to support the diagnosis of various diseases including rare disorders. ML algorithms can learn complex patterns from large medical imaging datasets to aid clinicians in disease diagnosis, prognosis, and detection of complications. Materials and Methods: A structured search was conducted in electronic databases, including PubMed, Scopus, Web of Science, and Google Scholar to identify peer-reviewed articles published between 2013 to 2023 related to ML-based diagnosis of rare diseases from medical images. A total of 187 articles were identified after removing duplicates. The titles and abstracts of retrieved articles were screened to determine their relevance based on the research objective. Finally, 51 full-text articles were selected for the final review. The key data extracted from selected studies included diseases, imaging modalities, ML algorithms, performance metrics, datasets, and limitations. Results: Most studies evaluated deep learning-based ML techniques, with convolutional neural networks (CNN) being the most applied algorithm. CNNs were mainly used to classify rare diseases based on specific imaging features in X-rays, CT, and MRI scans. Diseases frequently investigated included cardiovascular, neurological, and rare genetic disorders. Publicly available datasets such as Deep Lesion, MSD, and ChestX-ray14 were commonly utilized. Most studies reported high classification accuracy ranging from 80-95% on test datasets. However, limited generalizability due to small private datasets and lack of external validation were limitations. Discussion: The results demonstrate the potential of deep learning for the image-based diagnosis of rare diseases. Features learned from large imaging data sets enabled CNNs to distinguish subtle abnormalities. Hybrid models combining CNNs with other techniques like recurrent neural networks further improved performance. Overall, ML showed promise as a decision support tool. However, more multi-center validation studies are needed using larger and diverse datasets before clinical adoption. Standardization of performance metrics and reporting is also required to establish reliability and generalizability. Conclusion: This comprehensive review provided insights into ongoing research applying ML to medical images for rare disease diagnosis. While initial results are encouraging, further work is still necessary before ML can be reliably used in clinical decision making. Larger collaborative efforts and open-source datasets are needed to advance the field. Standardization of ML frameworks can also promote reproducible research and comparison of different methodologies.

Machine-learning to analyze human-building interactions: How do people use mobile solar protections?

Buildings significantly impact urban energy consumption. Mobile passive solutions, such as manual solar protections, can mitigate heat gains, but their effectiveness depends on occupants’ decisions. Analyzing occupant adjustments of manual systems is challenging due to the complexity of human behavior. While multiple studies have observed human-building interactions, analyzing large buildings over extended periods remains a technical challenge. This study examines 359 manually controlled solar protections over a year using systematic photographic data. To manage the large volume of images, an unsupervised machine learning algorithm was employed to cluster similar solar protection positions for each window, which were then manually tagged to identify their positions. Results show that solar protections are mostly closed year-round (28 % aperture on average), with minimal differences between occupied and non-occupied days and between warm and cool seasons. Notably, one-third of the solar protections were not operated throughout the year. Our results provide insights into the usage of manual façade systems and offer valuable knowledge for energy simulation. Methodologically, the use of machine-learning algorithms presents a new way to process large datasets of images, emphasizing the importance of prioritizing quality over quantity and strategic data collection. Future research can apply this methodology to different building types and climates to gain broader insights into occupant behavior and solar protection interactions.

Machine learning-enabled computer vision for plant phenotyping: a primer on AI/ML and case study on stomatal patterning.

Artificial intelligence and machine learning (AI/ML) can be used to automatically analyze large image datasets. One valuable application of this approach is estimation of plant trait data contained within images. Here we review 39 papers that describe the development and/or application of such models for estimation of stomatal traits from epidermal micrographs. In doing so, we hope to provide plant biologists with a foundational understanding of AI/ML and summarize the current capabilities and limitations of published tools. While most models show human-level performance for stomatal density (SD) quantification at superhuman speed, they are often likely to be limited in how broadly they can be applied across phenotypic diversity associated with genetic, environmental or developmental variation. Other models can make predictions across greater phenotypic diversity and/or additional stomatal/epidermal traits, but require significantly greater time investment to generate ground-truth data. We discuss the challenges and opportunities presented by AI/ML-enabled computer vision analysis, and make recommendations for future work to advance accelerated stomatal phenotyping.

