Recent advances in imaging and microscopy techniques have led to a surge in biological data. Different imaging modalities such as computed tomography (CT), magnetic resonance imaging (MRI), confocal microscopy, and serial sectioning microscopy (both optical and electron microscopy) have enabled the observation of morphological characteristics of biological systems and organs at multiple spatial and temporal scales (e.g., neuronal networks, vascular networks, airways, and villi). The different imaging modalities and scales, and the amount of data (easily reaching several terabytes) pose serious computational and mathematical challenges in image analysis and modeling. For example, segmentation, tracing, and reconstruction of densely packed objects in 3D is computationally intensive and error-prone, and inference of nano-scale morphological properties from micro- or macro-scale imaging data require the discovery of invariant statistical properties across scales. This minisymposium will focus on latest advances in multiscale and space-time imaging methods, image processing/analysis algorithms, and modeling frameworks that will enable the interpretation and integration of data from multiple scales and spatial-temporal correlations in biological systems. Also relevant to this symposium are strategies for data and model sharing that will require further abstractions for general multiscale and space-time frameworks for imaging and image analysis.

Notes: (1) "*" marks confirmed speakers. (2) See Session_6_Speakers for the abstracts.

Back to Minisymposium_planning.

ABSTRACTS

• Choe: High-throughput Imaging using Knife-Edge Scanning Microscope and Array Tomography, and Fast Algorithms for Multiscale Image Analysis and Reconstruction

Yoonsuck Choe (1), Louise C. Abbott (2), Donghyeop Han (1), Pei-San Huang (2), John Keyser (1), Jaerock Kwon (1),

David Mayerich (1), Zeki Melek (1), Bruce McCormick (1), Kristina Micheva (3), and Stephen J. Smith (3)

(1) Department of Computer Science, Texas A&M University

(2) Department of Veterinary Integrative Biosciences, Texas A&M University

(3) Department of Molecular and Cellular Physiology, Stanford University

Recent advances in serial-sectioning microscopy have enabled high-throughput imaging of massive volumes of biological microstructure at a very high resolution. One example is the Knife-Edge Scanning Microscope (KESM) we developed at Texas A&M, which is one of the few that combines serial sectioning and imaging in an integrated process. The KESM is capable of imaging biological tissue (about 1 cm³) at 300 nm x 300 nm x 500 nm resolution within 100 hours, generating data at a rate of 180 MB/s. The resulting data per organ (e.g., a mouse brain) can easily exceed tens of terabytes. Another example is a complementary microscopy method, Array Tomography, developed by our team at Stanford. Array Tomography allows the imaging of molecularly labeled objects in the brain tissue using fluorescent microscopy, and registered high-resolution imaging with scanning electron microscopy, enabling an accurate mapping across nano- to micro scale. Due to the massive amounts of data at multiple scales, morphological reconstruction algorithms that are fast, resource efficient, and accurate become necessary. We will present our latest results in large-scale microscopic neuronal circuit data acquisition in the mouse brain using KESM and Array Tomography, and discuss the fast algorithms we developed for tracing and analyzing neuronal morphology.

Presentation slides

• Ailiani: Quantitative Analysis of Gut Motion in an Animal Model using Dynamic MRI and 3-D Live-wire Image Segmentation

Amit C. Ailiani (1), Thomas Neuberger (1), Gino Banco (2), Yanxing Wang (2), Nadine B. Smith(1), James G. Brasseur (2,1), Andrew G. Webb (1)

