Objective: To present the potential application of an automated image analysis pipeline called WSIQC for the cleaning and quality assessment of brain histology images.

Background: Whole Slide Images (WSI) from brain tissue can be used to better understand the pathology of Parkinson’s disease (PD) and to develop artificial intelligence-based assessment models. However, this is possible only with high-quality and artefact-free WSI. Despite careful preparation and cleaning of glass slides prior to image scanning, there are always some residual artefacts. WSIQC [1] is an image analysis pipeline, implemented on H&E breast tissue images, which can assess the quality of WSI and clean them in an automated way by removing different artefacts. The successful application of this pipeline on brain tissue images could drastically reduce the time and resources required to assess images captured from glass slides without the need for extensive computational resources. The aim of this study is to evaluate the performance of WSIQC on brain tissue images compared to breast tissue images, assess whether the pipeline can be used as an effective automated cleaning approach for brain images and use it to evaluate the WSI quality of images produced at the Parkinson’s UK Tissue Bank at Imperial.

Method: The default WSIQC pipeline was tested on 674 PD, 570 non-neurological controls and 57 Multiple Sclerosis (MS) brain tissue H&E histology images from the Parkinson’s UK Tissue Bank at Imperial, and 73 breast tissue images from a non-curated publicly available database [2,3]. WSIQC performance was compared between brain and breast tissue images. The quality of the datasets was compared according the percentage of artefacts and the number of high-quality tiles.

Results: WSIQC fully detected and removed contrast issues, coverslip edges, dust, dye and ink of different colours, fungi, bubbles, and areas with broken glass from brain tissue images, with a performance equivalent to breast tissues. The pipeline was also able to identify images incorrectly labelled as being stained with H&E, and areas with demyelination or low staining. Among the datasets studied, the PD and control data had the best quality, followed by MS and breast tissues.

Conclusion: Our results confirm that WSIQC can be used as an effective, automated, approach to clean and assess the quality of brain tissue images.

References: [1] Giunchiglia, V., Takats, Z., and McKenzie J. “WSIQC: whole slide image pre-processing pipeline for artifact removal and quality control” (in preparation).
[2] http://histoqcrepo.com/
[3] Janowczyk, Andrew, et al. “HistoQC: an open-source quality control tool for digital pathology slides.” JCO clinical cancer informatics 3 (2019): 1-7

« Back to 2022 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/an-automated-data-cleaning-approach-to-remove-preparation-artefacts-from-brain-histology-slide-images/