Medical Imaging

Computational Models of the Human Airway Tree

Margrit Betke has worked in the area of medical image analysis for over a decade. Most recently, she developed subject-specific airway tree models of the human lung to analyze the airways responsible for ventilation defects in asthmatics. Her collaboration with Kenneth Lutchen at Boston University's Biomedical Engineering Department and Mitchell Albert at Brigham and Women's Hospital resulted in the 2009 dissertation of Betke's PhD student William Mullally and a number of publications, most notably a 2009 article in the Annals of Biomedical Engineering:

W. Mullally, M. Betke, M. Albert, and K. Lutchen. "Explaining Clustered Ventilation Defects via a Minimal Number of Airway Closure Locations." Annals of Biomedical Engineering. 37(2):286-300, February 2009. Pdf file. Html open access.
W. Mullally, A. Milutinovic, M. Betke, M. Albert, and K. Lutchen "Personalized Airway Trees from a Generative Model, Lung Atlas, and Hyperpolarized Helium MRI." MICCAI 2006 Workshop "From Statistical Atlases to Personalized Models: Understanding Complex Diseases in Populations and Individuals." Copenhagen, Denmark, October 6, 2006. 4 pp. pdf.
W. Mullally, M. Betke, C. Bellardine, and K. Lutchen. "Locally Switching Between Cost Functions in Iterative Non-Rigid Registration." In Y. Liu, T. Jiang and C. Zhang, editors. Computer Vision for Biomedical Image Applications. First International Workshop, CVBIA 2005, Beijing, China, October 21, 2005. Proceedings. Lecture Notes in Computer Science 3765, pp. 367-377. Springer Verlag. Abstract. pdf.

Computational models of the human lung have been developed to study lung physiology and have been used to identify the airways responsible for mechanical dysfunction in asthmatics. Tgavalekos et al. (2007) used models anatomically consistent with the human lung to link ventilation defects to the heterogeneous closure of small airways. Their approach implicitly assumed a high degree of independence between airway closures as indicated by the low compactness of the airway structures mapped to individual ventilation defects. Venegas et al. (2005), however, have found that significant mutual dependence of airways may play a role in patchy ventilation of asthmatics. This led Mullally et al. (2009) to explore the question to what extent anatomically consistent models can be built which do not implicitly assume high independence of airways but instead allow for the mutual dependence of airways responsible for ventilation defects. They proposed an algorithm for generating subject-specific airway-tree models that minimize the number of airways that must be closed or severely constricted to cause observed ventilation defects. They also proposed novel approaches for measuring the compactness of airway structures. The approach showed that anatomically consistent models which link compact airway structures to ventilation defects can be built. The model also shows that some ventilation defects may be caused by closures of larger airways than previously reported.

Prediction of Tumor Motion

In collaboration with scientists Massachusetts General Hospital, Betke's group developed a lung and abdominal tumor tracking system that analyzes tumor motion due to respiration. Prediction of tumor motion is important for automatic and adaptive administration of radiation during therapy.

Fluoroscopy is currently used in treatment planning for patients undergoing radiation therapy. Radiation oncologists would like to maximize the amount of dose the tumor receives and minimize the amount delivered to the surrounding tissues. During treatment, patients breathe freely and so the tumor location will not be fixed. This makes calculating the amount of dose delivered to the tumor, and verifying that the tumor actually receives that dose, difficult. Betke's group first developed a correlation-based method of tracking the two-dimensional (2D) motion of internal markers (surgical clips) placed around a tumor. The group then developed a method to model the average and maximum three-dimensional (3D) motion of the clips given two orthogonal fluoroscopy videos of the same patients that were taken sequentially. In preliminary experiments, it was shown that both trackers had small errors in estimating marker positions (Brewer et al., MICCAI 2004). If imaging is possible during treatment, such motion models may be used for beam guided radiation.

The approach to correlate tumor motion models to a set of external markers for use in respiratory gating was also examined by the group (Gierga et al. 2005). Although tumor motion generally correlated well with external fiducial marker motion, relatively large underlying tumor motion can occur compared with external-marker motion and variations in the tumor position for a given marker position. Treatment margins should be determined on the basis of a detailed understanding of tumor motion, as opposed to relying only on external-marker information.

M. Betke, J. Ruel, G. C. Sharp, S. B. Jiang, D. P. Gierga, and G. T. Y. Chen. "Tracking and prediction of tumor movement in the abdomen." In A. Fred and A. Lourenço, editors, Pattern Recognition in Information Systems: Proceedings of the 6th International Workshop on Pattern Recogntion in Information Systems - PRIS 2006, pages 27-37, Paphos, Cyprus, May 2006. INSTICC Press. pdf.

