Research

My research interests are focused on images and how to extract information from them. This includes :

• computer vision and image analysis
• statistics
• technologies to acquire images: optical sensors, scanners, IRMs, mass-spectrometry, etc.
• physics, chemistry and biology for the interpretation of image features.

In all of these, there are models that describe the objects of interest and we need to assess whether the data conforms to the models.

More information on these topics is available below, or on dedicated sections in the left menu.

Recently, I have also become interested in "Reproducible Research" and how to preserve computational research results and methods in the long term. More information in the Lepton section.

Computer vision

The human vision system is extremely efficient at processing visual information: extracting features, delimiting objects regardless of textures, inferring object shape, etc. Computer vision tries to analyse how a computer can perform the same tasks.

I am interested in low-level processing of visual information such as extracting segments, contours or clusters in an image. I use a contrario detection, i.e. I assume that the human visual system sees image features as exceptional with regards to a model of the background.

Quantitative image analysis

In many applications, data recorded by instruments during experiments are images or are structured as images, i.e. maps from a set with spatial coordinates to a set with intensity coordinates.

Quantitative in this context means that the objective is not only to detect interesting features in the image, but also to

• estimate their characteristics, such as volume for cancerous tissue,
• estimate the confidence level and the precision level.

For example, I have experience in:

• mass-spectrometry based proteomics. This means analysing biological samples containing thousands of proteins to identify the components and measure their concentration. The standard experiment is to use 2D chromatography gels, but the current trend is towards the use of liquid-chromatography / mass-spectrometry instruments.
• next-generation DNA sequencing. High throughput sequencing technologies measure light signals when nucleic acids are bound to a probe.
• 3D medical imaging. I am currently investigating the combination of scanner and PET images of the same subject.

Reproducible research

Many details are often omitted in current scientific publications. As a result, it is a challenge to reproduce the methods and results of many interesting papers. A growing trend in computational science is to design tools and frameworks to ensure that published results can be reproduced by a third-party / interested reader or reviewer. I propose a tool called Lepton.

Sociology

The real world is a source of many interesting problems. I am currently contributing to the statistical analysis of a population survey with fellow sociologists. The group of people in the study are "young children educators" (éducateur de jeunes enfants). In France, this corresponds to a specific diploma and specific job activities focused on education rather than day care or animation.