Christophe Tilmant

Nonsupervised Ranking of Different Segmentation Approaches: Application to the Estimation of the Left Ventricular Ejection Fraction From Cardiac Cine MRI Sequences

A statistical methodology is proposed to rank several estimation methods of a relevant clinical parameter when no gold standard is available. Based on a regression without truth method, the proposed approach was applied to rank eight methods without using any a priori information regarding the reliability of each method and its degree of automation. It was only based on a prior concerning the statistical distribution of the parameter of interest in the database. The ranking of the methods relies on figures of merit derived from the regression and computed using a bootstrap process. The methodology was applied to the estimation of the left ventricular ejection fraction derived from cardiac magnetic resonance images segmented using eight approaches with different degrees of automation: three segmentations were entirely manually performed and the others were variously automated. The ranking of methods was consistent with the expected performance of the estimation methods: the most accurate estimates of the ejection fraction were obtained using manual segmentations. The robustness of the ranking was demonstrated when at least three methods were compared. These results suggest that the proposed statistical approach might be helpful to assess the performance of estimation methods on clinical data for which no gold standard is available.

Background subtraction adapted to PTZ cameras by keypoint density estimation.

Constant Guillot1 constant.guillot@cea.fr Maxime Taron1 maxime.taron@cea.fr Patrick Sayd1 patrick.sayd@cea.fr Quoc-Cuong Pham1 quoc-cuong.pham@cea.fr Christophe Tilmant2 christophe.tilmant@lasmea.univ-bpclermont.fr Jean-Marc Lavest2 J-Marc.Lavest@u-clermont1.fr 1 CEA LIST Laboratoire Vision et Ingenierie des Contenus, BP 94, Gif-sur-Yvette, F-91191 France 2 LASMEA UMR 6602, PRES Clermont Universite/CNRS, 63177 Aubiere cedex, France

Camera-to-camera mapping for hybrid pan-tilt-zoom sensors calibration

Video surveillance becomes more and more extended in industry and often involves automatic calibration system to remain efficient. In this paper, a video-surveillance system that uses stationary-dynamic cameras devices is presented. The static camera is used to monitor a global scene. When it detects a moving object, the Pan-Tilt-Zoom (PTZ) camera is controlled to be centered on this object. We describe a method of camera-to-camera calibration, integrating zoom calibration in order to command the angles and the zoom of the PTZ camera. This method enables to take into account the intrinsic camera parameters, the 3D scene geometry and the fact that the mechanism of inexpensive camera does not fit the classical geometrical model. Finally, some experiment results attest the accuracy of the proposed solution.

Measuring the size of neoplasia in colonoscopy using depth-from-defocus

Colonoscopy is the reference medical examination for the diagnosis and treatment of neoplasia in gastroenterology. During the examination, the expert explores the colon cavity with a gastroscope in order to detect neoplasias - abnormal growths of tissue - and to diagnose which ones could be malignant. The Paris classification of superficial neoplastic lesions is the gold standard set of criteria for this type of diagnosis. One of the major criteria is the size. However, this is tremendously difficult to accurately estimate from images. This is because the absolute scale of the observed tissues is not directly conveyed in the 2D endoscopic image. We propose an image-based method to estimate the size of neoplasias. The core idea is to combine Depth-From-Focus (DFF) and Depth-From-Defocus (DFD). This allows us to recover the absolute scale by automatically detecting the blur/unblur breakpoint while the expert pulls the gastroscope away from a neoplasia. Our method is passive: it uses the image data only and thus does not require hardware modification of the gastroscope. We report promising experimental results on phantom and patient datasets.

Improved Estimation of Cardiac Function Parameters Using a Combination of Independent Automated Segmentation Results in Cardiovascular Magnetic Resonance Imaging.

This work aimed at combining different segmentation approaches to produce a robust and accurate segmentation result. Three to five segmentation results of the left ventricle were combined using the STAPLE algorithm and the reliability of the resulting segmentation was evaluated in comparison with the result of each individual segmentation method. This comparison was performed using a supervised approach based on a reference method. Then, we used an unsupervised statistical evaluation, the extended Regression Without Truth (eRWT) that ranks different methods according to their accuracy in estimating a specific biomarker in a population. The segmentation accuracy was evaluated by estimating six cardiac function parameters resulting from the left ventricle contour delineation using a public cardiac cine MRI database. Eight different segmentation methods, including three expert delineations and five automated methods, were considered, and sixteen combinations of the automated methods using STAPLE were investigated. The supervised and unsupervised evaluations demonstrated that in most cases, STAPLE results provided better estimates than individual automated segmentation methods. Overall, combining different automated segmentation methods improved the reliability of the segmentation result compared to that obtained using an individual method and could achieve the accuracy of an expert.

