Scholarlink

Cell counting in microscopy images plays a crucial role in biomedical research and diagnostics. This paper introduces an enhanced deep learning approach for cell counting using Fully Convolutional Regression Networks (FCRN) and U-net architectures, augmented with residual concatenation connections incorporating Neural Arithmetic Logic Units (NALU) and Neural Accumulator (NAC) modules. These numerically biased units provide improved generalization, particularly in high-density cell images. The methodology is validated on synthetic datasets and modified BBBC005 benchmark datasets, showing significant improvements in both interpolation and extrapolation tasks. The proposed models demonstrate superior accuracy and robustness compared to baseline architectures, offering a more effective solution for real-world applications.

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

Enhancing Cell Counting in Microscopy Images with Residual Concatenation of NALU and NAC Modules in Deep Learning Models

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Dinesh Besiahgari

摘要

Cell counting in microscopy images plays a crucial role in biomedical research and diagnostics. This paper introduces an enhanced deep learning approach for cell counting using Fully Convolutional Regression Networks (FCRN) and U-net architectures, augmented with residual concatenation connections incorporating Neural Arithmetic Logic Units (NALU) and Neural Accumulator (NAC) modules. These numerically biased units provide improved generalization, particularly in high-density cell images. The methodology is validated on synthetic datasets and modified BBBC005 benchmark datasets, showing significant improvements in both interpolation and extrapolation tasks. The proposed models demonstrate superior accuracy and robustness compared to baseline architectures, offering a more effective solution for real-world applications.