Scholarlink

Background <p>Automated colorectal cancer (CRC) histopathology classification remains challenging due to variations in datasets, staining conditions, and tissue morphology across institutions. Many prior studies apply standard CNN architectures with fixed hyperparameters, leaving limited examination of how model choice and optimization strategies affect performance robustness across heterogeneous CRC data.</p> Methods <p>We evaluate eight transfer learning models on three-class CRC datasets and propose CRC-BayTune, applying Bayesian optimization to tune key training parameters, including learning rate, batch size, with fine-tuning depth. All models are assessed in patch-level experimental settings, and statistical significance is quantified using Friedman tests, repeated-measures ANOVA, and post hoc analyses. Robustness is assessed by introducing controlled Gaussian noise perturbations. Grad-CAM provides qualitative visual explanations by highlighting regions that contribute to model predictions.</p> Results <p>DenseNet201, InceptionV3, InceptionResNetV2, and ResNet50V2 achieved the highest median MCC values of 0.984, 0.982, 0.975, and 0.983, respectively. Statistical analysis confirms that both model architecture (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\textrm{p}=0.059\)</EquationSource> </InlineEquation>, Friedman) and hyperparameter configuration (<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(\textrm{p}=0.027\)</EquationSource> </InlineEquation>, RM-ANOVA) significantly affect performance. Models with deeper feature hierarchies demonstrated more stable convergence and smaller accuracy degradation under noise.</p> Conclusion <p>The results show that systematic hyperparameter tuning can improve the training stability and classification performance of standard CNN models compared with fixed configurations in CRC histopathology tasks. The findings underscore that model performance in this setting is sensitive to choices such as learning rate, batch size, and fine-tuning depth, and that evaluating these factors explicitly can support more reliable use of deep learning models in computational pathology.</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

Transfer learning with Bayesian optimization for colorectal cancer histopathology classification

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月12日

查看全文
  • Houda Saif ALGhafri,
  • Chia S. Lim

摘要

Background

Automated colorectal cancer (CRC) histopathology classification remains challenging due to variations in datasets, staining conditions, and tissue morphology across institutions. Many prior studies apply standard CNN architectures with fixed hyperparameters, leaving limited examination of how model choice and optimization strategies affect performance robustness across heterogeneous CRC data.

Methods

We evaluate eight transfer learning models on three-class CRC datasets and propose CRC-BayTune, applying Bayesian optimization to tune key training parameters, including learning rate, batch size, with fine-tuning depth. All models are assessed in patch-level experimental settings, and statistical significance is quantified using Friedman tests, repeated-measures ANOVA, and post hoc analyses. Robustness is assessed by introducing controlled Gaussian noise perturbations. Grad-CAM provides qualitative visual explanations by highlighting regions that contribute to model predictions.

Results

DenseNet201, InceptionV3, InceptionResNetV2, and ResNet50V2 achieved the highest median MCC values of 0.984, 0.982, 0.975, and 0.983, respectively. Statistical analysis confirms that both model architecture ( \(\textrm{p}=0.059\) , Friedman) and hyperparameter configuration ( \(\textrm{p}=0.027\) , RM-ANOVA) significantly affect performance. Models with deeper feature hierarchies demonstrated more stable convergence and smaller accuracy degradation under noise.

Conclusion

The results show that systematic hyperparameter tuning can improve the training stability and classification performance of standard CNN models compared with fixed configurations in CRC histopathology tasks. The findings underscore that model performance in this setting is sensitive to choices such as learning rate, batch size, and fine-tuning depth, and that evaluating these factors explicitly can support more reliable use of deep learning models in computational pathology.