Scholarlink

<p>Early detection is essential for the successful treatment of skin cancer (SC), one of the most prevalent cancers worldwide. Dermatologists frequently deal with issues that might impair diagnostic results, such as high data requirements, the possibility of human error, and stringent time constraints. In this paper, enhancing early skin cancer detection and diagnosis with a rotation-invariant coordinate convolutional neural network for medical advancements (ESCDD-RIC-CNN-RMA) is proposed. Firstly, the image is collected from the HAM 10000 dataset. Then, the images are pre-processed using an inverse unscented Kalman filter (I-UKF) for resizing, noise reduction, and normalization. Then, the pre-processed images are fed into a high-order time-reassigned synchrony squeezing transform (HTSST) for feature extraction. HTSST is used to extract relevant features from lesion images like colour, shape, texture, and border irregularity. Then, the extracted features are given to a rotation-invariant coordinate convolutional neural network (RIC-CNN) for skin cancer diagnosis. It classifies like dermatofibroma (DF), vascular lesion (VASC), benign keratosis (BKL), basal cell carcinoma (BCC), actinic keratosis (AKIEC), melanocytic nevus (NV), and melanoma (MEL). The proposed method, implemented in Python, demonstrates substantial improvements in accuracy, precision, recall, specificity, and F1 score. The proposed ESCDD-RIC-CNN-RMA method achieves the best performance with 98.10% accuracy, 98.50% specificity, and 99% recall, demonstrating consistent high performance and minimal false positives, compared with existing methods, like Pre-operative high-frequency ultrasound: a reliable management tool in auricular and nasal non-melanoma skin cancer (POHFU-RMT-ANNMS), skin cancer classification utilizing a convolutional neural network: an exploration into deep learning (SCC-CNN-EDL), and artificial intelligence‑driven enhanced skin cancer diagnosis: leveraging convolutional neural networks with discrete wavelet transformation(AI-SCD-CNN-DWT).</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhancing early skin cancer detection and diagnosis through a rotation-invariant coordinate convolutional neural network for revolutionary medical advancements

  • 期刊论文

  • 发表日期：

    2025年12月17日

查看全文
  • Shweta Ashish Koparde,
  • Deepa Abin,
  • Sonali Mahendra Kothari,
  • Jayesh Mohanrao Sarwade,
  • Kapil Netaji Vhatkar

摘要

Early detection is essential for the successful treatment of skin cancer (SC), one of the most prevalent cancers worldwide. Dermatologists frequently deal with issues that might impair diagnostic results, such as high data requirements, the possibility of human error, and stringent time constraints. In this paper, enhancing early skin cancer detection and diagnosis with a rotation-invariant coordinate convolutional neural network for medical advancements (ESCDD-RIC-CNN-RMA) is proposed. Firstly, the image is collected from the HAM 10000 dataset. Then, the images are pre-processed using an inverse unscented Kalman filter (I-UKF) for resizing, noise reduction, and normalization. Then, the pre-processed images are fed into a high-order time-reassigned synchrony squeezing transform (HTSST) for feature extraction. HTSST is used to extract relevant features from lesion images like colour, shape, texture, and border irregularity. Then, the extracted features are given to a rotation-invariant coordinate convolutional neural network (RIC-CNN) for skin cancer diagnosis. It classifies like dermatofibroma (DF), vascular lesion (VASC), benign keratosis (BKL), basal cell carcinoma (BCC), actinic keratosis (AKIEC), melanocytic nevus (NV), and melanoma (MEL). The proposed method, implemented in Python, demonstrates substantial improvements in accuracy, precision, recall, specificity, and F1 score. The proposed ESCDD-RIC-CNN-RMA method achieves the best performance with 98.10% accuracy, 98.50% specificity, and 99% recall, demonstrating consistent high performance and minimal false positives, compared with existing methods, like Pre-operative high-frequency ultrasound: a reliable management tool in auricular and nasal non-melanoma skin cancer (POHFU-RMT-ANNMS), skin cancer classification utilizing a convolutional neural network: an exploration into deep learning (SCC-CNN-EDL), and artificial intelligence‑driven enhanced skin cancer diagnosis: leveraging convolutional neural networks with discrete wavelet transformation(AI-SCD-CNN-DWT).