Anne Grote

On-Chip Imaging of Schistosoma haematobium Eggs in Urine for Diagnosis by Computer Vision

Background Microscopy, being relatively easy to perform at low cost, is the universal diagnostic method for detection of most globally important parasitic infections. As quality control is hard to maintain, misdiagnosis is common, which affects both estimates of parasite burdens and patient care. Novel techniques for high-resolution imaging and image transfer over data networks may offer solutions to these problems through provision of education, quality assurance and diagnostics. Imaging can be done directly on image sensor chips, a technique possible to exploit commercially for the development of inexpensive “mini-microscopes”. Images can be transferred for analysis both visually and by computer vision both at point-of-care and at remote locations. Methods/Principal Findings Here we describe imaging of helminth eggs using mini-microscopes constructed from webcams and mobile phone cameras. The results show that an inexpensive webcam, stripped off its optics to allow direct application of the test sample on the exposed surface of the sensor, yields images of Schistosoma haematobium eggs, which can be identified visually. Using a highly specific image pattern recognition algorithm, 4 out of 5 eggs observed visually could be identified. Conclusions/Significance As proof of concept we show that an inexpensive imaging device, such as a webcam, may be easily modified into a microscope, for the detection of helminth eggs based on on-chip imaging. Furthermore, algorithms for helminth egg detection by machine vision can be generated for automated diagnostics. The results can be exploited for constructing simple imaging devices for low-cost diagnostics of urogenital schistosomiasis and other neglected tropical infectious diseases.

Detection of lobular structures in normal breast tissue

BACKGROUND Ongoing research into inflammatory conditions raises an increasing need to evaluate immune cells in histological sections in biologically relevant regions of interest (ROIs). Herein, we compare different approaches to automatically detect lobular structures in human normal breast tissue in digitized whole slide images (WSIs). This automation is required to perform objective and consistent quantitative studies on large data sets. METHODS In normal breast tissue from nine healthy patients immunohistochemically stained for different markers, we evaluated and compared three different image analysis methods to automatically detect lobular structures in WSIs: (1) a bottom-up approach using the cell-based data for subsequent tissue level classification, (2) a top-down method starting with texture classification at tissue level analysis of cell densities in specific ROIs, and (3) a direct texture classification using deep learning technology. RESULTS All three methods result in comparable overall quality allowing automated detection of lobular structures with minor advantage in sensitivity (approach 3), specificity (approach 2), or processing time (approach 1). Combining the outputs of the approaches further improved the precision. CONCLUSIONS Different approaches of automated ROI detection are feasible and should be selected according to the individual needs of biomarker research. Additionally, detected ROIs could be used as a basis for quantification of immune infiltration in lobular structures.

Assessment of tumour viability in human lung cancer xenografts with texture-based image analysis

Aims To build and evaluate an automated method for assessing tumour viability in histological tissue samples using texture features and supervised learning. Methods H&E-stained sections (n=56) of human non-small cell lung adenocarcinoma xenografts were digitised with a whole-slide scanner. A novel image analysis method based on local binary patterns and a support vector machine classifier was trained with a set of sample regions (n=177) extracted from the whole-slide images and tested with another set of images (n=494). The extracted regions, or single-tissue entity images, were chosen to represent as pure as possible examples of three morphological tissue entities: viable tumour tissue, non-viable tumour tissue and mouse host tissue. Results An agreement of 94.5% (area under the curve=0.995, kappa=0.90) was achieved to classify the single-tissue entity images in the test set (n=494) into the viable tumour and non-viable tumour tissue categories. The algorithm assigned 250 of the 252 non-viable and 219 of the 242 of viable sample regions to the correct categories, respectively. This corresponds to a sensitivity of 90.5% and specificity of 99.2%. Conclusions The proposed image analysis-based tumour viability assessment resulted in a high agreement with expert annotations. By providing extraction of detailed information of the tumour microenvironment, the automated method can be used in preclinical research settings. The method could also have implications in cancer diagnostics, cancer outcome prognostics and prediction.

Exploring the spatial dimension of estrogen and progesterone signaling: detection of nuclear labeling in lobular epithelial cells in normal mammary glands adjacent to breast cancer

