Building Better Deep Learning Models Through Dataset Fusion: A Case Study in Skin Cancer Classification with Hyperdatasets.

In the last few years, Deep Learning (DL) has been showing superior performance in different modalities of bio-medical image analysis. Several DL architectures have been proposed for classification, segmentation, and detection tasks in medical imaging and computational pathology. In this paper, we propose a new DL architecture, the NABLA-N network (∇N-Net), with better feature fusion techniques in decoding units for dermoscopic image segmentation tasks. The ∇N-Net has several advances for segmentation tasks. First, this model ensures better feature representation for semantic segmentation with a combination of low to high-level feature maps. Second, this network shows better quantitative and qualitative results with the same or fewer network parameters compared to other methods. In addition, the Inception Recurrent Residual Convolutional Neural Network (IRRCNN) model is used for skin cancer classification. The proposed ∇N-Net network and IRRCNN models are evaluated for skin cancer segmentation and classification on the benchmark datasets from the International Skin Imaging Collaboration 2018 (ISIC-2018). The experimental results show superior performance on segmentation tasks compared to the Recurrent Residual U-Net (R2U-Net). The classification model shows around 87% testing accuracy for dermoscopic skin cancer classification on ISIC2018 dataset.

Skin cancer ranks among the most commonly diagnosed forms of cancer worldwide, with millions of new cases identified annually. Over 2 million cases of non-melanoma skin cancer and approximately 132,000 melanoma cases are reported each year . The study aims is to develop an efficient deep learning model using CNN’s architectures to accurately classify skin lesions as malignant or benign based on dermoscopic images. This study explored the use of Convolutional Neural Network (CNN) models using AlexNet and ResNet architectures for skin cancer classification. The performance of both models were evaluated based on key metrics such as accuracy, precision, recall, and F1-score. AlexNet was trained on 1,200 skin lesion images and validated on 600 images,training spanned 10 epochs. ResNet was similarly trained and validated after 10 epochs. Both architectures were compared under identical conditions (same dataset, pre-processing, optimizer, and hyperparameters). ResNet outperformed AlexNet in all performance metrics of accuracy as 91.00% and 59.00% respectively. This research explored the performance of CNN(ResNet50) and CNN(AlexNet). This validates that deeper architectures with residual connections, batch normalization, and optimized feature extraction are more effective for binary skin cancer classification.

Skin cancer diagnosis challenges dermatologists due to its complex visual variations across diagnostic categories. Convolutional neural networks (CNNs), specifically the Efficient Net B0-B7 series, have shown superiority in multiclass skin cancer classification. This study addresses the limitations of visual examination by presenting a tailored preprocessing pipeline designed for Efficient Net models. Leveraging transfer learning with pre-trained ImageNet weights, the research aims to enhance diagnostic accuracy in an imbalanced multiclass classification context. The study develops a specialized image preprocessing pipeline involving image scaling, dataset augmentation, and artifact removal tailored to the nuances of Efficient Net models. Using the Efficient Net B0-B7 dataset, transfer learning fine-tunes CNNs with pre-trained ImageNet weights. Rigorous evaluation employs key metrics like Precision, Recall, Accuracy, F1 Score, and Confusion Matrices to assess the impact of transfer learning and fine-tuning on each Efficient Net variant's performance in classifying diverse skin cancer categories. The research showcases the effectiveness of the tailored preprocessing pipeline for Efficient Net models. Transfer learning and fine-tuning significantly enhance the models' ability to discern diverse skin cancer categories. The evaluation of eight Efficient Net models (B0-B7) for skin cancer classification reveals distinct performance patterns across various cancer classes. While the majority class, Benign Kertosis, achieves high accuracy (>87%), challenges arise in accurately classifying Eczema classes. Melanoma, despite its minority representation (2.42% of images), attains an average accuracy of 80.51% across all models. However, suboptimal performance is observed in predicting warts molluscum (90.7%) and psoriasis (84.2%) instances, highlighting the need for targeted improvements in accurately identifying specific skin cancer types. The study on skin cancer classification utilizes EfficientNets B0-B7 with transfer learning from ImageNet weights. The pinnacle performance is observed with EfficientNet-B7, achieving a groundbreaking top-1 accuracy of 84.4% and top-5 accuracy of 97.1%. Remarkably efficient, it is 8.4 times smaller than the leading CNN. Detailed per-class classification exactitudes through Confusion Matrices affirm its proficiency, signaling the potential of EfficientNets for precise dermatological image analysis.