Can artificial intelligence aid the urologists in detecting bladder cancer?

ABSTRACT Introduction: The emergence of artificial intelligence (AI)-based support system endoscopy, including cystoscopy, has shown promising results by training deep learning algorithms with large datasets of images and videos. This AI-aided cystoscopy has the potential to significantly transform the urological practice by assisting the urologists in identifying malignant areas, especially considering the diverse appearance of these lesions. Methods: Four databases, the PubMed, ProQuest, EBSCOHost, and ScienceDirect were searched, along with a manual hand search. Prospective and retrospective studies, experimental studies, cross-sectional studies, and case–control studies assessing the utilization of AI for the detection of bladder cancer through cystoscopy and comparing with the histopathology results as the reference standard were included. The following terms and their variants were used: “artificial intelligence,” “cystoscopy,” and “bladder cancer.” The risk of bias was assessed using the Quality Assessment of Diagnostic Accuracy Studies-2 tool. A random effects model was used to calculate the pooled sensitivity and specificity. The Moses–Littenberg model was used to derive the Summary Receiver Operating Characteristics (SROC) curve. Results: Five studies were selected for the analysis. Pooled sensitivity and specificity were 0.953 (95% confidence interval [CI]: 0.908–0.976) and 0.957 (95% CI: 0.923–0.977), respectively. Pooled diagnostic odd ratio was 449.79 (95% CI: 12.42–887.17). SROC curve (area under the curve: 0.988, 95% CI: 0.982–0.994) indicated a strong discriminating power of AI-aided cystoscopy in differentiation normal or benign bladder lesions from the malignant ones. Conclusions: Although the utilization of AI for aiding in the detection of bladder cancer through cystoscopy remains questionable, it has shown encouraging potential for enhancing the detection rates. Future studies should concentrate on identification of the patients groups which could derive maximum benefit from accurate identification of the bladder cancer, such as those with intermediate or high-risk invasive tumors.

RFMiD: Retinal Image Analysis for multi-Disease Detection challenge

In the last decades, many publicly available large fundus image datasets have been collected for diabetic retinopathy, glaucoma, and age-related macular degeneration, and a few other frequent pathologies. These publicly available datasets were used to develop a computer-aided disease diagnosis system by training deep learning models to detect these frequent pathologies. One challenge limiting the adoption of a such system by the ophthalmologist is, computer-aided disease diagnosis system ignores sight-threatening rare pathologies such as central retinal artery occlusion or anterior ischemic optic neuropathy and others that ophthalmologists currently detect. Aiming to advance the state-of-the-art in automatic ocular disease classification of frequent diseases along with the rare pathologies, a grand challenge on “Retinal Image Analysis for multi-Disease Detection” was organized in conjunction with the IEEE International Symposium on Biomedical Imaging (ISBI - 2021). This paper, reports the challenge organization, dataset, top-performing participants solutions, evaluation measures, and results based on a new “Retinal Fundus Multi-disease Image Dataset” (RFMiD). There were two principal sub-challenges: disease screening (i.e. presence versus absence of pathology — a binary classification problem) and disease/pathology classification (a 28-class multi-label classification problem). It received a positive response from the scientific community with 74 submissions by individuals/teams that effectively entered in this challenge. The top-performing methodologies utilized a blend of data-preprocessing, data augmentation, pre-trained model, and model ensembling. This multi-disease (frequent and rare pathologies) detection will enable the development of generalizable models for screening the retina, unlike the previous efforts that focused on the detection of specific diseases.

