Scholarlink

<p>Brain tumors are said to be one of the deadliest diseases in the world. Determining the identity and classification of a brain tumor is an essential step in improving our understanding of its underlying causes. The location of brain cancers can be determined using a range of diagnostic imaging methods. The new paradigm of automated systems for medical picture recognition has been made possible by the advancement of deep learning techniques. Magnetic resonance imaging (MRI) is the most extensively employed tool of screening for brain tumors, although sonography, X-rays, and other techniques can also be utilized. The CNN models are a revolution in image classification tasks. However, optimizing them for lightweight applications is still a challenge. This study focuses on deploying a time-efficient CNN that can be deployed using limited resources. The proposed model showed 99% accuracy during training and 97% accuracy during testing using a dataset of 3264 MRIs. A detailed ablation study was conducted where the model’s depth and hyperparameter tuning were justified, and the model’s computational efficiency and time complexity were discussed. Models proved very efficient with a time complexity of 31.80&#xa0;M and performed well on all metrics, namely precision of 97.26%, recall of 97%, and F1-score of 97.24%. When comparing its outcomes with other prominent state-of-the-art methods, the proposed model detects brain malignancies early with low computational cost.</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

An improved CNN model for characterizing brain tumors using deep learning

  • 期刊论文

  • 发表日期：

    2026年5月2日

查看全文
  • Ajay Indian,
  • Gaurav Meena,
  • Dikshant Sharma

摘要

Brain tumors are said to be one of the deadliest diseases in the world. Determining the identity and classification of a brain tumor is an essential step in improving our understanding of its underlying causes. The location of brain cancers can be determined using a range of diagnostic imaging methods. The new paradigm of automated systems for medical picture recognition has been made possible by the advancement of deep learning techniques. Magnetic resonance imaging (MRI) is the most extensively employed tool of screening for brain tumors, although sonography, X-rays, and other techniques can also be utilized. The CNN models are a revolution in image classification tasks. However, optimizing them for lightweight applications is still a challenge. This study focuses on deploying a time-efficient CNN that can be deployed using limited resources. The proposed model showed 99% accuracy during training and 97% accuracy during testing using a dataset of 3264 MRIs. A detailed ablation study was conducted where the model’s depth and hyperparameter tuning were justified, and the model’s computational efficiency and time complexity were discussed. Models proved very efficient with a time complexity of 31.80 M and performed well on all metrics, namely precision of 97.26%, recall of 97%, and F1-score of 97.24%. When comparing its outcomes with other prominent state-of-the-art methods, the proposed model detects brain malignancies early with low computational cost.