nach oben

2017 | Buch

Imaging for Patient-Customized Simulations and Systems for Point-of-Care Ultrasound

International Workshops, BIVPCS 2017 and POCUS 2017, Held in Conjunction with MICCAI 2017, Québec City, QC, Canada, September 14, 2017, Proceedings

herausgegeben von: M. Jorge Cardoso, Tal Arbel, Prof. Dr. João Manuel R.S. Tavares, Stephen Aylward, Prof. Dr. Shuo Li, Emad Boctor, Dr. Gabor Fichtinger, Prof. Kevin Cleary, Bradley Freeman, Luv Kohli, Deborah Shipley Kane, Matt Oetgen, Sonja Pujol

Verlag: Springer International Publishing

Buchreihe : Lecture Notes in Computer Science

Enthalten in: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Einloggen, um Zugang zu erhalten

Über dieses Buch

This book constitutes the refereed joint proceedings of the International Workshop on Bio-Imaging and Visualization for Patient-Customized Simulations, BIVPCS 2017, and the International Workshop on Point-of-Care Ultrasound, POCUS 2017, held in conjunction with the 20th International Conference on Medical Imaging and Computer-Assisted Intervention, MICCAI 2017, in Québec City, QC, Canada, in September 2017.

The 12 full papers presented at BIVPCS 2017 and the 7 full papers presented at POCUS 2017 were carefully reviewed and selected. The papers feature research from complementary fields such as signal and image processing, mechanics, computational vision, mathematics, physics, informatics, computer graphics, bio-medical-practice, psychology and industry as well as ultrasound image systems applications.

Inhaltsverzeichnis

Frontmatter

Erratum to: A Novel Ultrasound Imaging Method for 2D Temperature Monitoring of Thermal Ablation

Chloé Audigier, Younsu Kim, Emad Boctor

International Workshop on Bio-Imaging and Visualization for Patient-Customized Simulations, BIVPCS 2017

Frontmatter

Cortical Envelope Modeling for Interactive Patient-Customized Curvilinear Reformatting in the Native Space

Abstract

Focal cortical dysplasia is one of the most common cause of medically refractory epilepsy. Its imaging features include cortical architectural abnormalities and abnormal structural arrangement at the interface between the grey matter and the white matter. It is well-known that curvilinear multiplanar reformatting (CMPR), consisting in re-slicing the brain almost prependicular to the inward folding gyri from the view of anatomical planes, enhances the visualization of these abnormalities. In this paper, we present yet another interactive modeling of a patient-customized cortical envelope with which we can automatically re-slice the brain volume in a fashion similar to CMPR. Although our proposal requires fewer user interactions in comparison with the previous proposals, we show that the outcomes of re-slicing match those of the conceived CMPR.

Wallace Souza Loos, Clarissa Lin Yasuda, Fernando Cendes, Shin-Ting Wu

Simulation of Patient-Specific Deformable Ultrasound Imaging in Real Time

Abstract

Intraoperative ultrasound is an imaging modality frequently used to provide delineation of tissue boundaries. This paper proposes a simulation platform that enables rehearsal of patient-specific deformable ultrasound scanning in real-time, using preoperative CT as the data source. The simulation platform was implemented within the GPU-accelerated NVIDIA FleX position-based dynamics framework. The high-resolution particle model is used to deform both surface and volume meshes. The latter is used to compute the barycentric coordinates of each simulated ultrasound image pixel in the surrounding volume, which is then mapped back to the original undeformed CT volume. To validate the computation of simulated ultrasound images, a kidney phantom with an embedded tumour was CT-scanned in the rest position and at five different levels of probe-induced deformation. Measures of normalised cross-correlation and similarity between features were adopted to compare pairs of simulated and ground truth images. The accurate results demonstrate the potential of this approach for clinical translation.