Synthetic Meets Authentic: Leveraging LLM Generated Datasets for YOLO11 and YOLOv10-Based Apple Detection through Machine Vision Sensors

Training machine learning (ML) models for artificial intelligence (AI) and computer vision-based object detection process typically requires large, labeled datasets, a process often burdened by significant human effort and high costs associated with imaging systems and image acquisition. This research aimed to simplify image data collection for object detection in orchards by avoiding traditional fieldwork with different imaging sensors. Utilizing OpenAI's DALLE, a large language model (LLM) for realistic image generation, we generated and annotated a cost-effective dataset. This dataset, exclusively generated by LLM, was then utilized to train two state-of-the-art deep learning models: YOLOV10 and YOLO11. The YOLO11 model for apple detection was trained with its five configurations (YOLO11n, YOLO11s, YOLO11m, YOLO11l and YOLO11x), and YOLOv10 model with its six configurations (YOLOv10n, YOLOv10s, YOLOv10m, YOLOv10b, YOLOv10l and YOLOv10x), which was then tested with real-world (outdoor orchard) images captured by a digital (Nikon D5100) camera and a consumer RGB-D camera (Microsoft Azure Kinect). YOLO11 outperformed YOLOv10 as YOLO11x and YOLO11n exhibited superior precision of 0.917 and 0.916, respectively. Furthermore, YOLO11l demonstrated the highest recall among its counterparts, achieving a recall of 0.889. Likewise, the YOLO11n variant excelled in terms of mean average precision (mAP@50), achieving the highest value of 0.958. Validation tests against actual images collected through a digital camera (Nikon D5100) over Scilate apple variety in a commercial orchard environment showed a highest precision of 0.874 for YOLO11s, recall of 0.877 for YOLO11l and mAP@50 of 0.91 for YOLO11x. Additionally, validation test against actual images collected through a Microsoft Azure camera over the same orchard showed a highest precision, recall and mAP@50 respectively of 0.924, 0.781 and 0.855 with YOLO11x. All variants of YOLO11 surprisingly demonstrated a pre-processing time of just 0.2 milliseconds (ms), which was faster than any variant of YOLOv10. The fastest inference time for the YOLO11n model using the training dataset generated by the language model was 3.2 ms, while YOLOv10n, fastest among YOLOv10 variants, had a longer inference time of 5.5 ms. Likewise, the fastest inference time for the sensor-based images was 7.1 ms (for Nikon D5100 camera images) and 4.7 ms (for Azure images) with YOLO11n. This study presents a pathway for generating large image datasets using LLM in challenging agricultural fields with minimal or no labor-intensive efforts in field data-collection, which could accelerate the development and deployment of computer vision and robotic technologies in orchard environments.

The Study of Echocardiography of Left Ventricle Segmentation Combining Transformer and Convolutional Neural Networks.

Accurate prediction of echocardiographic parameters is essential for diagnosis and treatment of cardiac disease, especially for segmentation of the left ventricle to obtain measurements such as left ventricular ejection fraction and volume. However, manually outlining left ventricle on echocardiographic images is a time-consuming and physician experience-dependent task. Therefore, it is crucial to develop an accurate and efficient automatic segmentation tool. Therefore, we aimed to explore a model to perform echocardiography of left ventricle segmentation by combining transformer and convolutional neural networks (CNN).ResNet-50 was used in CNN branch. The encoder-decoder architecture was used for transformer branch, which was fused to the corresponding feature maps of the CNN branches. Fusion module was used to effectively combine feature information from the CNN and transformer. Bridge attention used to increase sensitivity and prediction accuracy of model. The entire network was trained end-to-end using the binary cross-entropy with logits loss L.In this work, we propose an automatic left ventricular (LV) segmentation model based on Transformer and CNN that efficiently captures global dependencies and spatial details and create a fusion module using CBAM that fuses Transformer and CNN features. In addition, attention is also computed using multi-level fusion features to obtain the final attention segmentation map. The model was trained and evaluated on a large cardiac image dataset, EchoNet-Dynamic, with test dice coefficient of 92.4%.The results show that our model can better segment left ventricle. We also tested our model on clinical patient ultrasound images, and visualization results proved effectiveness of the model.

