Scholarlink

Researchers have always been interested in inventing ways to detect medical issues using AI technologies to fasten the procedure and to receive the most accurate and specific diagnosis results. Kidney stone is the most common yet fatal disease found in the human body. If not detected at the right time, it can cause severe pains with intense infections and might also lead to permanent damage to the kidney. Several researchers have proposed their work on various artificial intelligence algorithms to detect kidney stones. The following study compares several deep learning algorithms as worked upon by various authors on the Kidney Stone Detection system, to find which algorithm performs best based on factors like accuracy and sensitivity with a slight insight into the imaging scanning techniques and various models of deep learning that could be used during detection. It is found that 3D-UNet outperformed all the algorithms with its 99.96% accuracy rate with Inception-ResNetV2 having a 100% sensitivity rate among all the models.

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

Comparative Review of Deep Learning Models for Renal Stone Detection and Classification

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Shivani Verma,
  • Pawan Kumar Singh,
  • Himani Gautam,
  • Shimpy Goyal

摘要

Researchers have always been interested in inventing ways to detect medical issues using AI technologies to fasten the procedure and to receive the most accurate and specific diagnosis results. Kidney stone is the most common yet fatal disease found in the human body. If not detected at the right time, it can cause severe pains with intense infections and might also lead to permanent damage to the kidney. Several researchers have proposed their work on various artificial intelligence algorithms to detect kidney stones. The following study compares several deep learning algorithms as worked upon by various authors on the Kidney Stone Detection system, to find which algorithm performs best based on factors like accuracy and sensitivity with a slight insight into the imaging scanning techniques and various models of deep learning that could be used during detection. It is found that 3D-UNet outperformed all the algorithms with its 99.96% accuracy rate with Inception-ResNetV2 having a 100% sensitivity rate among all the models.