Augmented Dataset Research Articles

Improving quality control of whole slide images by explicit artifact augmentation

The problem of artifacts in whole slide image acquisition, prevalent in both clinical workflows and research-oriented settings, necessitates human intervention and re-scanning. Overcoming this challenge requires developing quality control algorithms, that are hindered by the limited availability of relevant annotated data in histopathology. The manual annotation of ground-truth for artifact detection methods is expensive and time-consuming. This work addresses the issue by proposing a method dedicated to augmenting whole slide images with artifacts. The tool seamlessly generates and blends artifacts from an external library to a given histopathology dataset. The augmented datasets are then utilized to train artifact classification methods. The evaluation shows their usefulness in classification of the artifacts, where they show an improvement from 0.10 to 0.01 AUROC depending on the artifact type. The framework, model, weights, and ground-truth annotations are freely released to facilitate open science and reproducible research.

A Geometric Approach to Textual Augmented Data Filtering

Abstract Data augmentation is necessary if the amount of training data is insufficient for supervised learning. For natural language processing tasks, obtaining good quality augmented data is not easy. This paper introduces GATFilter, a novel method for filtering out inappropriate augmented textual data for text classification (TC). Utilizing geometric concepts, more specifically the principle component and convex hull analyses, this method adeptly preserves the semantic integrity of words within augmented texts. GATFilter is versatile and applicable across various types of textual augmentation methods. Experiments using several datasets and augmentation strategies showed that classifiers trained with GATFilter-filtered augmented data sets showed improvements in key performance metrics, including accuracy, precision, recall, and F1 score. The method’s efficacy is notably influenced by the quality of the underlying augmentation techniques, indicating its potential to complement and refine various text augmentation strategies. Furthermore, our analysis showed that GATFilter is particularly able to amplify the effectiveness of methods that generate good quality augmented data. GATFilter is openly available online on Github1, and as a Python package2.

Simulation data-driven fault diagnosis method for metro traction motor bearings under small samples and missing fault samples

Abstract Traction motor bearings are crucial for guaranteeing the safe operation of metro vehicles. However, in the metro traction motor bearing fault diagnosis, there are usually problems of small samples and missing fault samples, leading to inaccurate results. Therefore, a novel bearing fault diagnosis method utilizing a track-vehicle-bearing coupled dynamic model and the improved deep convolutional generative adversarial network-multiscale convolutional neural network with mixed attention (IDCG-MAMCNN) model is proposed in this paper. The IDCG-MAMCNN model combines an improved deep convolutional generative adversarial network (IDCGAN) with a multi-scale convolutional neural network with mixed attention (MA-MCNN). Specifically, simulation data is first provided by the coupled dynamic model to supplement missing fault samples. Secondly, the IDCGAN, along with a training method that involves pre-training models with simulation samples and fine-tuning models with experimental samples, is introduced to generate high-quality samples and augment experimental samples under small samples. Lastly, the MA-MCNN serves as the classification model, trained with the augmented dataset comprising experimental, simulation, and generated samples. The fault diagnosis performance of the proposed method is evaluated on the experimental samples of two bearing datasets under small samples and various conditions of missing fault samples. It has been demonstrated by the experimental results that the proposed method exhibits robust fault diagnosis performance and generates high-quality samples under small samples and missing fault samples. Furthermore, the proposed method showcases its adaptability to different operation speeds.

Are aortic biomechanical properties early markers of dilatation in patients with Marfan syndrome? A systematic review and meta-analysis.

Although tissue stiffness is known to play an important role in aortic dilatation, the current guidelines for offering preventative surgery in patients with Marfan syndrome rely solely on the aortic diameter. In this systematic review and meta-analysis, we analyze and compare literature on in vivo aortic stiffness measures in Marfan patients. Our aim is to assess the potential of these measurements as early indicators of aortic dilatation. Following the PRISMA guidelines, we collected literature on diameter and three in vivo stiffness measures: Pulse wave velocity (PWV), -stiffness index (SI) and distensibility, at five different aortic locations in patients with Marfan syndrome. Results were reviewed and compared against each other. For meta-analysis, an augmented dataset was created by combining data from the literature. Regression with respect to age and statistical comparisons were performed. Thirty articles reporting data from 1925 patients with Marfan and 836 patients without Marfan were reviewed. PWV was found to be higher in Marfan, but only in dilated aortas. Distensibility was found to be lower even in non-dilated aortas, and its decrease was associated with higher chances of developing aortic dilatation. -SI was higher in Marfan patients and was positively correlated with the rate of aortic dilatation, emphasizing its role as a valuable indicator. In our meta-analysis, all stiffness measures showed a significant variation with age. Distensibility and -stiffness index were different in Marfan patients at all locations, and the difference was more pronounced after accounting for age-related variation. From the literature, -SI and distensibility emerge as the best predictors of future aortic dilatation. Our meta-analysis quantifies age-related changes in aortic stiffness and highlights the importance of accounting for age in comparing these measurements. Missing diameter values in the literature limited our analysis. Further investigation of criteria combining stiffness and diameter is recommended to better assist clinical decisions for prophylactic surgery.