Early detection of monkeypox: Analysis and optimization of pretrained deep learning models using the Sparrow Search Algorithm

The global spread of monkeypox across over 40 countries is a major health challenge. Its symptoms resemble those of chickenpox and measles, complicating early diagnosis. When PCR tests are unavailable, computational lesion detection offers a promising option. This research presents a non-invasive diagnostic method using the Sparrow Search Algorithm (SpaSA). SpaSA is applied to improve accuracy in medical imaging tasks. Deep learning, especially CNNs, effectively addresses challenges in medical imaging. By combining CNNs with optimization, we analyze large imaging datasets. This study evaluates several pre-trained models on two datasets: “Monkeypox 2022” and “Images of Monkeypox.” The method classifies patients as either “normal” or “pox” (including monkeypox, chickenpox, smallpox, cowpox, and measles). The novelty lies in using SpaSA to optimize pre-trained CNNs' performance. The Xception model achieves 97.66% accuracy without optimization. SpaSA enhances VGG19's performance, reaching 99.87% accuracy. With SpaSA, VGG19 reached 99.87% accuracy in 120 seconds. SpaSA had the fastest time (120s) and lowest failure rate (5%). It outperformed Grid Search, Random Search, Bayesian Optimization, WOA, GA, PSO, and ACO. This shows mathematical optimization's impact on medical imaging. It sets new standards in healthcare diagnostics using SpaSA and CNNs. This research shows the significant role of optimization in improving medical imaging. It sets new standards for healthcare diagnostics using SpaSA and CNNs.

A Modified Cane for the Blind using Computer Vision

This study aimed to develop “A Modified Cane for the Blind using Computer Vision” that can detect possible objects and obstacles from the surroundings. A collection of a large image dataset to develop a model has been divided into three (3) datasets with the ratio of 70% for training, 20% for validation, and 10% for testing. The model was used for a mobile application to detect objects from the surroundings. A series of designs for a mobile application and a modified cane were created until the final prototype was made. The ultrasonic sensors, micro-controllers, and other electronic components were connected and tested for effectiveness and accuracy. The final prototype of a modified cane and the mobile application were combined to create a system called “a modified cane for the blind using computer vision.” The effectiveness of the system was evaluated by five (5) IT experts and five (5) blind persons in terms of functional suitability, reliability, performance efficiency, usability, security, compatibility, maintainability, portability, and the overall result with a mean of 4.46. The results of the evaluation show that the overall mean was “Excellent.” This implies that the system is effective and meets ISO 25010 standards in terms of software quality.

Simulated analogues II: a new methodology for non-parametric matching of models to observations

ABSTRACT Star formation is a multiscale problem, and only global simulations that account for the connection from the molecular cloud-scale gas flow to the accreting protostar can reflect the observed complexity of protostellar systems. Star-forming regions are characterized by supersonic turbulence, and as a result, it is not possible to simultaneously design models that account for the larger environment and in detail reproduce observed stellar systems. Instead, the stellar inventories can be matched statistically, and the best matches found that approximate specific observations. Observationally, a combination of single-dish telescopes and interferometers are now able to resolve the nearest protostellar objects on all scales from the protostellar core to the inner $10\, \mathrm{au}$. We present a new non-parametric methodology which uses high-resolution simulations and post-processing methods to match simulations and observations using deep learning. Our goal is to perform a down-selection from large data sets of synthetic images to a ranked list of best-matching candidates with respect to the observation. This is particularly useful for binary and multiple stellar systems that form in turbulent environments. The objective is to accelerate the rate at which we can do such comparisons, remove biases from hand-picking matches, and contribute to identifying the underlying physical processes that drive the creation and evolution of observed protostellar systems.

NecroGlobalGCN: Integrating micronecrosis information in HCC prognosis prediction via graph convolutional neural networks

