Scholarlink

<p>Lung adenocarcinoma (LUAD) represents one of the causes of cancer-associated death that is usually diagnosed at its advanced stages, when it is too late to treat it. Timely and correct diagnosis is essential to enhance patient survival and outcomes. This paper aims to present a deep learning model that is a combination of a modified AlexNet and Bilateral Long Short-Term Memory (Bi-LSTM) to predict LUAD. The methodology involves data preprocessing to deal with imbalance and normalization, optimization of feature selection to determine the most relevant genes and tuning of hyperparameters to optimize model performance. The proposed approach has a high predictive accuracy of 98.58% and 99.01% on two independent LUAD datasets, which proves its potential to be used in clinical practice and its high superiority to the current methods. Although it tells some details of the algorithms, they are provided in such a limited amount that they provide a picture of the robustness and reliability of the framework, which directly leads to its applicability to early LUAD detection and better patient care.</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

Modified AlexNet for accurate lung adenocarcinoma prediction with enhanced UMAP and optimized feature selection

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月18日

查看全文
  • B. Jyothi,
  • L. Mary Gladence

摘要

Lung adenocarcinoma (LUAD) represents one of the causes of cancer-associated death that is usually diagnosed at its advanced stages, when it is too late to treat it. Timely and correct diagnosis is essential to enhance patient survival and outcomes. This paper aims to present a deep learning model that is a combination of a modified AlexNet and Bilateral Long Short-Term Memory (Bi-LSTM) to predict LUAD. The methodology involves data preprocessing to deal with imbalance and normalization, optimization of feature selection to determine the most relevant genes and tuning of hyperparameters to optimize model performance. The proposed approach has a high predictive accuracy of 98.58% and 99.01% on two independent LUAD datasets, which proves its potential to be used in clinical practice and its high superiority to the current methods. Although it tells some details of the algorithms, they are provided in such a limited amount that they provide a picture of the robustness and reliability of the framework, which directly leads to its applicability to early LUAD detection and better patient care.