Short-Term Photovoltaic Power Prediction Based on a Digital Twin Model

Due to the influence of meteorological conditions, shipboard photovoltaic (PV) systems have problems such as large fluctuation and inaccurate prediction of the output power. In this paper, a short-term PV power prediction method based on a novel digital twin (DT) model and BiLSTM is proposed. Firstly, a PV mechanism model and a data-driven model were established, in which the data-driven model was updated iteratively in real time using the sliding time window update method; then, these two models were converged to construct a PV DT model according to the DS evidence theory. Secondly, a BiLSTM model was built to make short-term predictions of the PV power using the augmented dataset of the DT model as an input. Finally, the method was tested and verified by experiments and further compared with main PV prediction methods. The research results indicate the following: firstly, the absolute error of the DT model was smaller than that of the mechanism model and the data-driven model, being as low as 5.62 W after the data update of the data-driven model; thus, the DT model realized data augmentation and high fidelity. Secondly, compared to several main PV prediction models, the PV DT model combined with BiLSTM had the lowest RMSE, MAE, and MAPE; the best followability; and the smallest absolute error under different weather conditions, which was especially obvious under cloudy weather conditions. In summary, the method can accurately predict the shipboard PV power, which has great theoretical significance and application value for improving the economy and reliability of solar ship operation.

Automated Classification of Collateral Circulation for Ischemic Stroke in Cone-Beam CT Images Using VGG11: A Deep Learning Approach

Background: Ischemic stroke poses significant challenges in diagnosis and treatment, necessitating efficient and accurate methods for assessing collateral circulation, a critical determinant of patient prognosis. Manual classification of collateral circulation in ischemic stroke using traditional imaging techniques is labor-intensive and prone to subjectivity. This study presented the automated classification of collateral circulation patterns in cone-beam CT (CBCT) images, utilizing the VGG11 architecture. Methods: The study utilized a dataset of CBCT images from ischemic stroke patients, accurately labeled with their respective collateral circulation status. To ensure uniformity and comparability, image normalization was executed during the preprocessing phase to standardize pixel values to a consistent scale or range. Then, the VGG11 model is trained using an augmented dataset and classifies collateral circulation patterns. Results: Performance evaluation of the proposed approach demonstrates promising results, with the model achieving an accuracy of 58.32%, a sensitivity of 75.50%, a specificity of 44.10%, a precision of 52.70%, and an F1 score of 62.10% in classifying collateral circulation patterns. Conclusions: This approach automates classification, potentially reducing diagnostic delays and improving patient outcomes. It also lays the groundwork for future research in using deep learning for better stroke diagnosis and management. This study is a significant advancement toward developing practical tools to assist doctors in making informed decisions for ischemic stroke patients.

Application of machine learning for film thickness prediction in elliptical EHL contact with varying entrainment angle

This contribution demonstrates the potential of machine learning (ML) algorithms in predicting elastohydrodynamic lubrication (EHL) film thickness in elliptical contact with varying direction of lubricant entrainment, ranging from wide to slender elliptical configurations. The input parameters pertain to worm gear contacts, which are characterized by slender-like elliptical contact between a steel and a soft metal component. The study encompasses generating a database using numerical Finite Element Method (FEM) simulations, training artificial neural network (ANN) models, and evaluating their performance in terms of bias and variance. Key outcomes include the successful training of the ANN models, detailed analysis of the impact of tailored architecture on the ANN models' performance, and the superiority of the ANN compared to other ML regression algorithms. The study further identifies key input parameters that influence prediction accuracy and introduces a strategic dataset augmentation procedure to increase local and overall prediction accuracy. This strategic dataset augmentation enhances model robustness and precision while providing insights for expanding databases collaboratively. It holds potential for broader applications of ML for performance prediction of tribological contacts, thus paving the way for advanced ML models that consider additional factors and collaborative databases refined by multiple research groups.

