SADASNet: A Selective and Adaptive Deep Architecture Search Network with Hyperparameter Optimization for Robust Skin Cancer Classification

Diagnostics, Год журнала: 2025, Номер 15(5), С. 541 - 541

Опубликована: Фев. 24, 2025

Background/Objectives: Skin cancer is a major public health concern, where early diagnosis and effective treatment are essential for prevention. To enhance diagnostic accuracy, researchers have increasingly utilized computer vision systems, with deep learning-based approaches becoming the primary focus in recent studies. Nevertheless, there notable research gap optimization of hyperparameters to design optimal learning architectures, given need high accuracy lower computational complexity. Methods: This paper puts forth robust metaheuristic optimization-based approach develop novel architectures multi-class skin classification. method, designated as SADASNet (Selective Adaptive Deep Architecture Search Network by Hyperparameter Optimization) algorithm, developed based on Particle Swarm Optimization (PSO) technique. The method adapted HAM10000 dataset. Innovative data augmentation techniques applied overcome class imbalance issues performance model. has been accommodate range image sizes, six different original models produced result. Results: achieved following highest metrics: 99.31% 97.58% F1 score, 97.57% recall, 97.64% precision, 99.59% specificity. Compared most advanced competitors reported literature, proposed demonstrates superior terms Furthermore, it maintains broad solution space during parameter optimization. Conclusions: With these outcomes, this aims classification contribute advancement learning.

Язык: Английский

---

WCFormer: An interpretable deep learning framework for heart sound signal analysis and automated diagnosis of cardiovascular diseases

Expert Systems with Applications, Год журнала: 2025, Номер unknown, С. 127238 - 127238

Опубликована: Март 1, 2025

Язык: Английский

---

Melanoma Skin Cancer Recognition with a Convolutional Neural Network and Feature Dimensions Reduction with Aquila Optimizer

Diagnostics, Год журнала: 2025, Номер 15(6), С. 761 - 761

Опубликована: Март 18, 2025

Background: Melanoma is a highly aggressive form of skin cancer, necessitating early and accurate detection for effective treatment. This study aims to develop novel classification system melanoma that integrates Convolutional Neural Networks (CNNs) feature extraction the Aquila Optimizer (AO) dimension reduction, improving both computational efficiency accuracy. Methods: The proposed method utilized CNNs extract features from images, while AO was employed reduce dimensionality, enhancing performance model. effectiveness this hybrid approach evaluated on three publicly available datasets: ISIC 2019, ISBI 2016, 2017. Results: For 2019 dataset, model achieved 97.46% sensitivity, 98.89% specificity, 98.42% accuracy, 97.91% precision, 97.68% F1-score, 99.12% AUC-ROC. On 2016 it reached 98.45% 98.24% 97.22% 97.84% 97.62% 98.97% 2017, results were 98.44% 98.86% 97.96% 98.12% 97.88% 99.03% outperforms existing advanced techniques, with 4.2% higher 6.2% improvement in 5.8% increase specificity. Additionally, reduced complexity by up 37.5%. Conclusions: deep learning-Aquila (DL-AO) framework offers efficient detection, making suitable deployment resource-constrained environments such as mobile edge computing platforms. integration DL metaheuristic optimization significantly enhances robustness, detection.

Язык: Английский

---

Utilizing convolutional neural networks and vision transformers for precise corn leaf disease identification

Neural Computing and Applications, Год журнала: 2024, Номер unknown

Опубликована: Дек. 5, 2024

Язык: Английский

---

A Hybrid Trio-Deep Feature Fusion Model for Improved Skin Cancer Classification: Merging Dermoscopic and DCT Images

Technologies, Год журнала: 2024, Номер 12(10), С. 190 - 190

Опубликована: Окт. 3, 2024

The precise and prompt identification of skin cancer is essential for efficient treatment. Variations in colour within lesions are critical signs malignancy; however, discrepancies imaging conditions may inhibit the efficacy deep learning models. Numerous previous investigations have neglected this problem, frequently depending on features from a singular layer an individual model. This study presents new hybrid model that integrates discrete cosine transform (DCT) with multi-convolutional neural network (CNN) structures to improve classification cancer. Initially, DCT applied dermoscopic images enhance correct distortions these images. After that, several CNNs trained separately Next, obtained two layers each CNN. proposed consists triple feature fusion. initial phase involves employing wavelet (DWT) merge multidimensional attributes first CNN, which lowers their dimension provides time–frequency representation. In addition, second concatenated. Afterward, subsequent fusion stage, merged first-layer combined second-layer create effective vector. Finally, third bi-layer various integrated. Through process training multiple both original photos DCT-enhanced images, retrieving separate layers, incorporating CNNs, comprehensive representation generated. Experimental results showed 96.40% accuracy after trio-deep shows merging can diagnostic accuracy. outperforms CNN models most recent studies, thus proving its superiority.

Язык: Английский

---

Skin Cancer Image Classification Using Artificial Intelligence Strategies: A Systematic Review

Journal of Imaging, Год журнала: 2024, Номер 10(11), С. 265 - 265

Опубликована: Окт. 22, 2024

The increasing incidence of and resulting deaths associated with malignant skin tumors are a public health problem that can be minimized if detection strategies improved. Currently, diagnosis is heavily based on physicians' judgment experience, which occasionally lead to the worsening lesion or needless biopsies. Several non-invasive imaging modalities, e.g., confocal scanning laser microscopy multiphoton microscopy, have been explored for cancer assessment, aligned different artificial intelligence (AI) assist in diagnostic task, several image features, thus making process more reliable faster. This systematic review concerns implementation AI methods tumor classification following PRISMA guidelines. In total, 206 records were retrieved qualitatively analyzed. Diagnostic potential was found techniques, particularly dermoscopy images, yielding results close perfection. Learning approaches support vector machines neural networks seem preferred, recent focus convolutional networks. Still, detailed descriptions training/testing conditions lacking some reports, hampering reproduction. use an expanding field, future work aiming construct optimal learning strategies. Ultimately, early could optimized, improving patient outcomes, even areas where healthcare scarce.

Язык: Английский

---

Advanced Skin Cancer Classification using EfficientNetB1: Enhancing Diagnosis and Interpretability

Опубликована: Сен. 18, 2024

Язык: Английский

---

Enhancing the performance of CNN models for pneumonia and skin cancer detection using novel fractional activation function

Applied Soft Computing, Год журнала: 2024, Номер unknown, С. 112500 - 112500

Опубликована: Ноя. 1, 2024

Язык: Английский

---