Loïc Peter - Academia.edu (original) (raw)

Papers by Loïc Peter

Research paper thumbnail of Automatic Event Detection within Thrombus Formation Based on Integer Programming

Lecture Notes in Computer Science, 2013

ABSTRACT After a blood vessel injury, blood platelets progressively aggregate on the damaged site... more ABSTRACT After a blood vessel injury, blood platelets progressively aggregate on the damaged site to stop the resulting blood loss. This natural mechanism called thrombosis can however be prone to malfunctions and lead to the complete obstruction of the blood vessel. Thrombosis disorders play a crucial role in coronary artery diseases and the identification of genetic risk predispositions would therefore considerably help their diagnosis and therapy. In vitro experiments are conducted in this purpose by perfusing blood from several donors over a surface of collagen fibres, which results in the progressive attachment of platelets. Based on the segmentation over time of these aggregates called thrombi, we propose in this paper an automatic method combining tracking and event detection which allows the extraction of characteristics of interest for each thrombus growth individually, in order to find a potential correlation between these growth features and blood donors genetic disorders. We demonstrate the benefits of our approach and the accuracy of its results through an experimental validation.

Research paper thumbnail of Automatic segmentation and tracking of thrombus formation within in vitro microscopic video sequences

2012 9th IEEE International Symposium on Biomedical Imaging (ISBI), 2012

ABSTRACT There is a need to more accurately link human genetic variance with thrombotic risk. Thr... more ABSTRACT There is a need to more accurately link human genetic variance with thrombotic risk. Thrombus formation results from adhesion of blood platelets to a site of injury, followed by their progressive aggregation and occasional embolization. To observe this in vitro, blood is perfused over a surface of collagen fibres, during video microscopy. This paper proposes three complementary gradient-based features which, if included in a regularized machine learning framework, yield the accurate segmentation of thrombi during such acquisitions. A novel tracking method of thrombi as deformable growing objects under split and merge conditions is also introduced.

Research paper thumbnail of Leveraging Random Forests for Interactive Exploration of Large Histological Images

Lecture Notes in Computer Science, 2014

The large size of histological images combined with their very challenging appearance are two mai... more The large size of histological images combined with their very challenging appearance are two main difficulties which considerably complicate their analysis. In this paper, we introduce an interactive strategy leveraging the output of a supervised random forest classifier to guide a user through such large visual data. Starting from a forest-based pixelwise estimate, subregions of the images at hand are automatically ranked and sequentially displayed according to their expected interest. After each region suggestion, the user selects among several options a rough estimate of the true amount of foreground pixels in this region. From these one-click inputs, the region scoring function is updated in real time using an online gradient descent procedure, which corrects on-the-fly the shortcomings of the initial model and adapts future suggestions accordingly. Experimental validation is conducted for extramedullary hematopoesis localization and demonstrates the practical feasibility of the procedure as well as the benefit of the online adaptation strategy.

Research paper thumbnail of Characterizing the mechanics of cultured cell monolayers

One-cell-thick monolayers are the simplest tissues in multicellular organisms, yet they fulfill c... more One-cell-thick monolayers are the simplest tissues in multicellular organisms, yet they fulfill critical roles in development and normal physiology. In early development, embryonic morphogenesis results largely from monolayer rearrangement and deformation due to internally generated forces. Later, monolayers act as physical barriers separating the internal environment from the exterior and must withstand externally applied forces. Though resisting and generating mechanical forces is an essential part of monolayer function, simple experimental methods to characterize monolayer mechanical properties are lacking. Here, we describe a system for tensile testing of freely suspended cultured monolayers that enables the examination of their mechanical behavior at multi-, uni-, and subcellular scales. Using this system, we provide measurements of monolayer elasticity and show that this is two orders of magnitude larger than the elasticity of their isolated cellular components. Monolayers could withstand more than a doubling in length before failing through rupture of intercellular junctions. Measurement of stress at fracture enabled a first estimation of the average force needed to separate cells within truly mature monolayers, approximately ninefold larger than measured in pairs of isolated cells. As in single cells, monolayer mechanical properties were strongly dependent on the integrity of the actin cytoskeleton, myosin, and intercellular adhesions interfacing adjacent cells. High magnification imaging revealed that keratin filaments became progressively stretched during extension, suggesting they participate in monolayer mechanics. This multiscale study of monolayer response to deformation enabled by our device provides the first quantitative investigation of the link between monolayer biology and mechanics.