(1) Department of Bioengineering, Pennsylvania State University

(2) Department of Mechanical Engineering, Pennsylvania State University

Conventional methods of quantifying segmental and peristaltic motion in animal models are highly invasive, involving the isolation of segments of small intestine either from dead or anesthetized animals. The present study was undertaken to non-invasively analyze these motions in the jejunum region of anesthetized rats using dynamic contrast enhanced magnetic resonance imaging (MRI), and quantify thses motion using spatiotemporal maps. Dynamic images (~ 1000 images at 6 frames per second) of the GE tract were acquired in vivo using a gradient echo imaging sequence. A semi-automated 2D spatial + time image segmentation algorithm based on 3D live wire and gradient vector flow snakes was implemented to accurately segment the dynamically acquired images. A 2D thinning algorithm was applied to the segmented images to compute the medial axis, which was used to compute the spatiotemporal map of the acquired image sequence. The spatiotemporal map, which is plot of the diameter of the gut along the length of the gut with respect to time, is a widely-used and a compact representation of the complex gut motion. The frequency (0.29 ± 0.05 Hz) and period of segmental contraction (3.14 ± 0.14 s) and average distance between segmental constrictions (4.56 ± 1.21 mm) were very similar in the jejunum region among different rats, showing little inter-animal variability. The frequency (0.46 ± 0.01 Hz) and period of peristaltic waves (2.18 ± 0.05 s) correlated well with the frequency (0.5 Hz) and period (2 s) of slow waves which have been described in previous literature, leading to the conclusion that in vivo short distance peristaltic type contractions are the result of slow waves generated by interstitial cells of cajal. The speed of propagation of peristalsis wave (4.34 ± 1.03 mms-1) was found to be reduced under in-vivo conditions compared to in-vitro speeds. These represent the first quantitative in vivo results of the small intestine using non-invasive dynamic MRI approach. In this approach the segments of the GI tract are not isolated or exteriorized but are in true physiological conditions.

• Hoffman: Model-based Assessment of the Lung by Imaging

Eric A. Hoffman (1) and Ching-Long Lin (2,3)

(1) Departments of Biomedical Engineering, Medicine and Radiology, The University of Iowa

(2) Department of Mechanical and Industrial Engineering, The University of Iowa

(3) IIHR-Hydroscience & Engineering, The University of Iowa

There has been a growing need for sensitive and objective measures of regional lung status both for detection of disease and for outcomes analysis. X-ray CT remains the imaging modality of choice for comprehensively evaluating the lung. Significant advances are being made in both temporal and spatial resolution. Scan apertures are at sub-cardiac cycle time frames, allowing for the imaging of not only anatomy but also ventilation and perfusion, providing structure-to-function correlations.

We have brought together a multi-disciplinary team of investigators to develop a model/atlas of the normal human lung based upon these new measures. This atlas includes the lungs, lobes sublobar segments and airway and vascular branching structures of the lung and attached to each level of this structure will be the normal range of the CT-based measures of regional lung physiology including ventilation, perfusion, etc. This model/atlas of the normal human lung provides the comparative basis for detecting and quantitating pulmonary pathology.

Imaging is serving an important role in the phenotyping of diseases such as Asthma, COPD in general, as well as other interstial lung pathologies. As we are identifying specific differences amongst these patient populations, there is a growing interest to determine if computational fluid dynamics can provide insights into particular distribution patterns of pathology. In this talk we will discuss the growing interplay between CFD studies and the phenotyping of lung disease.

• Kerckhoffs: Image-based patient-specific multi-scale modeling of the failing heart

R. Kerckhoffs, S. Narayan, and A. McCulloch

UCSD

The most significant recent advance in management of heart failure (HF) with a conduction block has come from CRT and defibrillator therapy. However, up to 30% of currently eligible patients fail to respond to CRT. Notably, exact estimation of the percentage of non-responders, or of factors predicting non-response, are difficult because of different criteria used to define 'response'. Although these factors make it difficult to select optimally select patients for CRT, this remains of paramount importance to reduce unnecessary implants, procedural risks and healthcare expenses.

We propose to develop new computational tools for optimally selecting patients for CRT, then for optimizing CRT to each individual patient. Having acquired IRB approval, two to five patients with NYHA Class III or IV will be recruited at the San Diego VA Medical Center. Implementing the strategy outlined above, we will develop patient-specific computational models of the human heart from detailed mapping, and test their ability to predict observed short-term functional improvements after CRT. The clinical procedures that we will perform include: cardiac contrast CT to obtain ventricular geometry; electroanatomic mapping of both ventricles to obtain ventricular activation patterns during native (dyssynchronous) ventricular activation and during CRT; cardiac ultrasound to obtain dynamic ventricular (pseudo-) volumes and invasive dynamic pressure measurements during cardiac catheterization. Pressure-volume loops will be obtained during maneuvers to alter preload and afterload. Three and six months follow ups will be performed to assess long-term outcomes, to determine the predictive accuracy of our models, and also to provide an opportunity for the patient-specific models to optimize CRT in a patient-tailored fashion.