Mafalda Camara, Erik Mayer, Ara Darzi, Philip Pratt

A Hybrid CNN Feature Model for Pulmonary Nodule Differentiation Task

Abstract

Pulmonary nodule differentiation is one of the most challenge tasks of computer-aided diagnosis(CADx). Both texture method and shape estimation approaches previously presented could provide good performance to some extent in the literature. However, no matter 2D or 3D textures extracted, they just tend to observe characteristics of the pulmonary nodules from a statistical perspective according to local features’ change, which hints they are helpless to work as global as the human who always be aware of the characteristics of given target as a combination of local features and global features, thus they have certain limitations. Enlightened by the currently prevailing learning ability of convolutional neural network (CNN) and previously contributions provided by texture features, we here presented a hybrid method for better to complete the differentiation task. It can be observed that our proposed multi-channel CNN model has a better discrimination in capacity according to the projection of distributions of extracted features and achieved a new record with AUC 97.04 on LIDC-IDRI database.

Tingting Zhao, Huafeng Wang, Lihong Li, Yifang Qi, Haoqi Gao, FangFang Han, Zhengrong Liang, Yanmin Qi, Yuan Cao

A 3D Ultrasound Informed Model of the Human Gastrocnemius Muscle

Abstract

Muscle fascicle structure characterises muscle function, which in turn plays a key role in computer simulation of muscle shape. In this study we use 3D ultrasound from human gastrocnemius muscle to identify and map the muscle fascicle orientation and deformation during passive motion in four subjects. This muscle fascicle description is integrated into a representative muscle volume element using a free-form deformation technique to create a muscle primitive that deforms according to the embedded muscle fascicles within. For each subject computed passive tensile force was used to optimise the constitutive behaviour so that the known deformation matched this load. Each subject was fit to match deformation at 25%, 50%, 75% and 100% of muscle stretch. The medial gastrocnemius muscle built from these muscle primitives exhibited a contractile shape that is consistent to that observed in human gastrocnemius contraction. This shape was evaluated against the same muscle embedded with muscle fascicles derived from diffusion-weighted magnetic resonance imaging and was in good qualitative agreement. Muscle primitives may be used as building blocks to build large muscle volumes for mechanics simulation, visualisation and medical education.

M. Alipour, K. Mithraratne, R. D. Herbert, J. Fernandez

Atlas-Based 3D Intensity Volume Reconstruction of Musculoskeletal Structures in the Lower Extremity from 2D Calibrated X-Ray Images

Abstract

In this paper, the reconstruction of 3D intensity volumes of femur, tibia and three muscles around the thigh region from a pair of calibrated X-ray images is addressed. We present an atlas-based 2D-3D intensity volume reconstruction approach by combining a 2D-2D non-rigid registration based 3D landmark reconstruction procedure with an adaptive regularization step. More specifically, an atlas derived from the CT acquisition of a healthy lower extremity, together with the input calibrated X-ray images are used to reconstruct those musculoskeletal structures. To avoid the potential penetration of the reconstructed femoral and tibial volumes that might be caused by reconstruction error, we come up with an articulated 2D-3D reconstruction strategy, which can effectively preserve knee joint structure. Another contribution from our work is the application of the proposed 2D-3D reconstruction pipeline to derive the patient-specific volumes of three thigh muscles around the thigh region.

Weimin Yu, Guoyan Zheng

Automatic Liver Lesion Segmentation in CT Combining Fully Convolutional Networks and Non-negative Matrix Factorization

Abstract

Automatic liver tumor segmentation is an important step towards digital medical research, clinical diagnosis and therapy planning. However, the existence of noise, low contrast and heterogeneity make the automatic liver tumor segmentation remaining an open challenge. In this work, we focus on a novel automatic method to segment liver tumor in abdomen images from CT scans by using fully convolutional networks (FCN) and non-negative matrix factorization (NMF). We train the FCN for semantic liver and tumor segmentation. The segmented liver and tumor regions of FCN are used as ROI and initialization for the NMF based tumor refinement, respectively. We refine the surfaces of tumors using a 3D deformable model which derived from NMF and driven by local cumulative spectral histograms (LCSH). The refinement is designed to obtain a smoother, more accurate and natural liver tumor surface. Experiments demonstrated that the proposed segmentation method achieves satisfactory results. Likewise, it has been notably observed that the computing time of the segmentation method is no more than one minute for each CT volume.

