Search

International audience

Segmentation multimodale de tumeurs pulmonaires et de ganglions lymphatiques par une méthode variationnelle

International audience; Compressed Sensing (CS) provides a new framework for signal sampling, exploiting redundancy and sparsity in incoherent bases. For images with homogeneous objects and background, CS provides an optimal reconstruction framework from a set of random projections in the Fourier domain, while constraining bounded variations in the spatial domain. In this paper, we propose a CS-based method to simultaneously acquire and denoise data based on statistical properties of the CS optimality, signal modeling and characteristics of noise reconstruction. Our approach has several advantages over traditional denoising methods, since it can under-sample, recover and denoise images simultaneously. We demonstrate with simulated and practical experiments on fluorescence images that we obtain images with similar or increased SNR even with reduced exposure times. Such results open the gate to new mathematical imaging protocols, offering the opportunity to reduce exposure time along with photo-toxicity and photo-bleaching and assist biological applications relying on fluorescence microscopy.

With relatively high frame rates and the ability to acquire volume data sets with a stationary transducer, 3D ultrasound systems, based on matrix phased array transducers, provide valuable three-dimensional information, from which quantitative measures of cardiac function can be extracted. Such analyses require segmentation and visual tracking of the left ventricular endocardial border. Due to the large size of the volumetric data sets, manual tracing of the endocardial border is tedious and impractical for clinical applications. Therefore the development of automatic methods for tracking three-dimensional endocardial motion is essential. In this study, we evaluate a four-dimensional optical flow motion tracking algorithm to determine its capability to follow the endocardial border in three dimensional ultrasound data through time. The four-dimensional optical flow method was implemented using three-dimensional correlation. We tested the algorithm on an experimental open-chest dog data set and a clinical data set acquired with a Philips' iE33 three-dimensional ultrasound machine. Initialized with left ventricular endocardial data points obtained from manual tracing at end-diastole, the algorithm automatically tracked these points frame by frame through the whole cardiac cycle. Finite element surfaces were fitted through the data points obtained by both optical flow tracking and manual tracing by an experienced observer for quantitative comparison of the results. Parameterization of the finite element surfaces was performed and maps displaying relative differences between the manual and semi-automatic methods were compared. The results showed good consistency with less than 10% difference between manual tracing and optical flow estimation on 73% of the entire surface. In addition, the optical flow motion tracking algorithm greatly reduced processing time (about 94% reduction compared to human involvement per cardiac cycle) for analyzing cardiac function in three-dimensional ultrasound data sets. A displacement field was computed from the optical flow output, and a framework for computation of dynamic cardiac information is introduced. The method was applied to a clinical data set from a heart transplant patient and dynamic measurements agreed with known physiology as well as experimental results.

We propose a Hybrid Segmentation Engine that consists of component modules, for automated segmentation of radiological patient and the Visible Human data. We integrate boundary-based and region-based segmentation methods to exploit the strength of each method hopefully to cover the weakness of the other method.

This paper presents new results for segmentation of 3D ultrasound using a robust and smooth segmentation method based on a homogeneity-driven three dimensional level set algorithm. The segmentation was applied to echocardiographic data from healthy volunteers acquired with two 3D ultrasound machines based on matrix-phased array technology. A comparison of ventricular volumes and geometry is performed based on manual and automatic methods of segmentation. Results showed good agreement of the segmentation methods and the ultrasound machines for ventricular geometry and quantitative assessment of cardiac function.

This project is part of the Northeast Structural Genomics Consortium (NESG). The goal of this consortium is to develop efficient and integrated technologies for high-throughput (HTP) protein production and 3D structure determination. This project focuses on the design of an image analysis system to classify protein crystal structures in a production oriented environment. The method presented performs classification of microscopic images as protein precipitates or crystals versus clear solutions. Preliminary experiments are presented, which provided high classification accuracy on large datasets of protein crystallization experiments using expert classification for ground truth.

This project is part of the Northeast Structural Genomics Consortium (NESG). The goal of this consortium is to develop efficient and integrated technologies for high-throughput (HTP) protein production and 3D structure determination. This project focuses on the design of an image analysis system to classify protein crystal structures in a production oriented environment. The method performs classification of microscopic images as clear droplets versus non-clear droplets (precipitates and crystals). Using expert classification for ground truth, current results show high classification accuracy with a large image datasets.