Addressing data scarcity using audio signal augmentation and deep learning for bolt looseness prediction

Deep learning models such as convolutional neural networks (CNNs) encounter challenges, including instability and overfitting, while predicting bolt looseness in data-scarce scenarios. In this study, we proposed a novel audio signal augmentation approach to classify bolt looseness in the event of data deficiency using CNN models. Audio signals at varied bolt torque conditions were extracted using the percussion method. Audio signal augmentation was performed using signal shifting and scaling strategies after segmenting the extracted audio signals. The unaugmented and augmented audio signals were transformed into scalograms using the continuous wavelet transform approach to train the CNN models. Upon training with augmented datasets, a promising improvement in the loss and accuracy of the CNN models in recognizing bolt looseness was noticed. One of the significant observations from the current study is that the implementation of audio signal augmentation improved the extrinsic generalization ability of the CNN models to classify bolt looseness. A maximum increase of 73.5% to identify bolt looseness in novel data was exhibited as compared to without augmentation. Overall, a maximum accuracy of 94.5% to classify bolt looseness in unseen data was demonstrated upon audio signal augmentation. In summary, the results affirm that the audio signal augmentation approach empowered the CNN models to predict bolt looseness in data-deficient scenarios accurately.

P-237 Fertilization prediction of mature oocyte based on microscopic image of first polar body morphology using YOLOv8

Abstract Study question Can a machine learning-based model predict successful fertilization of oocyte using first polar body morphology? Summary answer With small imbalance oocyte dataset, the most successful model to predict fertilization of mature oocyte is YOLOv8lcls with 97.47% accuracy, 100.00% specificity, and 94.87% sensitivity. What is known already The development of oocytes is crucial in the development and fertility of embryos in In Vitro Fertilization (IVF). There are several internal and external factors that could affect the fertilization competency of oocytes, one of which is the first polar body (PBI). The importance of the quality of PBI in ICSI has long been studied and has been evaluated by scoring systems regarding Istanbul Consensus. However, there is still a debate regarding the influence of PBI on the developmental competency of oocytes. Study design, size, duration This study involved the analysis of retrospective data including the microscopic light image of meiosis II (MII) oocytes at the time of Intracytoplasmic sperm injection (ICSI), along with polar body scoring by embryologists and the fertilization outcome of 1006 MII oocytes from a single public IVF clinic, between 2020-2022. After ICSI, fertilized oocytes were classified into 2 groups: 2PN and non-2PN (e.g., 0PN, 1PN, > =3PN). Participants/materials, setting, methods Light microscopic images of oocytes in the metaphase of meiosis II (MII) stage are captured, followed by the prediction of the likelihood of 2PN development after fertilization through intra-cytoplasmic sperm injection (ICSI). The image has its PBI segmented by the CNN and then cropped out of the original image. The PBI image is then fed into the prediction CNN to obtain the final result. The model performance was determined by evaluating accuracy, sensitivity, and specificity. Main results and the role of chance There were 1006 images of MII oocytes included in this study. After ICSI, 860 oocytes were fertilized resulting in the 2PN stage embryos and were classified as 2PN, whereas 146 oocytes were not fertilized or abnormally fertilized and were classified as non-2PN. Among these images, 782 oocytes received a PBI score of A, 677 of which are 2PN and 105 are non-2PN. The B grade consists of 224 images, of which are 183 2PN, and 41 non-2PN. The dataset was split into 3 subsets for training (70%), validation (10%), and testing (20%). The images are augmented by horizontal flipping to increase the amount of data and allow the model to observe several positions of the PBI. The YOLOv8 segmentation model is applied to segment the PBI and locate its position. The image is then rotated to move the PBI to the top and the PBI is cropped based on its position. Lastly, the cropped PBI image is used to predict the development by using the YOLOv8 Classify model trained on an augmented dataset. The YOLOv8l-cls model achieves the highest accuracy at 97.47%, specificity at 100.00%, and sensitivity at 94.87%. Limitations, reasons for caution The model does not include information on sperm quality, which also essentially contributes to the fertilization process. 2PN stage prediction is just the initial and surrogate outcome in IVF treatment. The model’s interpretation of the PBI score remains inconclusive. Therefore, additional experiments using dataset with balanced PBI grading are recommended. Wider implications of the findings The developed model is the first step in paving the way to establish and implement tools for assisting embryologists in selecting and prioritizing the high-potential oocyte for fertilization. Additional factors such as patient characteristics, other features of oocytes, and male factors should be further integrated to improve model performance. Trial registration number NA