Shenhai Zheng, Bin Fang, Laquan Li, Mingqi Gao, Yi Wang, Kaiyi Peng

Constructing Detailed Subject-Specific Models of the Human Masseter

Abstract

We investigate the structural details of the human masseter and their contribution to force-transmission necessary for mastication through a computational modelling study. We compare two subject-specific models, constructed using data acquired by a dissection and digitization procedure on cadaveric specimens. Despite architectural differences between the two masseters, we find that in both instances it is necessary to capture the combination of the multipennate nature of the muscle fibres, as well as the increased aponeurosis stiffness, in order to reproduce adequate clenching forces. We also demonstrate the feasibility of deformably registering these architectural templates to target muscle surfaces in order to create new subject-specific models.

C. Antonio Sánchez, Zhi Li, Alan G. Hannam, Purang Abolmaesumi, Anne Agur, Sidney Fels

MRI-Based Heart and Torso Personalization for Computer Modeling and Simulation of Cardiac Electrophysiology

Abstract

In the last decade, electrophysiological models for in-silico simulations of cardiac electrophysiology have gained much attention in the research field. However, to translate them to clinical uses, the models need personalization based on recordings from the patient. In this work, we explore methodologies for the patient-specific personalization of torso and heart geometric models based on standard clinical cardiac magnetic resonance acquisitions to enable simulations. The inclusion of the torso and its internal structures allows simulations of the human ventricular electrophysiological activity from the ionic level to the body surface potentials and to the electrocardiogram.

Ernesto Zacur, Ana Minchole, Benjamin Villard, Valentina Carapella, Rina Ariga, Blanca Rodriguez, Vicente Grau

Rapid Prediction of Personalised Muscle Mechanics: Integration with Diffusion Tensor Imaging

Abstract

Diffusion Tensor Imaging (DTI) has been widely used to characterise the 3D fibre architecture in both neural and muscle mechanics. However, the computational expense associated with continuum models make their use in graphics and medical visualisation intractable. This study presents an integration of continuum muscle mechanics with partial least squares regression to create a fast mechano-statistical model. We use the human triceps surae muscle as an example informed though DTI. Our statistical models predicted muscle shape (within 0.063 mm RMS error), musculotendon force (within 1% error), and tissue strain (within 8% max error during contraction). Importantly, the presented framework may play a role in addressing computational cost of predicting detailed muscle information through popular rigid body solvers such as OpenSIM.

J. Fernandez, K. Mithraratne, M. Alipour, G. Handsfield, T. Besier, J. Zhang

Approaches to Brain Tissue Quantification with Comparison on Supporting the Detection of Age-Related Dementia in MRI

Abstract

In this paper, a comparison of two different approaches is given for brain tissue segmentation using various sources of techniques, the level-set thresholding segmentation with sparse model (LTSSM) and the segmentation with the k-means clustering (SKMC), in magnetic resonance imaging (MRI). In the LTSSM approach, the system searches for level-set thresholding in the working subsets recursively for segmentation. Unlike the LTSSM approach, the SKMC approach applies the k-means clustering to group the brain tissue objects into three classes (grey matter, white matter and cerebrospinal fluid), and then segment the three groups in the different components in the RGB color space. At the validation stage, both approaches of the LTSSM and the SKMC are implemented using the real-time OASIS data for comparison purpose. The experimental results demonstrate the robustness of both approaches for brain tissue segmentation with comparison, in terms of the Dice similarity and sensitivity in MRI.

Peifang Guo

Relating Atrial Appendage Flow Stasis Risk from Computational Fluid Dynamics to Imaging Based Appearance Paradigms for Cardioembolic Risk

Abstract

Emboli originating from the left atrial appendage are a major cause of transient ischemic attack and cardioembolic stroke. Whereas this risk has been shown to be correlated with left atrial appendage morphology (Cactus, Chicken Wing, Windsock, and Cauliflower shapes) determined from 3D imaging, this clinical correlation is found wanting with regard to a biomechanically grounded underlying basis for thrombosis based on intra-atrial hemodynamics. We define a novel probabilistic risk stratification paradigm for intra-atrial flow stasis based on personalized computational fluid dynamics.

