Peter Horvath's Portfolio

Publications | Research interest |  About   |   Home

High content analysis Advanced Cell Classifier - ACC is a data analyser program to evaluate cell-based high-content screens. The basic aim is to provide a very accurate analysis with minimal user interaction using advanced machine learning methods. Genome-scale high content screen analysis - The analysis of the first few Swiss academic whole human genome-wide siRNA screens. The analysis involves quality control, data management, image analysis, feature extraction, cell classification using machine learning, statistical analysis (quality control), and hit detection. Illumination correction of HCS images - My team is interested in developing methods to estimate most accurately vignetting and background functions and correct illumination problems. Our main goal is to estimate the functions via the images and not to use reference objects. Ph.D. project of Filippo Piccinini. Segmentation and identification of co-cultured cells - We developed a method to detect and distinguish different cell-types in co-cultures. The method is based on a novel spot detection algorithm and classification of the detected spots. Machine learning methods for high content screening (life sciences) - These projects develop methods and software to answer the following questions: Can one suggest the "best" machine learning method for a given biological data set? Can these methods be generally good for other biological data? To what extent the human factor influences the final decisions? How can we introduce novel techniques such as regression, semi-supervised learning, active learning? PosDoc project of Kevin Smith. Statistical methods for screening - We are mainly interested in two fields. Firstly, the normalization and correction of screening data to eliminate biological and liquid handling effects (eg. plate effects, cell number differences, ...). Secondly, to generate reliable and biologically meaningful hit list for replicate experiments or those containing multiple oligo sequences.   Microscopic image analysis Semi-automatic contour tracking - We develope a semi-automated program using active contours to track ring-like objects (osteoclast cells). The program also performs temporal, morphological, intensity-, and texture-based statistics. Tracking cells on phase contrast images - CellTracker is a program to perform automated and manual cell migration detection. Main goals are: automated image quality enhancement using background subtraction and alignment correction; automated detection and tracking of cells; manual tracking and editing cell paths. Master's thesis of Martin Maag. Cell segmentation and splitting in 3D - We developed methods to split blobs in 3D using gradient vector flows. Master's thesis of Christoph Faigle. Statistical analysis of non-sister kinetochore's motion coupling - We are interested in the organization of the metaphase plate and the statistical correlation analysis of non-sister kinetochore migration. Master's thesis of Katalin Virag. Fundamental image analysis Gas of Circles Active contour model - Circular object segmentation. This model is a tool to describe a set of circles with an approximately fixed radius. The model is based on the higher-order active contour (HOAC) framework. For certain ranges of the parameters, the model creates stable circles with an approximately fixed radius. The images we use are color-infrared (CIR) and panchromatic aerial images. Experiments show that the models outperform other traditional circle detection methods. The model can also be applied to the detection of other circular objects, e.g. in cell biology, nanotechnology, medical imaging, satellite images. (INRIA Sophia Antipolis, University of Szeged). 3D tracking on 2D images (BacTrack) - We developed a model to perform 3D tracking on microscopic image sequences using the Point Spread Function of the acquisition system. Master's thesis of Qasim Bukhari.