Research paper thumbnail of Random Forests for Phase Detection in Surgical Workflow Analysis

Information Processing in Computer-Assisted Interventions, 2014

Research paper thumbnail of Epithelial repair is a two-stage process driven first by dying cells and then by their neighbours

Development, 2014

Epithelial cells maintain an essential barrier despite continuously undergoing mitosis and apopto... more Epithelial cells maintain an essential barrier despite continuously undergoing mitosis and apoptosis. Biological and biophysical mechanisms have evolved to remove dying cells while maintaining that barrier. Cell extrusion is thought to be driven by a multicellular filamentous actin ring formed by neighbouring cells, the contraction of which provides the mechanical force for extrusion, with little or no contribution from the dying cell. Here, we use live confocal imaging, providing time-resolved three-dimensional observations of actomyosin dynamics, to reveal new mechanical roles for dying cells in their own extrusion from monolayers. Based on our observations, the clearance of dying cells can be subdivided into two stages. The first, previously unidentified, stage is driven by the dying cell, which exerts tension on its neighbours through the action of a cortical contractile F-actin and myosin ring at the cell apex. The second stage, consistent with previous studies, is driven by a multicellular F-actin ring in the neighbouring cells that moves from the apical to the basal plane to extrude the dying cell. Crucially, these data reinstate the dying cell as an active physical participant in cell extrusion.

Research paper thumbnail of Learning from Multiple Experts with Random Forests: Application to the Segmentation of the Midbrain in 3D Ultrasound

Lecture Notes in Computer Science, 2013

In the field of computer aided medical image analysis, it is often difficult to obtain reliable g... more In the field of computer aided medical image analysis, it is often difficult to obtain reliable ground truth for evaluating algorithms or supervising statistical learning procedures. In this paper we present a new method for training a classification forest from images labelled by variably performing experts, while simultaneously evaluating the performance of each expert. Our approach builds upon state-of-the-art randomized classification forest techniques for medical image segmentation and recent methods for the fusion of multiple expert decisions. By incorporating the performance evaluation within the training phase, we obtain a novel forest framework for learning from conflicting expert decisions, accounting for both inter-and intra-expert variability. We demonstrate on a synthetic example that our method allows to retrieve the correct segmentation among other incorrectly labelled images, and we present an application to the automatic segmentation of the midbrain in 3D transcranial ultrasound images.

Research paper thumbnail of Automatic Event Detection within Thrombus Formation Based on Integer Programming

Lecture Notes in Computer Science, 2013

ABSTRACT After a blood vessel injury, blood platelets progressively aggregate on the damaged site... more ABSTRACT After a blood vessel injury, blood platelets progressively aggregate on the damaged site to stop the resulting blood loss. This natural mechanism called thrombosis can however be prone to malfunctions and lead to the complete obstruction of the blood vessel. Thrombosis disorders play a crucial role in coronary artery diseases and the identification of genetic risk predispositions would therefore considerably help their diagnosis and therapy. In vitro experiments are conducted in this purpose by perfusing blood from several donors over a surface of collagen fibres, which results in the progressive attachment of platelets. Based on the segmentation over time of these aggregates called thrombi, we propose in this paper an automatic method combining tracking and event detection which allows the extraction of characteristics of interest for each thrombus growth individually, in order to find a potential correlation between these growth features and blood donors genetic disorders. We demonstrate the benefits of our approach and the accuracy of its results through an experimental validation.

Research paper thumbnail of Automatic segmentation and tracking of thrombus formation within in vitro microscopic video sequences

2012 9th IEEE International Symposium on Biomedical Imaging (ISBI), 2012

ABSTRACT There is a need to more accurately link human genetic variance with thrombotic risk. Thr... more ABSTRACT There is a need to more accurately link human genetic variance with thrombotic risk. Thrombus formation results from adhesion of blood platelets to a site of injury, followed by their progressive aggregation and occasional embolization. To observe this in vitro, blood is perfused over a surface of collagen fibres, during video microscopy. This paper proposes three complementary gradient-based features which, if included in a regularized machine learning framework, yield the accurate segmentation of thrombi during such acquisitions. A novel tracking method of thrombi as deformable growing objects under split and merge conditions is also introduced.