Inventory Prediction Using a Modified Multi-Dimensional Collaborative Wrapped Bi-Directional Long Short-Term Memory Model

Inventory prediction is concerned with the forecasting of future demand for products in order to optimize inventory levels and supply chain management. The challenges include demand volatility, data quality, multi-dimensional interactions, lead time variability, seasonal trends, and dynamic pricing. Nevertheless, these models suffer from numerous shortcomings, and in this research, we propose a new model, MMCW-BiLSTM (modified multi-dimensional collaboratively wrapped BiLSTM), for inventory prediction. The MMCW-BiLSTM model reflects a considerable leap in inventory forecasting by combining a number of components in order to consider intricate temporal dependencies and incorporate feature interactions. The MMCW-BiLSTM makes use of BiLSTM layers, collaborative attention mechanisms, and a multi-dimensional attention approach to learn from augmented datasets consisting of the original features and the extracted time series data. Moreover, adding a Taylor series transformation allows for a more precise description of the features in the model, thus improving the prediction precision. The results show that the models make the least mistakes when they use the AV demand forecasting dataset, with MAE values of 1.75, MAPE values of 2.89, MSE values of 6.76, and RMSE values of 2.6. Similarly, when utilizing the product demand dataset, the model also achieves the lowest error values for these metrics at 1.97, 3.91, 8.76, and 2.96. Likewise, when utilizing the dairy goods sales dataset, the model also achieves the lowest error values for these metrics at 2.54, 3.69, 10.39, and 3.22.

Brain tumor classification in VIT-B/16 based on relative position encoding and residual MLP.

Brain tumors pose a significant threat to health, and their early detection and classification are crucial. Currently, the diagnosis heavily relies on pathologists conducting time-consuming morphological examinations of brain images, leading to subjective outcomes and potential misdiagnoses. In response to these challenges, this study proposes an improved Vision Transformer-based algorithm for human brain tumor classification. To overcome the limitations of small existing datasets, Homomorphic Filtering, Channels Contrast Limited Adaptive Histogram Equalization, and Unsharp Masking techniques are applied to enrich dataset images, enhancing information and improving model generalization. Addressing the limitation of the Vision Transformer's self-attention structure in capturing input token sequences, a novel relative position encoding method is employed to enhance the overall predictive capabilities of the model. Furthermore, the introduction of residual structures in the Multi-Layer Perceptron tackles convergence degradation during training, leading to faster convergence and enhanced algorithm accuracy. Finally, this study comprehensively analyzes the network model's performance on validation sets in terms of accuracy, precision, and recall. Experimental results demonstrate that the proposed model achieves a classification accuracy of 91.36% on an augmented open-source brain tumor dataset, surpassing the original VIT-B/16 accuracy by 5.54%. This validates the effectiveness of the proposed approach in brain tumor classification, offering potential reference for clinical diagnoses by medical practitioners.

Development of an early detection and automatic targeting system for cotton weeds using an improved lightweight YOLOv8 architecture on an edge device

Traditional means of weed removal, such as human work or the use of pesticides, frequently require significant amounts of effort, incur high expenses, and can negatively impact the environment. This study introduces a modified version of the YOLOv8 nano architecture that is suitable for running on edge devices for real-time applications. The proposed model uses an augmented version of the well-known CottonWeedDet12 dataset consisting of a total of 16,944 images with characteristic annotations to develop a model capable of correctly distinguishing 12 different cotton weed classes with an increased mean average precision of 97.6 % that is about 1.2 % more than the model trained using original, unaugmented dataset. The final selected model uses a convolutional block attention module (CBAM) and a unique C3Ghost block within the YOLOv8 backbone, which together increase the model's reliability for more accurate predictions with reduced computational complexity. Upon training with the augmented dataset, the proposed model with only 3.6 million parameters was able to achieve an mAP@50 score of 97.6 %, which surpasses all previous studies conducted using this dataset. Additionally, a high F1 score of 94.4 % proves that the model has a good balance between recall and precision. Class Activation Map (CAM) approaches such as EigenCAM, Grad-CAM++, and LayerCAM explainable AI (XAI) showed promising results for each of the customized models upon testing their interpretability for cotton weed detection. Furthermore, based on this model, a fast and cost-efficient targeting system was developed using a yaw-pitch mechanism for automatic weed tracking and herbicide spraying.