Background and ObjectiveHepatocellular carcinoma (HCC) ranks fourth in cancer mortality, underscoring the importance of accurate prognostic predictions to improve postoperative survival rates in patients. Although micronecrosis has been shown to have high prognostic value in HCC, its application in clinical prognosis prediction requires specialized knowledge and complex calculations, which poses challenges for clinicians. It would be of interest to develop a model to help clinicians make full use of micronecrosis to assess patient survival. MethodsTo address these challenges, we propose a HCC prognosis prediction model that integrates pathological micronecrosis information through Graph Convolutional Neural Networks (GCN). This approach enables GCN to utilize micronecrosis, which has been shown to be highly correlated with prognosis, thereby significantly enhancing prognostic stratification quality. We developed our model using 3622 slides from 752 patients with primary HCC from the FAH-ZJUMS dataset and conducted internal and external validations on the FAH-ZJUMS and TCGA-LIHC datasets, respectively. ResultsOur method outperformed the baseline by 8.18% in internal validation and 9.02% in external validations. Overall, this paper presents a deep learning research paradigm that integrates HCC micronecrosis, enhancing both the accuracy and interpretability of prognostic predictions, with potential applicability to other pathological prognostic markers. ConclusionsThis study proposes a composite GCN prognostic model that integrates information on HCC micronecrosis, collecting large dataset of HCC histopathological images. This approach could assist clinicians in analyzing HCC patient survival and precisely locating and visualizing necrotic tissues that affect prognosis. Following the research paradigm outlined in this paper, other prognostic biomarker integration models with GCN could be developed, significantly enhancing the predictive performance and interpretability of prognostic model.

Lung Cancer Detection Using Convolutional Neural Networks

Lung cancer is among the most prevalent and deadly cancers worldwide. Accurate diagnosis and early detection are critical for improving lung cancer patient outcomes and survival rates. Thanks to developments in medical imaging technology, computer-aided diagnosis (CAD) systems have shown a great deal of promise in helping radiologists identify and diagnose lung cancer from medical images. Here, we present the use of convolutional neural networks (CNNs) to create an early detection system (CAD) for lung cancer. The suggested approach uses lung computed tomography (CT) scans as input and use a CNN architecture to extract high-level features from the pictures. We use transfer learning to enhance a CNN model trained on a large dataset of CT images. The CNN model has been taught to determine if a specific CT image contains lung cancer or not. We evaluate the performance of the proposed CAD system on a dataset of CT scans of the lungs from different institutions. The trial's results show that our CNN-based CAD system can reliably and precisely identify lung cancer from CT scans. We also show the comparative performance of our proposed system against the state-of-the-art machine learning methods for lung cancer prediction. In conclusion, the suggested CNN-based deep learning-based CAD system has produced encouraging results for lung cancer detection from CT scans. The approach might help radiologists identify and classify lung cancer early on, leading to better patient outcomes and survival rates. The viability and usefulness of the suggested approach in clinical practice require more study

The people behind the papers - Jake Turley.

During wound healing, epithelial cells divide, migrate and change shape. However, it is unclear how much these different cell behaviours contribute to the closure of the wound. A new paper in Development applies deep learning models to analyse large imaging datasets from the Drosophila pupa and quantify the contributions of the different cell behaviours. To learn more about the story behind the paper, we caught up with first author Jake Turley. Jake was a PhD student in Bristol, UK, but is now a Research Fellow at the Mechanobiology Institute, National University of Singapore.

Optimizing Lung Condition Categorization through a Deep Learning Approach to Chest X-ray Image Analysis

Background: Evaluating chest X-rays is a complex and high-demand task due to the intrinsic challenges associated with diagnosing a wide range of pulmonary conditions. Therefore, advanced methodologies are required to categorize multiple conditions from chest X-ray images accurately. Methods: This study introduces an optimized deep learning approach designed for the multi-label categorization of chest X-ray images, covering a broad spectrum of conditions, including lung opacity, normative pulmonary states, COVID-19, bacterial pneumonia, viral pneumonia, and tuberculosis. An optimized deep learning model based on the modified VGG16 architecture with SE blocks was developed and applied to a large dataset of chest X-ray images. The model was evaluated against state-of-the-art techniques using metrics such as accuracy, F1-score, precision, recall, and area under the curve (AUC). Results: The modified VGG16-SE model demonstrated superior performance across all evaluated metrics. The model achieved an accuracy of 98.49%, an F1-score of 98.23%, a precision of 98.41%, a recall of 98.07% and an AUC of 98.86%. Conclusion: This study provides an effective deep learning approach for categorizing chest X-rays. The model’s high performance across various lung conditions suggests its potential for integration into clinical workflows, enhancing the accuracy and speed of pulmonary disease diagnosis.

