Time filter

Source Type

Vienna, Austria

Heindl A.,Medical University of Vienna | Seewald A.K.,Seewald Solutions | Thalhammer T.,Medical University of Vienna | Bises G.,Medical University of Vienna | And 7 more authors.
Cytometry Part A

Automated microscopic image analysis of immunofluorescence-stained targets on tissue sections is challenged by autofluorescent elements such as erythrocytes, which might interfere with target segmentation and quantification. Therefore, we developed an automated system (Automated REcognition of Tissue-associated Erythrocytes; ARETE) for in silico exclusion of erythrocytes. To detect erythrocytes in transmission images, a cascade of boosted decision trees of Haar-like features was trained on 8,640/4,000 areas (15 × 15 pixels) with/without erythrocytes from images of placental sections (4 μm). Ground truth data were generated on 28 transmission images. At least two human experts labelled the area covered by erythrocytes. For validation, output masks of human experts and ARETE were compared pixel-wise against a mask obtained from majority voting of human experts. F1 score, specificity, and Cohen's κ coefficients were calculated. To study the influence of erythrocyte-derived autofluorescence, we investigated the expression levels of a protein (receptor for advanced glycated end products; RAGE) in placenta and number of Ki-67-positive/cytokeratin 8-positive epithelial cells in colon sections. ARETE exhibited high sensitivity (99.87%) and specificity (99.81%) on a training-subset and processed transmission images (1,392 × 1,024 pixels) within 4 sec. ARETE and human expert's F1-scores were 0.55 versus 0.76, specificities 0.85 versus 0.92 and Cohen's κ coefficients 0.41 versus 0.68. A ranking of Cohen's κ coefficient by the scale of Fleiss certified "good agreement" between ARETE and the human experts. Applying ARETE, we demonstrated 4-14% false-positive RAGE-expression in placenta, and 18% falsely detected proliferative epithelial cells in colon, caused by erythrocyte-autofluorescence. ARETE is a fast system for in silico reduction of erythrocytes, which improves automated image analysis in research and diagnostic pathology. © 2013 International Society for Advancement of Cytometry. Source

Seewald A.K.,Seewald Solutions | Gansterer W.N.,University of Vienna
Computers and Security

We develop and discuss automated and self-adaptive systems for detecting and classifying botnets based on machine learning techniques and integration of human expertise. The proposed concept is purely passive and is based on analyzing information collected at three levels: (i) the payload of single packets received, (ii) observed access patterns to a darknet at the level of network traffic, and (iii) observed contents of TCP/IP traffic at the protocol level. We illustrate experiments based on real-life data collected with a darknet set up for this purpose to show the potential of the proposed concept for Levels (i) and (ii). As darknets cannot capture TCP/IP traffic data, we use a small spamtrap in our experiments at Level (iii). Strictly speaking, this approach for Level (iii) is not purely passive. However, traffic moving through a network could potentially be analyzed in a similar way to also obtain a purely passive system at this level. © 2009 Elsevier Ltd. All rights reserved. Source

Here we present a constrained object recognition task that has been robustly solved largely with simple machine learning methods, using a small corpus of about 100 images taken under a variety of lighting conditions. The task was to analyze images from a hand-held mobile phone camera showing an endgame position for the Japanese board game Go. The presented system would already be sufficient to reconstruct the full Go game record from a video record of the game and thus is complementary to Seewald (2003), which focuses on solving the same task using different sensors. The presented system is robust to a variety of lighting conditions, works with cheap low-quality cameras, and is resistant to changes in board or camera position without the need for any manual calibration. Source

Rogojanu R.,Medical University of Vienna | Thalhammer T.,Medical University of Vienna | Thiem U.,Medical University of Vienna | Heindl A.,Medical University of Vienna | And 6 more authors.
BioMed Research International

In colorectal cancer (CRC), an increase in the stromal (S) area with the reduction of the epithelial (E) parts has been suggested as an indication of tumor progression. Therefore, an automated image method capable of discriminating E and S areas would allow an improved diagnosis. Immunofluorescence staining was performed on paraffin-embedded sections from colorectal tumors (16 samples from patients with liver metastasis and 18 without). Noncancerous tumor adjacent mucosa (n=5) and normal mucosa (n=4) were taken as controls. Epithelial cells were identified by an anti-keratin 8 (K8) antibody. Large tissue areas (5-63 mm2/slide) including tumor center, tumor front, and adjacent mucosa were scanned using an automated microscopy system (TissueFAXS). With our newly developed algorithms, we showed that there is more K8-immunoreactive E in the tumor center than in tumor adjacent and normal mucosa. Comparing patients with and without metastasis, the E/S ratio decreased by 20% in the tumor center and by 40% at tumor front in metastatic samples. The reduction of E might be due to a more aggressive phenotype in metastasis patients. The novel software allowed a detailed morphometric analysis of cancer tissue compartments as tools for objective quantitative measurements, reduced analysis time, and increased reproducibility of the data. © 2015 R. Rogojanu et al. Source

Heindl A.,Medical University of Vienna | Dekan S.,Medical University of Vienna | Ellinger I.,Medical University of Vienna | Seewald A.K.,Seewald Solutions
2010 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC'10

Analyzing in-situ tissue structures with complex shapes and textures such as multinuclear cells or cells without nuclei is still a challenge for currently available imageprocessing software. This work aims to provide a versatile system to solve such tasks provided that the structures of interests were detected by immunofluorescence microscopy. Images were automatically acquired using slide-based microscopy. Human domain-experts manually marked up tissue samples to evaluate the performance of the computer generated masks. From precision and recall a balanced F-score was computed to measure the correlation between experts and algorithm output. Exhaustive parameter optimization was conducted to ensure that the optimal input parameters were applied during evaluation of the developed algorithm. This procedure significantly increased the performance compared to manually chosen input parameters. We present an approach that can handle huge tissue areas and does not rely on nuclei detection. Once a markup has been created, the algorithm can be parameter-optimized on ground-truth data for the chosen tissue sample. Thereafter, the resulting settings could be applied automatically to the respective stitched image. Concluding, we provide new insights in physiological and pathopysiological cellular mechanisms by automating the in-situ analysis of proteins in intact tissues. © 2010 IEEE. Source

Discover hidden collaborations