Scholarlink

<p>Cancer continues to be a primary cause of death worldwide, highlighting the critical need for early diagnosis methods. Automated, quick, and efficient technologies are critical to this endeavor, yet considerable gaps remain in this field. A comprehensive review was undertaken to examine seven cancer types characterized by elevated prevalence and mortality: lung, prostate,&#xa0;brain, skin, breast,&#xa0;leukemia, and&#xa0;colorectal&#xa0;cancer. The study aimed to reveal gaps in the existing research and compare traditional machine learning (TML) with deep learning (DL) methodologies, since such contrasts have been not much explored. A total of 320 publications were carefully chosen for study, including 150 that focused on TML methods and 170 that address DL techniques for the classification of cancer. Diverse parameters were used to assess these investigations, encompassing publication year, employed databases, data&#xa0;sample, classifier, modalities, and evaluation metrics. Separate evaluations were conducted for each cancer type and methodology, yielding 14 unique review tables. The assessment of each cancer type using ML/DL independently relied on four standard criteria: High&#xa0;performance (&gt; 99%), Limited&#xa0;performance (&lt; 85%), key findings, and key challenges. These studies were accompanied by a brief descriptive outline of the features, classifiers, public databases, and evaluation metrics that were utilized in the review process. The study concluded by offering general conclusions that highlighted the overall findings, overall challenges observed during&#xa0;the investigation. This thorough review seeks to improve clinical applications and guide future research initiatives in the persistent fight against 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

Transformative Advances in AI for Precise Cancer Detection: A Comprehensive Review of Non-Invasive Techniques

  • 期刊论文

  • 发表日期：

    2025年1月11日

查看全文
  • Hari Mohan Rai,
  • Joon Yoo,
  • Serhii Dashkevych

摘要

Cancer continues to be a primary cause of death worldwide, highlighting the critical need for early diagnosis methods. Automated, quick, and efficient technologies are critical to this endeavor, yet considerable gaps remain in this field. A comprehensive review was undertaken to examine seven cancer types characterized by elevated prevalence and mortality: lung, prostate, brain, skin, breast, leukemia, and colorectal cancer. The study aimed to reveal gaps in the existing research and compare traditional machine learning (TML) with deep learning (DL) methodologies, since such contrasts have been not much explored. A total of 320 publications were carefully chosen for study, including 150 that focused on TML methods and 170 that address DL techniques for the classification of cancer. Diverse parameters were used to assess these investigations, encompassing publication year, employed databases, data sample, classifier, modalities, and evaluation metrics. Separate evaluations were conducted for each cancer type and methodology, yielding 14 unique review tables. The assessment of each cancer type using ML/DL independently relied on four standard criteria: High performance (> 99%), Limited performance (< 85%), key findings, and key challenges. These studies were accompanied by a brief descriptive outline of the features, classifiers, public databases, and evaluation metrics that were utilized in the review process. The study concluded by offering general conclusions that highlighted the overall findings, overall challenges observed during the investigation. This thorough review seeks to improve clinical applications and guide future research initiatives in the persistent fight against cancer.