Prediction method of human defecation based on informer audio data augmentation and improved residual network

Defecation care for disabled patients is a major challenge in health management. Traditional post-defecation treatment will bring physical injury and negative emotions to patients, while existing pre-defecation forecasting care methods are physically intrusive. On the basis of exploring the mechanism of defecation intention generation, and based on the characteristic analysis and clinical application of bowel sounds, it is found that the generation of desire to defecate and bowel sounds are correlated to a certain extent. Therefore, a deep learning-based bowel sound recognition method is proposed for human defecation prediction. The wavelet domain based Wiener filter is used to filter the bowel sound data to reduce other noise. Statistical analysis, fast Fourier transform and wavelet packet transform are used to extract the integrated features of bowel sound in time, frequency and time-frequency domain. In particular, an audio signal expansion data algorithm based on the Informer model is proposed to solve the problem of poor generalization of the training model caused by the difficulty of collecting bowel sound in reality. An improved one-dimensional residual network model (1D-IResNet) for defecation classification prediction is designed based on multi-domain features. The experimental results show that the proposed bowel sound augmentation strategy can effectively improve the data sample size and increase the sample diversity. Under the augmented dataset, the training speed of the 1D-IResNet model is accelerated, and the classification accuracy reaches 90.54 %, the F1 score reaches 83.88 %, which achieves a relatively good classification stability while maintaining a high classification index.

Ensemble Learning with Pre-Trained Transformers for Crash Severity Classification: A Deep NLP Approach

Transfer learning has gained significant traction in natural language processing due to the emergence of state-of-the-art pre-trained language models (PLMs). Unlike traditional word embedding methods such as TF-IDF and Word2Vec, PLMs are context-dependent and outperform conventional techniques when fine-tuned for specific tasks. This paper proposes an innovative hard voting classifier to enhance crash severity classification by combining machine learning and deep learning models with various word embedding techniques, including BERT, RoBERTa, Word2Vec, and TF-IDF. Our study involves two comprehensive experiments using motorists’ crash data from the Missouri State Highway Patrol. The first experiment evaluates the performance of three machine learning models—XGBoost (XGB), random forest (RF), and naive Bayes (NB)—paired with TF-IDF, Word2Vec, and BERT feature extraction techniques. Additionally, BERT and RoBERTa are fine-tuned with a Bidirectional Long Short-Term Memory (Bi-LSTM) classification model. All models are initially evaluated on the original dataset. The second experiment repeats the evaluation using an augmented dataset to address the severe data imbalance. The results from the original dataset show strong performance for all models in the “Fatal” and “Personal Injury” classes but a poor classification of the minority “Property Damage” class. In the augmented dataset, while the models continued to excel with the majority classes, only XGB/TFIDF and BERT-LSTM showed improved performance for the minority class. The ensemble model outperformed individual models in both datasets, achieving an F1 score of 99% for “Fatal” and “Personal Injury” and 62% for “Property Damage” on the augmented dataset. These findings suggest that ensemble models, combined with data augmentation, are highly effective for crash severity classification and potentially other textual classification tasks.

DETECTING ANOMALIES IN VIDEOS USING MEMORY-AUGMENTED AUTOENCODER WITH KEY FRAME SELECTION

In this article, we propose a novel method to train a memory-augmented autoencoder in supervised mode by generating pseudo abnormal videos based on key frame selection techniques. Most video anomaly detection methods employ a machine learning model to learn patterns of normal videos. Any video where the patterns significantly deviate from the learnt patterns is considered an anomaly. However, developing an effective machine learning model for video anomaly detection is a challenging task due to the deficiency of anomalies. Specifically, abnormal samples are often much rarer and harder to collect than normal samples. To address this problem, we propose a novel approach using key frame selection techniques to generate pseudo anomalies. The generated pseudo anomalies are then combined with normal data to create the augmented dataset. The memory-augmented autoencoder is then trained on the augmented datasets. The experimental results show that the AUC scores of the proposed solution are higher than those of the base network architecture from 0.20% to 1.31% on three well-known datasets for video anomaly detection.

