Cell Detection for Asthma on Partially Annotated Whole Slide Images

Abstract

Asthma is a chronic inflammatory disorder of the lower respiratory tract and naturally occurs in humans and animals including horses. The annotation of an asthma microscopy whole slide image (WSI) is an extremely labour-intensive task due to the hundreds of thousands of cells per WSI. To overcome the limitation of annotating WSI incompletely, we developed a training pipeline which can train a deep learningbased object detection model with partially annotated WSIs and compensate class imbalances on the fly. With this approach we can freely sample from annotated WSIs areas and are not restricted to fully annotated extracted sub-images of the WSI as with classical approaches. We evaluated our pipeline in a cross-validation setup with a fixed training set using a dataset of six equine WSIs of which four are partially annotated and used for training, and two fully annotated WSI are used for validation and testing. Our WSI-based training approach outperformed classical sub-image-based training methods by up to 15% mAP and yielded human-like performance when compared to the annotations of ten trained pathologists.

First steps on Gamification of Lung Fluid Cells Annotations in the Flower Domain

“…The lung fluid cell data set consists of six cytological samples of equine bronchoalveolar lavage fluid (BALF), which were cytocentrifugated, and stained using May Grunwald Giemsa. This data set was provided by Marzahl et al [3]. Afterwards, the glass slides were digitized using a linear scanner (Aperio ScanScope CS2, Leica Biosystems, Germany) at a magnification of 400x with a resolution of 0.25 µm/px [3].…”

“…This data set was provided by Marzahl et al [3]. Afterwards, the glass slides were digitized using a linear scanner (Aperio ScanScope CS2, Leica Biosystems, Germany) at a magnification of 400x with a resolution of 0.25 µm/px [3]. Two of the six WSIs are fully annotated and four partially annotated by a trained pathologist.…”

“…Two of the six WSIs are fully annotated and four partially annotated by a trained pathologist. The 87,738 by bounding box annotated cells are distributed unbalanced between neutrophils, multinuclear, mast cells, macrophages, lymphocytes, erythrocyte and eosinophile [3]. Example images for the individual cell types can be seen in the top row of Fig.…”

“…One platform enabling crowdsourcing is the EXACT online annotation tool [2]. Among other data sets it hosts a data set with partly annotated lung fluid cells [3], which were localized using the RetinaNet architecture. However, a large portion of the cell types are still unlabeled of annotations.…”

X‐ray microscopy and automatic detection of defects in through silicon vias in three‐dimensional integrated circuits

“…The RetinaNet architecture is often used in microscopy image analysis 41‐43 because it provides a decent trade‐off between performance and complexity and proved to be applicable in class imbalance conditions 42,44 . Consequently, as this study encounters large volumetric XRM data and class imbalance conditions, a RetinaNet is used for the detection and classification of TSVs.…”