Combining Transfer Learning and Ensemble Algorithms for Improved Citrus Leaf Disease Classification

Accurate categorization and timely control of leaf diseases are crucial for citrus growth. We proposed the Multi-Models Fusion Network (MMFN) for citrus leaf diseases detection based on model fusion and transfer learning. Compared to traditional methods, the algorithm (integrating transfer learning Alexnet, VGG, and Resnet) we proposed can address the issues of limited categories, slow processing speed, and low recognition accuracy. By constructing efficient deep learning models and training and optimizing them with a large dataset of citrus leaf images, we ensured the broad applicability and accuracy of citrus leaf disease detection, achieving high-precision classification. Herein, various deep learning algorithms, including original Alexnet, VGG, Resnet, and transfer learning versions Resnet34 (Pre_Resnet34) and Resnet50 (Pre_Resnet50) were also discussed and compared. The results demonstrated that the MMFN model achieved an average accuracy of 99.72% in distinguishing between diseased and healthy leaves. Additionally, the model attained an average accuracy of 98.68% in the classification of multiple diseases (citrus huanglongbing (HLB), greasy spot disease and citrus canker), insect pests (citrus leaf miner), and deficiency disease (zinc deficiency). These findings conclusively illustrate that deep learning model fusion networks combining transfer learning and integration algorithms can automatically extract image features, enhance the automation and accuracy of disease recognition, demonstrate the significant potential and application value in citrus leaf disease classification, and potentially drive the development of smart agriculture.

A New Method for the Detection of Siliceous Microfossils on Sediment Microscope Slides Using Convolutional Neural Networks

AbstractDiatom communities preserved in sediment samples are valuable indicators for understanding the past and present dynamics of phytoplankton communities, and their response to environmental changes. These studies are traditionally achieved by counting methods using optical microscopy, a time‐consuming process that requires taxonomic expertise. With the advent of automated image acquisition workflows, large image data sets can now be acquired, but require efficient preprocessing methods. Detecting diatom frustules on microscope images is a challenge due to their low relief, diverse shapes, and tendency to aggregate, which prevent the use of traditional thresholding techniques. Deep learning algorithms have the potential to resolve these challenges, more particularly for the task of object detection. Here we explore the use of a Faster Region‐based Convolutional Neural Network model to detect siliceous biominerals, including diatoms, in microscope images of a sediment trap series from the Mediterranean Sea. Our workflow demonstrates promising results, achieving a precision score of 0.72 and a recall score of 0.74 when applied to a test set of Mediterranean diatom images. Our model performance decreases when used to detect fragments of these microfossils; it also decreases when particles are aggregated or when images are out of focus. Microfossil detection remains high when the model is used on a microscope image set of sediments from a different oceanic basin, demonstrating its potential for application in a wide range of contemporary and paleoenvironmental studies. This automated method provides a valuable tool for analyzing complex samples, particularly for rare species under‐represented in training data sets.

Generating Synthetic Data for Medical Imaging.

Artificial intelligence (AI) models for medical imaging tasks, such as classification or segmentation, require large and diverse datasets of images. However, due to privacy and ethical issues, as well as data sharing infrastructure barriers, these datasets are scarce and difficult to assemble. Synthetic medical imaging data generated by AI from existing data could address this challenge by augmenting and anonymizing real imaging data. In addition, synthetic data enable new applications, including modality translation, contrast synthesis, and professional training for radiologists. However, the use of synthetic data also poses technical and ethical challenges. These challenges include ensuring the realism and diversity of the synthesized images while keeping data unidentifiable, evaluating the performance and generalizability of models trained on synthetic data, and high computational costs. Since existing regulations are not sufficient to guarantee the safe and ethical use of synthetic images, it becomes evident that updated laws and more rigorous oversight are needed. Regulatory bodies, physicians, and AI developers should collaborate to develop, maintain, and continually refine best practices for synthetic data. This review aims to provide an overview of the current knowledge of synthetic data in medical imaging and highlights current key challenges in the field to guide future research and development.