Figure - available from: Multimedia Tools and Applications
This content is subject to copyright. Terms and conditions apply.
Accuracy metric of the classification based, respectively on, color, shape, texture, fusion features and the proposed approach
Source publication
Malignant melanoma is one of the most serious and deadly types of skin cancer, fortunately it is treatable if detected at an early stage. Many Computer-Aided Diagnosis (CAD) systems are proposed in the literature to assist in detecting this type of cancer. The vast majority of them are performed in four main steps, which are preprocessing, image se...
Similar publications
Customer churn is an important issue in increasing both the long and short-term revenues of a company. If companies identify customers’ churn behavior, they can prevent customer loss, ensure customer loyalty, and in turn gain better financial returns. The telecommunication sector is a customer-oriented sector that requires customer retention to sur...
Gynaecological cancer is a prevalent type of cancer affecting women worldwide, and early detection is critical for successful treatment and prevention of cancer recurrence. In our study, we introduce an innovative method to predict cervical cancer risk using a combination of deep learning models. Our approach employs a nested ensemble technique, wh...
Speech Emotion Recognition (SER) is the process of recognizing and classifying emotions expressed through speech. SER greatly facilitates personalized and empathetic interactions, enhances user experiences, enables sentiment analysis, and finds applications in psychology, healthcare, entertainment, and gaming industries. However, accurately detecti...
Early and accurate diagnosis has the potential to slow the progression of dementia and is a highly critical step in determining the appropriate treatment plan. Dementia is a condition in which various neurological anomalies occur. These anomalies can be detected with Magnetic Resonance Imaging (MRI) imaging. MRI imaging technique is widely used to...
Melanoma, the most lethal type of skin cancer, presents a substantial public health challenge. Detecting melanoma promptly is paramount for enhancing patient outcomes and preserving lives. In this research, we introduce an ensemble-learning model with Weighted Voting Classifier for Dermoscopic Image Classification with the goal of enhancing the acc...
Citations
... Some commonly used CNN architectures are AlexNet [5], VGG (Visual Geometry Group) [6], ResNet (Residual Neural Network) [7], DenseNet [8], EfficientNet [9]. Some recent work for analysis of melanoma had been carried out by various deep learning approaches [10][11][12][13][14][15]. are typically inversely proportional to the class frequencies. ...
... where n is the number of samples and β ∈ [0, 1) is a hyperparameter can be calculated as Eq. (12). ...
Accurate skin disease detection is one of the most challenging tasks due to high-class imbalance and limited labeled datasets. Recently Deep Convolutional Neural Network (DCNN) with ensemble learning has achieved significant popularity in skin cancer classification. However, implementing DCNN models with ensemble learning is not feasible for deployment on portable diagnostic devices due to the limitation in computing resources and computing time. This paper proposes a Channel Attention and Adaptive Class Balanced Focal Loss function based lightweight Deep CNN model (CACBL-Net) for handling the issues of data imbalance and limited computing resources of portable diagnostic devices, such as mobile phones or tablets. Channel attention explores interdependencies between channels by recalibrating channel-wise feature responses. To deal with the issue of high-class imbalance, the proposed method used an adaptive class balance focal loss function which can quickly concentrate the model on complex cases while automatically downweighting the contribution of easy examples during training. The proposed CACBL-Net is validated on three popular skin cancer datasets which are HAM-10000, PAD-UFES-20, and MED-NODE. Dermoscopic, non-dermoscopic and smartphone images are taken from all three datasets for experimental work. The quantitative findings indicate that the proposed CACBL-Net model achieved a sensitivity of 90.60%, 91.88%, and 91.31% for the HAM-10000, PAD-UFES-20, and MED-NODE datasets, respectively. Additionally, the average prediction time per patient was recorded at 0.006, 0.010, and 0.011 s. These results demonstrate superior performance compared to other state-of-the-art deep learning models. The experimental finding suggested that the proposed method can achieve a significant performance at a low cost of computational resources and inference time, which makes it potentially feasible for deployment in portable diagnostic devices for automated diagnosis of skin lesions.
... In recent years, ensemble-learning techniques, specifically voting classifiers, have shown considerable promise in improving the accuracy and reliability of melanoma and other skin cancer diagnosis [3,4]. Utilizing voting classifier ensembles for dermoscopic image classification is an intriguing approach to harness the power of ensemble learning, ultimately providing dermatologists with a robust tool for more precise melanoma diagnosis [5]. ...
... Feature extraction was accomplished through two fundamental techniques: the Gray-Level Co-Occurrence Matrix (GLCM) and the Local Binary Pattern (LBP). Leveraging the GLCM Matrix, we derived the essential parameters detailed in our previous work [5], which include Energy (Angular Second Moment (ASM)), Entropy, Inverse Difference Moment (IDM), Contrast, Correlation, Variance, Mean, Standard Deviation, Skewness, and Kurtosis. ...
Melanoma, the most lethal type of skin cancer, presents a substantial public health challenge. Detecting melanoma promptly is paramount for enhancing patient outcomes and preserving lives. In this research, we introduce an ensemble-learning model with Weighted Voting Classifier for Dermoscopic Image Classification with the goal of enhancing the accuracy of melanoma diagnosis. Our approach integrates the outputs of three distinct Machine Learning (ML) classifiers, all based on the Support Vector Machines (SVMs) algorithm. Each classifier is trained separately on color, shape, and textural dermoscopic images features. By assigning appropriate weights to these individual classifiers, our ensemble learning proposal leverages their complementary strengths, facilitating a synergistic decision-making process. This not only enhances classification accuracy but also bolsters the model's resilience against variations in image characteristics. To estimate the optimal weights for each classifier, we employ genetic algorithms based on their respective performance metrics. We evaluate our methodology using the publicly available PH2 dermoscopic database and achieve an impressive 99% of accuracy, 97.5% of sensitivity, and 99% of specificity. Our findings demonstrate that employing the ensemble learning weighted voting technique outperforms using each ML architecture independently and voting majority technique. In conclusion, our study underscores the potential of ensemble learning with weighted voting, in dermoscopic image classification for melanoma diagnosis, offering a promising avenue for enhancing diagnostic accuracy and, consequently, patient outcomes.