Synthetic Genitourinary Image Synthesis via Generative Adversarial Networks: Enhancing Artificial Intelligence Diagnostic Precision.

In the realm of computational pathology, the scarcity and restricted diversity of genitourinary (GU) tissue datasets pose significant challenges for training robust diagnostic models. This study explores the potential of Generative Adversarial Networks (GANs) to mitigate these limitations by generating high-quality synthetic images of rare or underrepresented GU tissues. We hypothesized that augmenting the training data of computational pathology models with these GAN-generated images, validated through pathologist evaluation and quantitative similarity measures, would significantly enhance model performance in tasks such as tissue classification, segmentation, and disease detection. To test this hypothesis, we employed a GAN model to produce synthetic images of eight different GU tissues. The quality of these images was rigorously assessed using a Relative Inception Score (RIS) of 1.27 ± 0.15 and a Fréchet Inception Distance (FID) that stabilized at 120, metrics that reflect the visual and statistical fidelity of the generated images to real histopathological images. Additionally, the synthetic images received an 80% approval rating from board-certified pathologists, further validating their realism and diagnostic utility. We used an alternative Spatial Heterogeneous Recurrence Quantification Analysis (SHRQA) to assess the quality of prostate tissue. This allowed us to make a comparison between original and synthetic data in the context of features, which were further validated by the pathologist's evaluation. Future work will focus on implementing a deep learning model to evaluate the performance of the augmented datasets in tasks such as tissue classification, segmentation, and disease detection. This will provide a more comprehensive understanding of the utility of GAN-generated synthetic images in enhancing computational pathology workflows. This study not only confirms the feasibility of using GANs for data augmentation in medical image analysis but also highlights the critical role of synthetic data in addressing the challenges of dataset scarcity and imbalance. Future work will focus on refining the generative models to produce even more diverse and complex tissue representations, potentially transforming the landscape of medical diagnostics with AI-driven solutions.

Causal prior-embedded physics-informed neural networks and a case study on metformin transport in porous media

This study introduces a novel approach to transport modelling by integrating experimentally derived causal priors into neural networks. We illustrate this paradigm using a case study of metformin, a ubiquitous pharmaceutical emerging pollutant, and its transport behaviour in sandy media. Specifically, data from metformin's sandy column transport experiment was used to estimate unobservable parameters through a physics-based model Hydrus-1D, followed by a data augmentation to produce a more comprehensive dataset. A causal graph incorporating key variables was constructed, aiding in identifying impactful variables and estimating their causal dynamics or “causal prior.” The causal priors extracted from the augmented dataset included underexplored system parameters such as the type-1 sorption fraction F, first-order reaction rate coefficient α, and transport system scale. Their moderate impact on the transport process has been quantitatively evaluated (normalized causal effect 0.0423, -0.1447 and -0.0351, respectively) with adequate confounders considered for the first time. The prior was later embedded into multilayer neural networks via two methods: causal weight initialization and causal prior regularization. Based on the results from AutoML hyperparameter tuning experiments, using two embedding methods simultaneously emerged as a more advantageous practice since our proposed causal weight initialization technique can enhance model stability, particularly when used in conjunction with causal prior regularization. amongst those experiments utilizing both techniques, the R-squared values peaked at 0.881. This study demonstrates a balanced approach between expert knowledge and data-driven methods, providing enhanced interpretability in black-box models such as neural networks for environmental modelling.

Bluish veil detection and lesion classification using custom deep learnable layers with explainable artificial intelligence (XAI)

Melanoma, one of the deadliest types of skin cancer, accounts for thousands of fatalities globally. The bluish, blue-whitish, or blue-white veil (BWV) is a critical feature for diagnosing melanoma, yet research into detecting BWV in dermatological images is limited. This study utilizes a non-annotated skin lesion dataset, which is converted into an annotated dataset using a proposed imaging algorithm (color threshold techniques) on lesion patches based on color palettes. A Deep Convolutional Neural Network (DCNN) is designed and trained separately on three individual and combined dermoscopic datasets, using custom layers instead of standard activation function layers. The model is developed to categorize skin lesions based on the presence of BWV. The proposed DCNN demonstrates superior performance compared to the conventional BWV detection models across different datasets. The model achieves a testing accuracy of 85.71 % on the augmented PH2 dataset, 95.00 % on the augmented ISIC archive dataset, 95.05 % on the combined augmented (PH2+ISIC archive) dataset, and 90.00 % on the Derm7pt dataset. An explainable artificial intelligence (XAI) algorithm is subsequently applied to interpret the DCNN's decision-making process about the BWV detection. The proposed approach, coupled with XAI, significantly improves the detection of BWV in skin lesions, outperforming existing models and providing a robust tool for early melanoma diagnosis.

