Enhancing Skin Cancer Diagnosis (SCD) Using Late Discrete Wavelet Transform (DWT) and New Swarm-Based Optimizers
Ramin Mousa, Saeed Chamani, Mohammad Morsali, Mohammad Kazzazi, Parsa Hatami, Soroush Sarabi · 2024
Read Enhancing Skin Cancer Diagnosis (SCD) Using Late Discrete Wavelet Transform (DWT) and New Swarm-Based Optimizers by Ramin Mousa, Saeed Chamani, Mohammad Morsali, Mohammad Kazzazi, Parsa Hatami, Soroush Sarabi (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-00472.txt.
Introduction
Ramin Mousa and co-authors sharpen skin cancer spotting with wave math and swarm search in a 2024 paper. Early calls matter: late skin cancer kills. Deep nets help, but their settings leave room to grow. The team's pipe runs photos through trained image nets, then a wave split that keeps both broad and fine marks.
A focus step weighs the marks that count most. Three fresh swarm tuners, named for gorilla troops, grey wolves, and foxes, set the net's sizes and weights. Top builds hit near ninety-eight in a hundred right on two test sets, a point or more past past ways.
Catalog record
| CREATOR | Ramin Mousa, Saeed Chamani, Mohammad Morsali, Mohammad Kazzazi, Parsa Hatami, Soroush Sarabi |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 62 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-00472
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Enhancing Skin Cancer Diagnosis (SCD) Using Late Discrete Wavelet Transform (DWT) and New Swarm-Based Optimizers Ramin Mousa, Saeed Chamani, Mohammad Morsali, Mohammad Kazzazi, Parsa Hatami, Soroush Sarabi Abstract Skin cancer (SC) stands out as one of the most life-threatening forms of cancer, with its danger amplified if not diagnosed and treated promptly. Early intervention is critical, as it allows for more effective treatment approaches, potentially halting the disease’s progression and improving patient outcomes. In recent years, Deep Learning (DL) has emerged as a powerful tool in the early detection and skin cancer diagnosis (SCD). Although the DL seems promising for the diagnosis of skin cancer, still ample scope exists for improving model efficiency and accuracy. This paper proposes a novel approach to skin cancer detection, utilizing optimization techniques in conjunction with pre-trained networks and wavelet transformations. First, normalized images will undergo pre-trained networks such as Densenet-121, Inception, Xception, and MobileNet to extract hierarchical features from input images. After feature extraction, the feature maps are passed through a Discrete Wavelet Transform (DWT) layer to capture low and high-frequency components. Then the self-attention module is integrated to learn global dependencies between features and focus on the most relevant parts of the feature maps. The number of neurons and optimization of the weight vectors are performed using three new swarm-based optimization techniques, such as Modified Gorilla Troops Optimizer (MGTO), Improved Gray Wolf Optimization (IGWO), and Fox optimization algorithm, to improve the efficacy of the model and diagnostic accuracy. Evaluation results demonstrate that optimizing weight vectors using optimization algorithms can enhance diagnostic accuracy and make it a highly effective approach for SCD. The proposed method demonstrates substantial improvements in accuracy, achieving top rates of 98.11% with the MobileNet + Wavelet + FOX and DenseNet + Wavelet + Fox combination on the ISIC-2016 dataset and 97.95% with the Inception + Wavelet + MGTO combination on the ISIC-2017 dataset, which improves accuracy by at least 1% compared to other methods. Introduction SCD is one of the dangerous diseases that is treatable with early diagnosis but still claims the lives of many people annually. In 2022, 331,722 people were diagnosed with skin cancer, and 58,667 of them lost their lives due to the disease, and these numbers are increasing every year [ 1 ] . Early diagnosis plays a crucial role in the treatment process and the patient’s chance of survival. It can significantly increase the likelihood of the disease being operable, allow treatment methods to be applied more effectively, and prevent the disease from progressing [ 2 ] . Skin cancer symptoms often appear as skin lesions that look very similar to harmless ones and can only be recognized by dermatologists. Because of this, many people with the disease ignore these skin changes and do nothing until the cancer has advanced. This delay in diagnosis can slow down treatment, make it less effective, and allow the disease to spread to other body parts. A significant reason for this delay is the difficulty of current diagnostic