Skin cancer is one of the most common cancers, and its early detection can have a huge impact on its outcomes. Deep learning, especially convolutional neural networks, perform well in processing massive amounts of data, especially image data in classifying skin cancer. In this paper, convolutional neural networks are mainly used to diagnose and classify 7 types of skin lesions, including melanoma, basal cell carcinoma, melanocytic nevus, actinic keratosis, and intraepithelial carcinoma, benign keratinoid lesions, dermatofibroma, and vascular lesions. First, the characteristics of skin lesion images are analyzed, using image processing technology and sampling technology to preprocess skin lesion images. Then the training parameters of imageNet network are adjusted through the idea of transfer learning on InceptionV3, ResNet50, DenseNet201, and other networks to perform training classification. Furthermore, different convolutional neural network models are built for classification. In order to validate the classification performance of various convolutional neural network models, this paper adopts ISIC 2017 HAM10000 dataset for experiments. The experimental results show that proper preprocessing is necessary when applying CNN for image classification. In classifying the 224*224 skin lesion images, the classical deep convolutional network with DenseNet201 achieved a remarkable performance classification accuracy of 99.12% for training and 86.91% for testing.

Among the many deep learning methods, the convolutional neural network (CNN) model has an excellent performance in image recognition. Research on identifying and classifying image datasets using CNN is ongoing. Animal species recognition and classification with CNN is expected to be helpful for various applications. However, sophisticated feature recognition is essential to classify quasi-species with similar features, such as the quasi-species of parrots that have a high color similarity. The purpose of this study is to develop a vision-based mobile application to classify endangered parrot species using an advanced CNN model based on transfer learning (some parrots have quite similar colors and shapes). We acquired the images in two ways: collecting them directly from the Seoul Grand Park Zoo and crawling them using the Google search. Subsequently, we have built advanced CNN models with transfer learning and trained them using the data. Next, we converted one of the fully trained models into a file for execution on mobile devices and created the Android package files. The accuracy was measured for each of the eight CNN models. The overall accuracy for the camera of the mobile device was 94.125%. For certain species, the accuracy of recognition was 100%, with the required time of only 455 ms. Our approach helps to recognize the species in real time using the camera of the mobile device. Applications will be helpful for the prevention of smuggling of endangered species in the customs clearance area.

BackgroundIt is difficult for orthodontists to accurately predict the growth trend of the mandible in children with anterior crossbite. This study aims to develop a deep learning model to automatically predict the mandibular growth result into normal or overdeveloped using cephalometric radiographs.MethodsA deep convolutional neural network (CNN) model was constructed based on the algorithm ResNet50 and trained on the basis of 256 cephalometric radiographs. The prediction behavior of the model was tested on 40 cephalograms and visualized by equipped with Grad-CAM. The prediction performance of the CNN model was compared with that of three junior orthodontists.ResultsThe deep-learning model showed a good prediction accuracy about 85%, much higher when compared with the 54.2% of the junior orthodontists. The sensitivity and specificity of the model was 0.95 and 0.75 respectively, higher than that of the junior orthodontists (0.62 and 0.47 respectively). The area under the curve value of the deep-learning model was 0.9775. Visual inspection showed that the model mainly focused on the characteristics of special regions including chin, lower edge of the mandible, incisor teeth, airway and condyle to conduct the prediction.ConclusionsThe deep-learning CNN model could predict the growth trend of the mandible in anterior crossbite children with relatively high accuracy using cephalometric images. The deep learning model made the prediction decision mainly by identifying the characteristics of the regions of chin, lower edge of the mandible, incisor teeth area, airway and condyle in cephalometric images.