Using the Infocus-Breakpoint to estimate the scale of neoplasia in colonoscopy

Colonoscopy is the reference medical examination for early diagnosis and treatment of colonic diseases. This minimally invasive technique allows gastroenterologists to explore the colon and remove neoplasias - abnormal growth of tissue - such as polyps which may transform into malignant tumors. Shape, texture and size of polyps are of particular interest for determining their nature. However, the size of neoplasias is difficult to estimate because the absolute scale of the observed tissue is not directly conveyed in the 2D endoscopic images. We here improve our Infocus-Breakpoint (IB) technique, which estimates an imagewise scale by detecting the blur/ unblur breakpoint in a video sequence. We simultaneously track a neoplasia with a 2D affine transformation and estimate the amount of defocus blur. This leads to an areawise scale estimate with better accuracy.

How big is this neoplasia? Live colonoscopic size measurement using the Infocus-Breakpoint

Colonoscopy is the reference medical examination for early diagnosis and treatment of colonic diseases. This minimally invasive technique allows endoscopists to explore the colon cavity and remove neoplasias - abnormal growths of tissue - which may develop into malignant tumors. The size, shape and appearance of a neoplasia are essential cues for diagnostic. However, the size is difficult to estimate because the absolute scale of the observed tissue is not directly conveyed in the 2D colonoscopic images. An erroneous size estimate may lead to inappropriate treatment. There currently exist no solutions to reproducible neoplasia size measurement adapted to colonoscopy. We propose a colonoscopic size measurement system for neoplasias. By using a simple planar geometry, the key technical problem is reduced to resolving scale. Our core contribution is introducing the Infocus-Breakpoint (IB) that allows us to resolve scale from a regular colonoscopic video. We define the IB as the lower limit of the colonoscopes depth of field. The IB corresponds to a precise colonoscope to tissue distance, called the reference depth, which we calibrate preoperatively. We detect the IB intraoperatively thanks to two novel modules: deformable Blur-Estimating Tracking (BET) and Blur-Model Fitting (BMF). With our system, the endoscopist may interactively measure the length and area of a neoplasia in a 2D colonoscopic image directly. Our system needs no hardware modification to standard monocular colonoscopes, yet reaching a size measurement accuracy of the order of a millimeter, as shown on several phantom and patient datasets.

Modular Ensemble Tracking

Object Tracking is a very important domain in computer vision. It was recently approached using classification techniques and still more recently using boosting methods. Boosting is a general method of producing an accurate prediction rule by combining rough and moderately inaccurate ones. We introduce in this paper a modular object tracking algorithm based on one of these boosting methods: Adaboost. Tracking is performed on homogeneous feature spaces and the final classification decision is obtained by combining the decisions made on each of these spaces. A classifier update stage is also introduced, that allows the method both to handle time-varying objects in real-time (using fast computable features) and to handle partial occlusions. We compare the performance of our algorithm with Ensemble Tracking algorithm [2] on several real video sequences.

Background subtraction for PTZ cameras performing a guard tour and application to cameras with very low frame rate

Pan Tilt Zoom cameras have the ability to cover wide areas with an adapted resolution. Since the logical downside of high resolution is a limited field of view, a guard tour can be used to monitor a large scene of interest. However, this greatly increases the duration between frames associated to a specific location. This constraint makes most background algorithms ineffective. In this article we propose a background subtraction algorithm suitable to cameras with very low frame rate. Its main interest consists in the resulting robustness to sudden illumination changes. The background model which describes a wide scene of interest consisting of a collection of images can thus be successfully maintained. This algorithm is compared with the state of the art and a discussion regarding its properties follows.

A mutual reference shape based on information theory

In this paper, we consider the estimation of a reference shape from a set of different segmentation results using both active contours and information theory. The reference shape is defined as the minimum of a criterion that benefits from both the mutual information and the joint entropy of the input segmentations and is then called a mutual shape. This energy criterion is here justified using similarities between information theory quantities and area measures, and presented in a continuous variational framework. This framework brings out some interesting evaluation measures such as the specificity and sensitivity. In order to solve this shape optimization problem, shape derivatives are computed for each term of the criterion and interpreted as an evolution equation of an active contour. Some synthetical examples allow us to cast the light on the difference between our mutual shape and an average shape. Our framework has been considered for the estimation of a mutual shape for the evaluation of cardiac segmentation methods in MRI.