methods, like sampling. Machine learning and deep learning can play a significant role in skin cancer detection by providing automated detection. The most critical issue in using learning algorithms for disease detection is their accuracy. Therefore, improving the accuracy of these algorithms is of utmost importance. As we know CNN models perform very well in classification problems. Thus, we use CNN as our base model. In this article, we increased the model’s accuracy compared to previous models. The key innovations we employed in this paper to enhance accuracy are the use of the swarm-based optimizers and the integration of the Wavelet Transform. In this article, we aimed to significantly increase the accuracy of the model used for disease diagnosis. First, we preprocessed the data in the dataset and resolved the data imbalance problem, and then the pictures were ready for giving as the input to the model. To improve the model’s accuracy, we employed innovations that can be broadly divided into the following two stages: • The images are fed into four pre-trained models. In the feature extraction section, we utilized the wavelet transform to extract suitable features for training the data. The output is then passed through a self-attention mechanism. Since the classification task is binary, the output is passed through a sigmoid function. • After completing all these stages, the network is fed into an ANN, where three advanced and modern optimizers (IGWO, FOX, MGTO) are used to update and modify the model’s parameters, and their performance is evaluated. These optimizers update the parameters optimally, which significantly improves the model’s accuracy and overall performance. In previous studies on SCD, the methods, tools, and optimizers mentioned above were not utilized this way. The combination of using wavelet transforms for feature extraction, evaluating multiple new optimizers, and analyzing four different models has led to significant improvements in our results compared to prior studies in this field. This approach could be an important step in advancing deep models for diagnosing various diseases based on medical images. Literature review In [ 3 ] , first, FairDisCo, a deep neural network (DNN) with contrastive learning, was proposed. This model includes a network branch that reduces sensitive details from model representations and another contrastive branch to improve representation learning and thus improve SCD diagnosis accuracy. Secondly, introduced MDViT, a multi-domain Vision Transformer (VIT) with domain adapters. MDViT aims to handle model data hunger and negative knowledge transfer (NKT), which can reduce model performance on domains with inter-domain heterogeneity. MDViT also uses mutual knowledge distillation to improve representation learning across domains. In [ 4 ] , first, noisy samples are removed from the data using an Adaptive Median Filter (AMF). Then, features are extracted from segmented data using Kernel Principal Component Analysis (KPCA). Finally, an Augmented May Fly optimized with the K-Nearest Neighbors (AMFO-KNN) technique improves SCD classification.In [ 5 ] , compared commonly used pre-trained models for detecting SCD using the same dataset to determine which one has the highest accuracy. This research was done to prevent biased results that can happen when using only one model and dataset in various methods by introducing which model is best to train the SCD datasets.In [ 6 ] , a new combined model, Multi-Model Fusion for Skin Cancer Detection (MMF-SCD), has been introduced. This model uses pre-trained VGG-16, VGG-19, and ResNet-50 models to extract features accurately. After feature extraction, feature optimization is performed, and the model predicts diseases. This method is proposed to deal with the accuracy limit in SCD classification that exists in several methods. In [ 7 ] , presented a new multi-scale cross-attention method called M-VAN Unet for skin lesion segmentation network, based on the Visual Attention Network (VAN). This method handles the challenge of learning both local and global features. Unlike Transformers, which were initially developed for natural language processing (NLP) and may destroy the multidimensional nature of images, M-VAN Unet preserves the 2D structure of the image and can capture low-level features more effectively. In addition, they introduced two attention mechanisms: MSC-Attention for multi-scale channel attention and LKA-Cross-Attention, a cross-attention mechanism based on large kernel attention (LKA). These mechanisms enable the capture of low-level features and simplify global information interaction. In [