Combining Datasets with Different Label Sets for Improved Nucleus Segmentation and Classification

Segmentation and classification of cell nuclei in histopathology images using deep neural networks (DNNs) can save pathologists' time for diagnosing various diseases, including cancers, by automating cell counting and morphometric assessments. It is now well-known that the accuracy of DNNs incr...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2023-10
Main Authors: Parulekar, Amruta, Kanwat, Utkarsh, Gupta, Ravi Kant, Chippa, Medha, Jacob, Thomas, Bameta, Tripti, Rane, Swapnil, Sethi, Amit
Format: Article
Language:English
Subjects:
Artificial neural networks
Classification
Datasets
Histopathology
Image segmentation
Labels
Medical imaging
Training
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Segmentation and classification of cell nuclei in histopathology images using deep neural networks (DNNs) can save pathologists' time for diagnosing various diseases, including cancers, by automating cell counting and morphometric assessments. It is now well-known that the accuracy of DNNs increases with the sizes of annotated datasets available for training. Although multiple datasets of histopathology images with nuclear annotations and class labels have been made publicly available, the set of class labels differ across these datasets. We propose a method to train DNNs for instance segmentation and classification on multiple datasets where the set of classes across the datasets are related but not the same. Specifically, our method is designed to utilize a coarse-to-fine class hierarchy, where the set of classes labeled and annotated in a dataset can be at any level of the hierarchy, as long as the classes are mutually exclusive. Within a dataset, the set of classes need not even be at the same level of the class hierarchy tree. Our results demonstrate that segmentation and classification metrics for the class set used by the test split of a dataset can improve by pre-training on another dataset that may even have a different set of classes due to the expansion of the training set enabled by our method. Furthermore, generalization to previously unseen datasets also improves by combining multiple other datasets with different sets of classes for training. The improvement is both qualitative and quantitative. The proposed method can be adapted for various loss functions, DNN architectures, and application domains.
ISSN:2331-8422