Scholarlink

This study deviates from the norm by examining the impact of batch size on CNN models that have been trained to analyze thoracic X-ray images. To analyze the NIH Chest X-ray dataset, we implemented two optimization techniques, namely stochastic gradient descent with momentum (SGDM) and adaptive moment estimation (ADAM), in addition to 18 distinct CNN models. Following a thorough examination of the impacts associated with batch sizes ranging from 16 to 1024, the researchers determined that a batch size of 512 yielded the highest performance for the majority of CNN models. By elucidating the intricate relationship between batch size and model performance, our exhaustive analytical investigation elucidates the complex dynamics underpinning CNNs using chest X-ray analysis as a case study. The findings shed light on the advantages of a specific sample size and contribute to our understanding of how this parameter impacts the accuracy and efficacy of convolutional neural network (CNN) models in the classification of medical images. Academics and practitioners can use our findings to enhance the performance of CNN in medical image analysis by effectively managing the intricate interplay between sample size and optimization methods, as suggested by our findings.

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

Batch Size Optimization in CNN Models for Chest X-ray Image Analysis: An Analytical Investigation

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Prabira Kumar Sethy,
  • Raman Sahu,
  • Preesat Biswas,
  • Akshay Shirole,
  • Santi Kumari Behera

摘要

This study deviates from the norm by examining the impact of batch size on CNN models that have been trained to analyze thoracic X-ray images. To analyze the NIH Chest X-ray dataset, we implemented two optimization techniques, namely stochastic gradient descent with momentum (SGDM) and adaptive moment estimation (ADAM), in addition to 18 distinct CNN models. Following a thorough examination of the impacts associated with batch sizes ranging from 16 to 1024, the researchers determined that a batch size of 512 yielded the highest performance for the majority of CNN models. By elucidating the intricate relationship between batch size and model performance, our exhaustive analytical investigation elucidates the complex dynamics underpinning CNNs using chest X-ray analysis as a case study. The findings shed light on the advantages of a specific sample size and contribute to our understanding of how this parameter impacts the accuracy and efficacy of convolutional neural network (CNN) models in the classification of medical images. Academics and practitioners can use our findings to enhance the performance of CNN in medical image analysis by effectively managing the intricate interplay between sample size and optimization methods, as suggested by our findings.