Improving skin cancer detection by Raman spectroscopy using convolutional neural networks and data augmentation.

Our previous studies have demonstrated that Raman spectroscopy could be used for skin cancer detection with good sensitivity and specificity. The objective of this study is to determine if skin cancer detection can be further improved by combining deep neural networks and Raman spectroscopy. Raman spectra of 731 skin lesions were included in this study, containing 340 cancerous and precancerous lesions (melanoma, basal cell carcinoma, squamous cell carcinoma and actinic keratosis) and 391 benign lesions (melanocytic nevus and seborrheic keratosis). One-dimensional convolutional neural networks (1D-CNN) were developed for Raman spectral classification. The stratified samples were divided randomly into training (70%), validation (10%) and test set (20%), and were repeated 56 times using parallel computing. Different data augmentation strategies were implemented for the training dataset, including added random noise, spectral shift, spectral combination and artificially synthesized Raman spectra using one-dimensional generative adversarial networks (1D-GAN). The area under the receiver operating characteristic curve (ROC AUC) was used as a measure of the diagnostic performance. Conventional machine learning approaches, including partial least squares for discriminant analysis (PLS-DA), principal component and linear discriminant analysis (PC-LDA), support vector machine (SVM), and logistic regression (LR) were evaluated for comparison with the same data splitting scheme as the 1D-CNN. The ROC AUC of the test dataset based on the original training spectra were 0.886±0.022 (1D-CNN), 0.870±0.028 (PLS-DA), 0.875±0.033 (PC-LDA), 0.864±0.027 (SVM), and 0.525±0.045 (LR), which were improved to 0.909±0.021 (1D-CNN), 0.899±0.022 (PLS-DA), 0.895±0.022 (PC-LDA), 0.901±0.020 (SVM), and 0.897±0.021 (LR) respectively after augmentation of the training dataset (p<0.0001, Wilcoxon test). Paired analyses of 1D-CNN with conventional machine learning approaches showed that 1D-CNN had a 1-3% improvement (p<0.001, Wilcoxon test). Data augmentation not only improved the performance of both deep neural networks and conventional machine learning techniques by 2-4%, but also improved the performance of the models on spectra with higher noise or spectral shifting. Convolutional neural networks slightly outperformed conventional machine learning approaches for skin cancer detection by Raman spectroscopy.

Maturity grading of jujube for industrial applications harnessing deep learning

PurposeTo develop and examine an efficient and reliable jujube grading model with reduced computational time, which could be utilized in the food processing and packaging industries to perform quick grading and pricing of jujube as well as for the other similar types of fruits.Design/methodology/approachThe whole process begins with manual analysis and collection of four jujube grades from the jujube tree, in addition to this jujube image acquisition was performed utilizing MVS which is further followed by image pre-processing and augmentation tasks. Eventually, classification models (i.e. proposed model, from scratch and pre-trained VGG16 and AlexNet) were trained and validated over the original and augmented datasets to discriminate the jujube into maturity grades.FindingsThe highest success rates reported over the original and augmented datasets were 97.53% (i.e. error of 2.47%) and 99.44% (i.e. error of 0.56%) respectively using Adam optimizer and a learning rate of 0.003.Research limitations/implicationsThe investigation relies upon a single view of the jujube image and the outer appearance of the jujube. In the future, multi-view image capturing system could be employed for the model training/validation.Practical implicationsDue to the vast functional derivatives of jujube, the identification of maturity grades of jujube is paramount in the fruit industry, functional food production industries and pharmaceutical industry. Therefore, the proposed model which is practically feasible and easy to implement could be utilized in such industries.Originality/valueThis research examines the performance of proposed CNN models for selected optimizer and learning rates for the grading of jujube maturity into four classes and compares them with the classical models to depict the sublime model in terms of accuracy, the number of parameters, epochs and computational time. After a thorough investigation of the models, it was discovered that the proposed model transcends both classical models in all aspects for both the original and augmented datasets utilizing Adam optimizer with learning rate of 0.003.