BackgroundComprehensive spatial assessment of hormone receptor immunohistochemistry staining in digital whole slide images of breast cancer requires accurate detection of positive nuclei within biologically relevant regions of interest. Herein, we propose a combination of automated region labeling at low resolution and subsequent detailed tissue evaluation of subcellular structures in lobular structures adjacent to breast cancer, as a proof of concept for the approach to analyze estrogen and progesterone receptor expression in the spatial context of surrounding tissue.MethodsRoutinely processed paraffin sections of hormone receptor-negative ductal invasive breast cancer were stained for estrogen and progesterone receptor by immunohistochemistry. Digital whole slides were analyzed using commercially available image analysis software for advanced object-based analysis, applying textural, relational, and geometrical features. Mammary gland lobules were targeted as regions of interest for analysis at subcellular level in relation to their distance from coherent tumor as neighboring relevant tissue compartment. Lobule detection quality was evaluated visually by a pathologist.ResultsAfter rule set optimization in an estrogen receptor-stained training set, independent test sets (progesterone and estrogen receptor) showed acceptable detection quality in 33% of cases. Presence of disrupted lobular structures, either by brisk inflammatory infiltrate, or diffuse tumor infiltration, was common in cases with lower detection accuracy. Hormone receptor detection tended towards higher percentage of positively stained nuclei in lobules distant from the tumor border as compared to areas adjacent to the tumor. After adaptations of image analysis, corresponding evaluations were also feasible in hormone receptor positive breast cancer, with some limitations of automated separation of mammary epithelial cells from hormone receptor-positive tumor cells.ConclusionsAs a proof of concept for object-oriented detection of steroid hormone receptors in their spatial context, we show that lobular structures can be classified based on texture-based image features, unless brisk inflammatory infiltration disrupts the normal morphological structure of the tubular gland epithelium. We consider this approach as prototypic for detection and spatial analysis of nuclear markers in defined regions of interest. We conclude that advanced image analysis at this level of complexity requires adaptation to the individual tumor phenotypes and morphological characteristics of the tumor environment.

Context-Based Normalization of Histological Stains Using Deep Convolutional Features

While human observers are able to cope with variations in color and appearance of histological stains, digital pathology algorithms commonly require a well-normalized setting to achieve peak performance, especially when a limited amount of labeled data is available. This work provides a fully automated, end-to-end learning-based setup for normalizing histological stains, which considers the texture context of the tissue. We introduce Feature Aware Normalization, which extends the framework of batch normalization in combination with gating elements from Long Short-Term Memory units for normalization among different spatial regions of interest. By incorporating a pretrained deep neural network as a feature extractor steering a pixelwise processing pipeline, we achieve excellent normalization results and ensure a consistent representation of color and texture. The evaluation comprises a comparison of color histogram deviations, structural similarity and measures the color volume obtained by the different methods.

Abstract 4815: Humanized single mouse trial: A preclinical platform feasible for immune-oncology drug screening and translational biomarker development

The field of cancer immunology is rapidly moving towards innovative therapeutic strategies. As a consequence the need for robust and predictive preclinical platforms arises just as well. The current project aims to establish a drug screening workflow bridging between innovative mouse models and clinical biomarker development. A total of 69 NOG (NOD/Shi-scid/IL-2Rγnull) mice were engrafted with CD34+ hematopoietic stem cells. Thereafter, tumor material from 11 different lung cancer patient derived xenograft models (NSCLC PDX) was implanted subcutaneously. Individual mice were treated with α-CTLA-4, α-PD-1 or the combination thereof. With n=1 per treatment arm and model the study design followed the screening approach of the single mouse trial (SMT). Infiltration of human immune cells was detected by flow cytometry (FC) and immunohistochemistry (IHC) in hematopoietic organs and tumor tissue. A computerized analysis for digitized whole-slide images of the samples was used to quantify the lymphocyte infiltration using color classification and morphological image processing techniques. All 3 treatment arms displayed a discrete activity pattern throughout the PDX panel. Tumor models with high tumor infiltrating lymphocyte (TIL) rates in the donor patient material tended to be more sensitive towards checkpoint inhibitor treatment as models with low rates. Numbers of TILs in the PDX detected by FC and IHC were significantly increased in the treatment groups as compared to control vehicle. In parallel, hematopoietic organs showed high (>25%) amounts of huCD45 cells in all groups and models. PDX models being sensitive towards checkpoint inhibitor treatment (responders) displayed a higher percentage of DAB+ nuclei in huCD45 IHC stains than non-responder models as determined by image analysis. Irrespective thereof, in responders as well as non-responders the treatment with checkpoint inhibitors enhanced the percentage of DAB+ nuclei. Whole-slide image analysis of the HE 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 4815. doi:10.1158/1538-7445.AM2017-4815

Image analysis of immune cell patterns in the human mammary gland during the menstrual cycle refines lymphocytic lobulitis

PurposeTo improve microscopic evaluation of immune cells relevant in breast cancer oncoimmunology, we aim at distinguishing normal infiltration patterns from lymphocytic lobulitis by advanced image analysis. We consider potential immune cell variations due to the menstrual cycle and oral contraceptives in non-neoplastic mammary gland tissue.MethodsLymphocyte and macrophage distributions were analyzed in the anatomical context of the resting mammary gland in immunohistochemically stained digital whole slide images obtained from 53 reduction mammoplasty specimens. Our image analysis workflow included automated regions of interest detection, immune cell recognition, and co-registration of regions of interest.ResultsIn normal lobular epithelium, seven CD8

Abstract 1698: Systems pathology for characterization of cancer model systems in a multicenter IMI-PREDECT project

^{+}