The RV-PBS (Ramakrishna Vivekananda Peripheral Blood Smear) dataset

---

Abstract

Automating blood cell counting and detection from smear slides holds significant potential for aiding doctors in disease diagnosis through blood tests. However, existing literature has not adequately addressed using whole slide data in this context. This study introduces the novel RV-PBS dataset, comprising ten distinct peripheral blood smear classes, each featuring multiple multi-class White Blood Cells per slide, specifically designed, for instance segmentation benchmarks. While conventional instance segmentation models like Mask R-CNN exhibit promising results in segmenting medical artifact instances, they face challenges in scenarios with limited samples and class imbalances within the dataset. This challenge prompted us to explore innovative techniques such as domain adaptation using a similar dataset to enhance the classification accuracy of Mask R-CNN, a novel approach in the domain of medical image analysis. Our study has successfully established a comprehensive pipeline capable of segmenting, detecting, and classifying blood samples from slides, striking an optimal balance between computational complexity and accurate classification of medical artifacts. This advancement enables precise cell counting and classification, facilitating doctors in refining their diagnostic analyses.

This repository hosts the dataset created for the paper titled Advancing instance segmentation and WBC classification in peripheral blood smear through domain adaptation: A study on PBC and the novel RV-PBS datasets.

Find the Figshare version of the datasets, models and other stuffs here

---

The dataset is annotated using CVAT. We are planning to release an extended version of this dataset soon. If you are a haematologist, then you could help us by annotating and adding more data. Please make sure that the data is ethically cleared before uploading new data in public servers, such as Github.

Data creation process

For this study, we have created a novel WBC dataset comprising 10 classes known as the ramakirshna vivekananda peripheral blood smear (RV-PBS) dataset. Air dried peripheral blood smears are stained by Leishman stain following standard protocol and examined under an oil immersion lens using 10X eyepiece magnification (final magnification–1000X). Photographs taken by iPhone XR 12-megapixel camera with f/1.8 aperture. The dataset comprises high resolution (4032 x 3024) images of blood smear slides.

Questions or want to discuss about something ?

Submit an issue.

Snapshot of dataset creation using CVAT

Some relevant stuffs from the paper

Please study the paper for getting more insights. Here are some snapshots from the paper:

Smear slides cropped dataset

Schematic diagram for extraction of cells ready to be sent to domain adaptation pipeline

Classification model used with different backbones

Results Table

Results Table

Results Table

Final output of the detection and segmentation pipeline for MaskRCNN and Domain Adaptation

Mask R-CNN losses

Domain Adaptation models

Results Table for Domain Adaptation

Domain Adaptation losses

Full pipeline

JSON outputs which can be used for automated annotation of new slides (Future work)

If you find this work useful, please consider citing

@article{PAL2024123660,
title = {Advancing instance segmentation and WBC classification in peripheral blood smear through domain adaptation: A study on PBC and the novel RV-PBS datasets},
journal = {Expert Systems with Applications},
volume = {249},
pages = {123660},
year = {2024},
issn = {0957-4174},
doi = {https://doi.org/10.1016/j.eswa.2024.123660},
url = {https://www.sciencedirect.com/science/article/pii/S0957417424005268},
author = {Jimut Bahan Pal and Aniket Bhattacharyea and Debasis Banerjee and Br. Tamal Maharaj},
keywords = {Automated blood test, Detection, Domain adaptation, Instance segmentation, Peripheral blood smear}
}