Skin cancers are one of the most common cancers in the world. Early detections and treatments of skin cancers can greatly improve the survival rates of patients. In this paper, a skin lesions classification system is developed with deep convolutional neural networks of ResNet50, which may help dermatologists to recognize skin cancers earlier. We utilize the ResNet50 as a pre-trained model. Then, by transfer learning, it is trained on our skin lesions dataset. Image preprocessing and dataset balancing methods are used to increase the accuracy of the classification model. In classification of skin diseases, our model achieves an overall accuracy of 83.74% on nine-class skin lesions. The experimental results show an impressive effect of the ResNet50 model in finegrained skin lesions classification and skin cancers recognition.

Skin cancer diagnosis faces challenges in efficiency and accuracy. This research addresses the need for improved non-invasive diagnostic tools by leveraging deep learning for multi-class skin cancer classification from dermoscopic images. A key focus is overcoming the limitations of imbalanced datasets, common in medical imaging, which can hinder model performance. We propose an optimal strategy utilizing a Convolutional Neural Network (CNN) transfer learning methodology. The process involves CNN-based segmentation to isolate relevant regions, followed by feature extraction and classification. We comparatively evaluated three pre-trained transfer learning techniques: DenseNet201, ResNet50, and VGG16, using the HAM10000 dataset (10,015 images across seven skin cancer classes). To mitigate severe class imbalance, Random Oversampling was employed, chosen for its simplicity and effectiveness in balancing the dataset and enhancing model generalization. Model performance was rigorously evaluated using accuracy, precision, recall, and F1-score. DenseNet201 consistently achieved superior performance, with an accuracy of 97% post-oversampling. It also exhibited the highest precision, recall, and F1-score across all models, confirming its effectiveness in classifying both majority and minority classes. Compared to previous studies on HAM10000, our DenseNet201 model's test accuracy of 96.52% is competitive or superior to reported accuracy of 90-92%. This highlights the synergistic effect of DenseNet201's efficient feature reuse and robust data balancing. This research provides a robust framework for advanced methodologies in skin cancer classification, particularly for imbalanced medical image datasets.

People around the world suffer from a wide range of illnesses. The most dangerous of them all is cancer. Of all the cancers, skin cancer is the one that is spreading the fastest. The reason for this is because skin cells grow abnormally. Skin cancer is spreading around the world in part due to an increase in UV light on Earth's surface. The two most prevalent kinds of skin cancer are benign and malignant. Despite undergoing costly and time-consuming therapies, skin cancer mortality rates do not decrease. Early identification of skin cancer in its early stages is useful in lowering the death rate. Deep learning is being utilized in the modern world to identify illnesses. Convolutional neural networks (CNNs) improve the accuracy of image classification in the detection of skin cancer. This study compares the operational procedures of several CNN models in order to identify which produces the best outcomes. CNN models are analyzed using pre-trained models such as VGG16, Res-Net50, and Inception Resnet V2. Due to differences in their layer numbers, these models operate in distinct ways. Certain models function better than others, depending on their layers and how they operate. A Kaggle picture dataset containing both benign and cancerous data has been extracted. There are 3287 photos of benign and malignant skin cancer in this dataset. We have obtained accurate results for VGG16 (85%), ResNet50 (69%), and Inception Resnet V2 (83%), using various techniques. This study compares these results according to the model's operation. We compare precision, working process, and model layer numbers.