Research paper thumbnail of Leveraging Random Forests for Interactive Exploration of Large Histological Images

Lecture Notes in Computer Science, 2014

The large size of histological images combined with their very challenging appearance are two mai... more The large size of histological images combined with their very challenging appearance are two main difficulties which considerably complicate their analysis. In this paper, we introduce an interactive strategy leveraging the output of a supervised random forest classifier to guide a user through such large visual data. Starting from a forest-based pixelwise estimate, subregions of the images at hand are automatically ranked and sequentially displayed according to their expected interest. After each region suggestion, the user selects among several options a rough estimate of the true amount of foreground pixels in this region. From these one-click inputs, the region scoring function is updated in real time using an online gradient descent procedure, which corrects on-the-fly the shortcomings of the initial model and adapts future suggestions accordingly. Experimental validation is conducted for extramedullary hematopoesis localization and demonstrates the practical feasibility of the procedure as well as the benefit of the online adaptation strategy.

Research paper thumbnail of Characterizing the mechanics of cultured cell monolayers

One-cell-thick monolayers are the simplest tissues in multicellular organisms, yet they fulfill c... more One-cell-thick monolayers are the simplest tissues in multicellular organisms, yet they fulfill critical roles in development and normal physiology. In early development, embryonic morphogenesis results largely from monolayer rearrangement and deformation due to internally generated forces. Later, monolayers act as physical barriers separating the internal environment from the exterior and must withstand externally applied forces. Though resisting and generating mechanical forces is an essential part of monolayer function, simple experimental methods to characterize monolayer mechanical properties are lacking. Here, we describe a system for tensile testing of freely suspended cultured monolayers that enables the examination of their mechanical behavior at multi-, uni-, and subcellular scales. Using this system, we provide measurements of monolayer elasticity and show that this is two orders of magnitude larger than the elasticity of their isolated cellular components. Monolayers could withstand more than a doubling in length before failing through rupture of intercellular junctions. Measurement of stress at fracture enabled a first estimation of the average force needed to separate cells within truly mature monolayers, approximately ninefold larger than measured in pairs of isolated cells. As in single cells, monolayer mechanical properties were strongly dependent on the integrity of the actin cytoskeleton, myosin, and intercellular adhesions interfacing adjacent cells. High magnification imaging revealed that keratin filaments became progressively stretched during extension, suggesting they participate in monolayer mechanics. This multiscale study of monolayer response to deformation enabled by our device provides the first quantitative investigation of the link between monolayer biology and mechanics.

Research paper thumbnail of Random Forests for Phase Detection in Surgical Workflow Analysis

Information Processing in Computer-Assisted Interventions, 2014

Research paper thumbnail of Epithelial repair is a two-stage process driven first by dying cells and then by their neighbours

Development, 2014

Epithelial cells maintain an essential barrier despite continuously undergoing mitosis and apopto... more Epithelial cells maintain an essential barrier despite continuously undergoing mitosis and apoptosis. Biological and biophysical mechanisms have evolved to remove dying cells while maintaining that barrier. Cell extrusion is thought to be driven by a multicellular filamentous actin ring formed by neighbouring cells, the contraction of which provides the mechanical force for extrusion, with little or no contribution from the dying cell. Here, we use live confocal imaging, providing time-resolved three-dimensional observations of actomyosin dynamics, to reveal new mechanical roles for dying cells in their own extrusion from monolayers. Based on our observations, the clearance of dying cells can be subdivided into two stages. The first, previously unidentified, stage is driven by the dying cell, which exerts tension on its neighbours through the action of a cortical contractile F-actin and myosin ring at the cell apex. The second stage, consistent with previous studies, is driven by a multicellular F-actin ring in the neighbouring cells that moves from the apical to the basal plane to extrude the dying cell. Crucially, these data reinstate the dying cell as an active physical participant in cell extrusion.

Research paper thumbnail of Learning from Multiple Experts with Random Forests: Application to the Segmentation of the Midbrain in 3D Ultrasound

Lecture Notes in Computer Science, 2013

In the field of computer aided medical image analysis, it is often difficult to obtain reliable g... more In the field of computer aided medical image analysis, it is often difficult to obtain reliable ground truth for evaluating algorithms or supervising statistical learning procedures. In this paper we present a new method for training a classification forest from images labelled by variably performing experts, while simultaneously evaluating the performance of each expert. Our approach builds upon state-of-the-art randomized classification forest techniques for medical image segmentation and recent methods for the fusion of multiple expert decisions. By incorporating the performance evaluation within the training phase, we obtain a novel forest framework for learning from conflicting expert decisions, accounting for both inter-and intra-expert variability. We demonstrate on a synthetic example that our method allows to retrieve the correct segmentation among other incorrectly labelled images, and we present an application to the automatic segmentation of the midbrain in 3D transcranial ultrasound images.