Scholarlink

<p>Cancer is the most life-threatening disease that occurs, while the uncontrollable enlargement of cells and tissues in the individual body. Skin cancer is the most frequently widespread form of cancer, which starts in the skin’s outermost layer. The complicated background and variety of lesion characteristics in the dermoscopy images pose important challenges. The existing approaches focus on improving the detection accuracy by utilizing with more complex and larger models. However, these methods neglect the high variability that arises from important inter-class overlapping and intra-class differences in characters. Recent enhancement in deep learning approaches were allowed the enhancement of Computer-Aided Diagnosis (CAD) systems, which promote earlier detection and uncover hidden causes of skin cancer. To meet this requirement, an automated skin cancer recognition approach is designed utilizing advanced deep learning approaches. Primitively, an adequate set of raw images is gathered on the standard data sources. Subsequently, the gathered input data are fed as input to the Transformer-based Dilated Recurrent Residual Unet++(TD-R2Unet++) for segmenting the abnormality or lesion region. Moreover, the resultant segmented data is given to the detection process here; the Deformable Adaptive Residual Attention Network (PD-ARAN) mechanism is designed to easily identify abnormal and irrelevant tissue and tumor growth within a restricted duration. Moreover, the improved Intellectual Position Updated Concept-based Orangutan Optimization Algorithm (IPUC-OOA) is proposed to get the best optimal result and it has the ability to timely optimize the significant parameters like epoch count, hidden neuron count, and learning rate in PD-ARAN mechanism. Thus, the system’s performance is investigated with different performance measures. In the estimation process, the proposed framework has achieved better detection outcomes 94.1% accuracy, 95.1% specificity, and 94.1% NPV values in terms of 48th batch size. In contrary, the designed framework achieves the best outcomes, which indicates the supremacy in diagnosing skin cancer.</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

Developing an Automated Model of Pyramidal Deformable Adaptive RAN and Unet++-Based Segmentation for Skin Cancer Disease Diagnosis

  • 期刊论文

  • 发表日期：

    2026年4月7日

查看全文
  • Malapati Venkata Narayana,
  • Nidamanuru Srihari Rao,
  • Krishna Kumar .N,
  • Dr Surya Kiran Chebrolu,
  • Naveen Reddy Nemili,
  • Supraja Ballari,
  • Ravindra Changala

摘要

Cancer is the most life-threatening disease that occurs, while the uncontrollable enlargement of cells and tissues in the individual body. Skin cancer is the most frequently widespread form of cancer, which starts in the skin’s outermost layer. The complicated background and variety of lesion characteristics in the dermoscopy images pose important challenges. The existing approaches focus on improving the detection accuracy by utilizing with more complex and larger models. However, these methods neglect the high variability that arises from important inter-class overlapping and intra-class differences in characters. Recent enhancement in deep learning approaches were allowed the enhancement of Computer-Aided Diagnosis (CAD) systems, which promote earlier detection and uncover hidden causes of skin cancer. To meet this requirement, an automated skin cancer recognition approach is designed utilizing advanced deep learning approaches. Primitively, an adequate set of raw images is gathered on the standard data sources. Subsequently, the gathered input data are fed as input to the Transformer-based Dilated Recurrent Residual Unet++(TD-R2Unet++) for segmenting the abnormality or lesion region. Moreover, the resultant segmented data is given to the detection process here; the Deformable Adaptive Residual Attention Network (PD-ARAN) mechanism is designed to easily identify abnormal and irrelevant tissue and tumor growth within a restricted duration. Moreover, the improved Intellectual Position Updated Concept-based Orangutan Optimization Algorithm (IPUC-OOA) is proposed to get the best optimal result and it has the ability to timely optimize the significant parameters like epoch count, hidden neuron count, and learning rate in PD-ARAN mechanism. Thus, the system’s performance is investigated with different performance measures. In the estimation process, the proposed framework has achieved better detection outcomes 94.1% accuracy, 95.1% specificity, and 94.1% NPV values in terms of 48th batch size. In contrary, the designed framework achieves the best outcomes, which indicates the supremacy in diagnosing skin cancer.