The uncertainty of wind power as the main obstacle of its integration into the power grid can be addressed by an accurate and efficient wind power forecast. Among the various wind power forecasting methods, machine learning (ML) algorithms, are recognized as a powerful wind power forecasting tool, however, their performance is highly dependent on the proper tuning of their hyperparameters. Common hyperparameter tuning methods such as grid search or random search are time-consuming, computationally expensive, and unreliable for complex models such as deep learning neural networks. Therefore, there is an urgent need for automatic methods to discover optimal hyperparameters for higher accuracy and efficiency of prediction models. In this study, a novel investigation is contributed to the field of wind power forecasting by a comprehensive comparison of three advanced techniques – Scikit-opt, Optuna, and Hyperopt – for hyperparameter optimization of Convolutional Neural Network (CNN) and Long Short-Term Memory Network (LSTM) models, a facet that, to our knowledge, has not been systematically explored in existing literature. The impact of these optimization techniques on the accuracy and efficiency of the CNN and LSTM models are assessed by comparing the root mean square error (RMSE) of the predictions and the required time to tune the models. The results show that the Optuna algorithm, using a Tree-structured Parzen Estimator (TPE) search method and Expected Improvement (EI) acquisition function, has the best efficiency for both CNN and LSTM models. In terms of accuracy, it is demonstrated that while for the CNN model all the optimization methods achieve similar performances, the LSTM model optimized by the Hyperopt algorithm, based on the annealing search method, results in the highest accuracy. In addition, for the first time in this research, the impact of the random initialization features on the performance of the forecasting models with neural networks is investigated. The proposed structures for deep learning models were examined to determine the most robust structure with the minimal sensitivity to the randomness. What we have discovered from the comparison of advanced hyperparameter optimization methods can be used by researchers to tune the time series-based forecasting models.

Skin cancer is one of the most dangerous types of cancer due to its immediate appearance and the possibility of rapid spread. It arises from uncontrollably growing cells, rapidly dividing cells in one area of the body, invading other bodily tissues, and spreading throughout the body. Early detection helps prevent cancer progress from reaching critical levels, reducing the risk of complications and the need for more aggressive treatment options. Convolutional neural networks (CNNs) revolutionize skin cancer diagnosis by extracting intricate features from images, enabling an accurate classification of lesions. Their role extends to early detection, providing a powerful tool for dermatologists to identify abnormalities in their nascent stages, ultimately improving patient outcomes. This study proposes a novel deep convolutional neural network (DCNN) approach to classifying skin cancer lesions. The proposed DCNN model is evaluated using two unbalanced datasets, namely HAM10000 and ISIC-2019. The DCNN model is compared with other transfer learning models, including VGG16, VGG19, DenseNet121, DenseNet201, and MobileNetV2. Its performance is assessed using four widely used evaluation metrics: accuracy, recall, precision, F1-score, specificity, and AUC. The experimental results demonstrate that the proposed DCNN model outperforms other deep learning (DL) models that utilized these datasets. The proposed DCNN model achieved the highest accuracy with the HAM10000 and ISIC-2019 datasets, reaching 98.5%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$98.5\\%$$\\end{document} and 97.1%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$97.1\\%$$\\end{document}, respectively. These experimental results show how competitive and successful the DCNN model is in overcoming the problems caused by class imbalance and raising skin cancer classification accuracy. Furthermore, the proposed model demonstrates superior performance, particularly excelling in terms of accuracy, compared to other recent studies that utilize the same datasets, which highlights the robustness and effectiveness of the proposed DCNN.

Recent advances in deep learning models have made them the state-of-the art method for image classification. Due to this success, they have been applied to many areas, such as satellite image processing, medical image interpretation, video processing, etc. Recently, deep learning models have been utilized for processing Ground Penetrating Radar (GPR) data as well. However, studies general focus on building new Convolutional Neural Network (CNN) models instead of utilizing baseline ones. This paper investigates the usefulness of existing baseline CNN models for classifying GPR B-scan images and aims to determine how well pre-trained models perform. To that end, a real bridge deck GPR data, DECKGPRHv1.0 dataset was used to evaluate the transfer learning performances of various CNN models. Different variants of the models in terms of varying depths and number of parameters were also considered and evaluated in a comparative manner. Although it is an older model, ResNet achieved the best results with 0.998 accuracy. The experimental results showed that there is generally a direct correlation between the simplicity of the model and its success. Overall, it is concluded that near perfect results are possible by just adapting pre-trained models to the problem without fine-tuning.