Soroosh Sanatkhani, Prahlad G. Menon

Deformable Multi-material 2-Simplex Surface Mesh for Intraoperative MRI-Ready Surgery Planning and Simulation, with Deep-Brain Stimulation Applications

Abstract

Printed and/or digital atlases are important tools for medical research and surgical intervention. While these atlases can provide guidance in identifying anatomical structures, they do not take into account the wide variations in the shape and size of anatomical structures that can occur from patient to patient. Accurate, patient-specific representations are especially important for surgical interventions like deep brain stimulation, where even small inaccuracies can result in dangerous complications. This research effort extends the discrete deformable 2-simplex mesh into the multi-material domain where geometry-based internal forces and image-based external forces are used in the deformation process. Multi-material 2-simplex meshes having shared boundaries are initialized from multi-material triangular surface meshes. A multi-material deformable framework is presented and used to segment anatomical structures of the deep brain region such as the subthalamic nucleus.

T. Rashid, S. Sultana, G. S. Fischer, J. Pilitsis, M. A. Audette

International Workshop on Point-of-Care Ultrasound: Algorithms, Hardware, and Applications, POCUS 2017

Frontmatter

Combining Automated Image Analysis with Obstetric Sweeps for Prenatal Ultrasound Imaging in Developing Countries

Abstract

Ultrasound imaging can be used to detect maternal risk factors, but it remains out of reach for most pregnant women in developing countries because there is a severe shortage of well-trained sonographers. In this paper we show the potential of combining the obstetric sweep protocol (OSP) with image analysis to automatically obtain information about the fetus. The OSP can be taught to any health care worker without any prior knowledge of ultrasound within a day, obviating the need for a well-trained sonographer to acquire the ultrasound images. The OSP was acquired from 317 pregnant women using a low-cost ultrasound device in St. Luke’s Hospital in Wolisso, Ethiopia. A deep learning network was used to automatically detect the fetal head in the OSP data. The fetal head detection was used to detect twins, determine fetal presentation and estimate gestational age without the need of a well-trained sonographer.

Thomas L. A. van den Heuvel, Hezkiel Petros, Stefano Santini, Chris L. de Korte, Bram van Ginneken

Automatic Estimation of the Optic Nerve Sheath Diameter from Ultrasound Images

Abstract

We present an algorithm to automatically estimate the diameter of the optic nerve sheath from ocular ultrasound images. The optic nerve sheath diameter provides a proxy for measuring intracranial pressure, a life threating condition frequently associated with head trauma. Early treatment of elevated intracranial pressures greatly improves outcomes and drastically reduces the mortality rate. We demonstrate that the proposed algorithm combined with a portable ultrasound device presents a viable path for early detection of elevated intracranial pressure in remote locations and without access to trained medical imaging experts.

Samuel Gerber, Maeliss Jallais, Hastings Greer, Matt McCormick, Sean Montgomery, Bradley Freeman, Deborah Kane, Deepak Chittajallu, Neal Siekierski, Stephen Aylward

Achieving Fluid Detection by Exploiting Shadow Detection Methods

Abstract

Ultrasound provides a useful and readily available imaging tool. The big challenge in acquiring a good ultrasound image are possible shadow artefacts that hide anatomical structures. This applies in particular to 3D ultrasound acquisitions, because shadow artefacts may be recorded outside the visualized image plane. There are only a few automatic methods for shadow artefact detection. In our work we like to introduce a new shadow detection method that is based on an adaptive thresholding approach. The development was attempted, after existing methods had been extended to separate shadow and fluid regions. The entire detection procedure utilizes only the ultrasound scan line information and some basic knowledge about the ultrasound propagation inside the human body. Applying our method, the ultrasound operator can retrieve combined information about shadow and fluid locations, that may be invaluable for image acquisition or diagnosis. The method can be applied to conventional 2D as well as 3D ultrasound images.

Matthias Noll, Julian Puhl, Stefan Wesarg

A Probe-Camera System for 3D Ultrasound Image Reconstruction

Abstract

This paper proposes a probe-camera system for 3D ultrasound (US) image reconstruction with probe-camera calibration and probe localization methods. The probe-camera calibration method employs an existing US phantom for convenience with a simple procedure. The probe localization method employs structure from motion (SfM) to estimate the camera motion. SfM is used to reconstruct 3D point clouds from multiple-view images and simultaneously estimate each camera position. Through experiments using the developed system, we demonstrate that the proposed method exhibits good performance to reconstruct 3D US volume.

