Related Experiment Video
Updated: May 12, 2025
08:40
Quantitation of Protein Expression and Co-localization Using Multiplexed Immuno-histochemical Staining and Multispectral Imaging
Published on: April 8, 2016
12.7K
Technical note: Impact of tissue section thickness on accuracy of cell classification with a deep learning network
Ida Skovgaard Christiansen1, Rasmus Hartvig1, Thomas Hartvig Lindkær Jensen1,2
1Department of Pathology, Copenhagen University Hospital, Rigshospitalet, Copenhagen, Denmark.
Journal of Pathology Informatics
|May 9, 2025
Summary
Thin histological microsections, especially those from automated microtomes, improve deep learning cell classification accuracy. This study quantifies morphological differences to optimize histopathology cell identification.
Area of Science:
- Histopathology
- Digital Pathology
- Machine Learning in Medicine
Background:
- Developing automated cell classification systems for routine histopathology is crucial for diagnostic efficiency.
- Deep learning (DL) models require high-quality input data, and microsection thickness can significantly impact performance.
Purpose of the Study:
- To determine the optimal histological microsection thickness for a DL-based cell classification system.
- To analyze morphological variations in cells due to differing microsection thicknesses.
Main Methods:
- Liver tissue sections were prepared at five manual thicknesses and one automated (DS) thickness.
- Hepatocytes and non-hepatocytes were annotated and classified using a DL convolutional neural network (ResNet).
- Morphological features were quantified and analyzed using random forest to understand thickness-related differences.
Main Results:
- Automated microtome (DS) sections yielded the highest validation accuracy with the fewest training cells.
- Thinner sections generally showed a trend towards greater classification efficiency.
- Variations in nuclear granularity were the key features for distinguishing cell types, and these were more pronounced in DS and thinner sections.
Conclusions:
- Microsections produced by automated microtomes (DS) and generally thinner sections are optimal for the developed cell classification system.
- Optimizing microsection thickness enhances the accuracy and efficiency of DL-based histopathological analysis.
