Differential cell counts using center-point networks achieves human-level accuracy and efficiency over segmentation
Authors:
Open Access
- 19 August 2021
- journal article
- research article
- Published by Springer Nature in Scientific Reports
- Vol. 11 (1) , 1-14
Abstract
Differential cell counts is a challenging task when applying computer vision algorithms to pathology. Existing approaches to train cell recognition require high availability of multi-class segmentation and/or bounding box annotations and suffer in performance when objects are tightly clustered. We present differential count network (“DCNet”), an annotation efficient modality that utilises keypoint detection to locate in brightfield images the centre points of cells (not nuclei) and their cell class. The single centre point annotation for DCNet lowered burden for experts to generate ground truth data by 77.1% compared to bounding box labeling. Yet centre point annotation still enabled high accuracy when training DCNet on a multi-class algorithm on whole cell features, matching human experts in all 5 object classes in average precision and outperforming humans in consistency. The efficacy and efficiency of the DCNet end-to-end system represents a significant progress toward an open source, fully computationally approach to differential cell count based diagnosis that can be adapted to any pathology need.Keywords
Funding Information
- National Health and Medical Research Council (1141479)
This publication has 34 references indexed in Scilit:
- Single-Cell Phenotype Classification Using Deep Convolutional Neural NetworksSLAS Discovery, 2016
- A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability ResearchJournal of Chiropractic Medicine, 2016
- Microscopy cell counting and detection with fully convolutional regression networksComputer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization, 2016
- Locality Sensitive Deep Learning for Detection and Classification of Nuclei in Routine Colon Cancer Histology ImagesIEEE Transactions on Medical Imaging, 2016
- U-Net: Convolutional Networks for Biomedical Image Segmentation2015
- DeepPose: Human Pose Estimation via Deep Neural Networks2014
- How Romanowsky stains work and why they remain valuable — including a proposed universal Romanowsky staining mechanism and a rational troubleshooting schemeBiotechnic & Histochemistry, 2011
- Performance Evaluation of the CellaVision DM96 SystemAmerican Journal of Clinical Pathology, 2005
- Bronchoalveolar lavage fluid differential cell count. How many cells should be counted?2002
- Differential cell analysis of cytocentrifuged bronchoalveolar fluid samples affected by the area counted.2000