Koichi Ito, Kouya Yodokawa, Takafumi Aoki, Jun Ohmiya, Satoshi Kondo

Ultrasound Augmentation: Rapid 3-D Scanning for Tracking and On-Body Display

Abstract

By using a laser projector and high speed camera, we can add three capabilities to an ultrasound system: tracking the probe, tracking the patient, and projecting information onto the probe and patient. We can use these capabilities to guide an untrained operator to take high quality, well framed ultrasound images for computer-augmented, point-of-care ultrasound applications.

Maeliss Jallais, Hastings Greer, Sam Gerber, Matt McCormick, Deepak Chittajallu, Neal Siekierski, Stephen Aylward

Overall Proficiency Assessment in Point-of-Care Ultrasound Interventions: The Stopwatch is not Enough

Abstract

With the shift in the medical education curriculum to a competency-based model, objective proficiency assessment is necessary. In this work, we use exploratory factor analysis to assess which primitive metrics convey unique information about proficiency in point-of-care ultrasound applications. We retrospectively validate the proposed methods on three datasets: FAST examination, femoral line, and lumbar puncture. We identify a minimal set of metrics for proficiency assessment in each application. Furthermore, we validate that overall proficiency assessment methods are unaffected by the removal of redundant metrics. This work demonstrates that proficiency in point-of-care ultrasound applications is multi-faceted, and that measuring completion time alone is not enough and application-specific metrics have added value in proficiency assessment.

Matthew S. Holden, Zsuzsanna Keri, Tamas Ungi, Gabor Fichtinger

A Novel Ultrasound Imaging Method for 2D Temperature Monitoring of Thermal Ablation

Abstract

Accurate temperature monitoring is a crucial task that directly affects the safety and effectiveness of thermal ablation procedures.

Compared to MRI, ultrasound-based temperature monitoring systems have many advantages, including higher temporal resolution, low cost, safety, mobility and ease of use. However, conventional ultrasound (US) images have a limited accuracy due to a weak temperature sensitivity. As a result, it is more challenging to fully meet the clinical requirements for assessing the completion of ablation therapy.

A novel imaging method for temperature monitoring is proposed based on the injection of virtual US pattern in the US brightness mode (B-mode) image coupled with biophysical simulation of heat propagation. This proposed imaging method does not require any hardware extensions to the conventional US B-mode system. The main principle is to establish a bi-directional US communication between the US imaging machine and an active element inserted within the tissue. A virtual pattern can then directly be created into the US B-mode display during the ablation by controlling the timing and amplitude of the US field generated by the active element. Changes of the injected pattern are related to the change of the ablated tissue temperature through the additional knowledge of a biophysical model of heat propagation in the tissue. Those changes are monitored during ablation, generating accurate spatial and temporal temperature maps.

We demonstrated in silico the method feasibility and showed experimentally its applicability on a clinical US scanner using ex vivo data. Promising results are achieved: a mean temperature error smaller than 4 \({^\circ }\mathrm{C}\) was achieved in all the simulation experiments. The system performance is tested under different configurations of noise in the data. The effect of error in the localization of the RFA probe is also evaluated.

Chloé Audigier, Younsu Kim, Emad Boctor

Backmatter

Titel: Imaging for Patient-Customized Simulations and Systems for Point-of-Care Ultrasound
herausgegeben von: M. Jorge Cardoso
Tal Arbel
Prof. Dr. João Manuel R.S. Tavares
Stephen Aylward
Prof. Dr. Shuo Li
Emad Boctor
Dr. Gabor Fichtinger
Prof. Kevin Cleary
Bradley Freeman
Luv Kohli
Deborah Shipley Kane
Matt Oetgen
Sonja Pujol
Verlag: Springer International Publishing
Electronic ISBN: 978-3-319-67552-7
Print ISBN: 978-3-319-67551-0
DOI: https://doi.org/10.1007/978-3-319-67552-7