Emotions expressed by humans can be identified from facial expressions, speech signals, or physiological signals. Among them, the use of physiological signals for emotion classification is a notable emerging area of research. In emotion recognition, a person’s electrocardiogram (ECG) and galvanic skin response (GSR) signals cannot be manipulated, unlike facial and voice signals. Moreover, wearables such as smartwatches and wristbands enable the detection of emotions in people’s naturalistic environment. During the COVID-19 pandemic, it was necessary to detect people’s emotions in order to ensure that appropriate actions were taken according to the prevailing situation and achieve societal balance. Experimentally, the duration of the emotion stimulus period and the social and non-social contexts of participants influence the emotion classification process. Hence, classification of emotions when participants are exposed to the elicitation process for a longer duration and taking into consideration the social context needs to be explored. This work explores the classification of emotions using five pretrained convolutional neural network (CNN) models: MobileNet, NASNetMobile, DenseNet 201, InceptionResnetV2, and EfficientNetB7. The continuous wavelet transform (CWT) coefficients were detected from ECG and GSR recordings from the AMIGOS database with suitable filtering. Scalograms of the sum of frequency coefficients versus time were obtained and converted into images. Emotions were classified using the pre-trained CNN models. The valence and arousal emotion classification accuracy obtained using ECG and GSR data were, respectively, 91.27% and 91.45% using the InceptionResnetV2 CNN classifier and 99.19% and 98.39% using the MobileNet CNN classifier. Other studies have not explored the use of scalograms to represent ECG and GSR CWT features for emotion classification using deep learning models. Additionally, this study provides a novel classification of emotions built on individual and group settings using ECG data. When the participants watched long-duration emotion elicitation videos individually and in groups, the accuracy was around 99.8%. MobileNet had the highest accuracy and shortest execution time. These subject-independent classification methods enable emotion classification independent of varying human behavior.

Deep learning methods have been developed and widely used in land use classification with remote sensing images. In addition, due to the different datasets used in different studies, there is a lack of direct comparison between different deep learning model applications in land use classification. The open source dataset DeepSat was used to build and test a convolutional neural network (CNN) model. The convolution kernels in the model were extracted to further study the specific features learned by the deep learning model. In addition, different CNN-based models were compared to explore the impacts of model structures on model accuracy. The major conclusions from the research are: (1) CNN model is effective in land use classification, with an accuracy of 0.9998 and 0.9991 for the SAT-4 and SAT-6 data, respectively; (2) CNN does have a “learning” ability that can extract the most critical and effective information from training datasets; and (3) for remote sensing land use classification, increasing the number of convolution kernels is better than adding more convolutional layers. Max pooling is better than average pooling. In addition, a localized response normalized layer can also improve model accuracy.

Detecting clinical keratoconus (KCN) poses a challenging and time-consuming task. During the diagnostic process, ophthalmologists are required to review demographic and clinical ophthalmic examinations in order to make an accurate diagnosis. This study aims to develop and evaluate the accuracy of deep convolutional neural network (CNN) models for the detection of keratoconus (KCN) using corneal topographic maps. We retrospectively collected 1758 corneal images (978 normal and 780 keratoconus) from 1010 subjects of the KCN group with clinically evident keratoconus and the normal group with regular astigmatism. To expand the dataset, we developed a model using Variational Auto Encoder (VAE) to generate and augment images, resulting in a dataset of 4000 samples. Four deep learning models were used to extract and identify deep corneal features of original and synthesized images. We demonstrated that the utilization of synthesized images during training process increased classification performance. The overall average accuracy of the deep learning models ranged from 99% for VGG16 to 95% for EfficientNet-B0. All CNN models exhibited sensitivity and specificity above 0.94, with the VGG16 model achieving an AUC of 0.99. The customized CNN model achieved satisfactory results with an accuracy and AUC of 0.97 at a much faster processing speed compared to other models. In conclusion, the DL models showed high accuracy in screening for keratoconus based on corneal topography images. This is a development toward the potential clinical implementation of a more enhanced computer-aided diagnosis (CAD) system for KCN detection, which would aid ophthalmologists in validating the clinical decision and carrying out prompt and precise KCN treatment.