D. P. Gierga, J. Brewer, G. C. Sharp, M. Betke, C. G. Willett, G. T. Y. Chen. "The correlation between internal and external markers for abdominal tumors: Implications for respiratory gating." International Journal of Radiation Oncology, Biology, Physics, 61:5, pp. 1551-1558, April 2005, pdf, abstract.

D. P. Gierga, G. T. Y. Chen, J. H. Kung, M. Betke, J. Lombardi, C. G. Willett, "Quantification of Respiration-induced Abdominal Tumor Motion and the Impact on IMRT Dose Distributions." International Journal on Radiation Oncology - Biology - Physics, 58:5, pp. 1584-1595, April 2004. pdf, abstract.

J. Brewer, M. Betke, D. P. Gierga, and G. T. Y. Chen. "Real-time 4D Tumor Tracking and Modeling From Internal and External Fiducials in Fluoroscopy." C. Barillot, D. R. Haynor (editors), Proceedings of the 7th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI 2004), Part II, LNCS 3217, pp. 594-601, Saint-Malo, France, September 2004, pdf, abstract.

Example-Based Image Registration via Boosted Classifiers

William Mullally, Stan Sclaroff, and Margrit Betke proposed a novel image registration framework which uses classifiers trained from examples of aligned images to achieve registration. Our approach was designed to register images of medical data where the physical condition of the patient has changed significantly and image intensities are drastically different. We used two boosted classifiers for each degree of freedom of image transformation. These two classifiers can both identify when two images are correctly aligned and provide an efficient means of moving towards correct registration for misaligned images. The classifiers capture local alignment information using multi-pixel comparisons and can therefore achieve correct alignments where approaches which rely on pixel-to-pixel comparisons, like correlation and mutual-information, fail. We tested our approach using images from CT scans acquired in a study of acute respiratory distress syndrome. We show a significant increase in registration accuracy in comparison to an approach using mutual information.

W. Mullally, S. Sclaroff, and M. Betke. Example-Based Image Registration via Boosted Classifiers, Department of Computer Science Technical Report BUCS-2009-007, Boston University. March 11, 2009. Abstract. pdf. ps.

Lung Fissure Segmentation

We proposed a shape-based curve-growing method for segmenting the major fissures in lungs imaged via thin-section computed tomography (CT). It can also be applied for segmenting objects with an open contour shape. We modeled the problem by a Bayesian network that is influenced by both the features that are extracted from the original CT images and prior knowledge of the shape of the fissure. The proposed technique takes advantage of the 3D structure of the fissure as a boundary between the lobes in the lungs. Prior shape knowledge obtained from the segmentation results on neighboring sections guides the evolution of the active curve. We proposed an approach in which the contribution of the prior-shape term of the energy function changes adaptively during the curve growing process based on an image entropy formulation. The proposed method effectively alleviates the problem of inappropriate weights of regularization terms, an effect that can occur with static regularization methods. We also proposed an effective method to initialize the curve automatically. We applied the shape-based curve-growing method to segment and visualize the lobes of the lungs on chest CT of ten patients with pulmonary nodules. The method had a linear-time worst-case complexity and segments the upper lung from the lower lung on a standard computer in less than 5 minutes. The experimental results demonstrated that the fissures segmented by the curve-growing method on average approximates the gold standard closely that is achieved by human subjects.

J. Wang, M. Betke, and J. P. Ko, "Pulmonary Fissure Segmentation on CT." Medical Image Analysis, 10(4):530-547, August 2006. PubMed Entry pdf.

Lung Cancer Detection

Margrit Betke, in collaboration with chest radiologist Jane P. Ko, developed the first algorithms for automatically detecting and measuring pulmonary nodule growth in CT (computed tomography) images. These growth measurements are essential for lung cancer diagnosis but are currently made by time-consuming, inaccurate and inconsistent manual methods. Facilitating the diagnosis of lung cancer is important, because early detection and resection of small, growing, pulmonary nodules can improve the 5-year survival rate of patients from 15% to 67%. Betke and Ko's method for automating the detection, comparison and volumetric quantification of pulmonary nodules was described in

M. Betke and J. P. Ko, "Detection of Pulmonary Nodules on CT and Volumetric Assessment over Time." In C. Taylor and A. Colchester, editors, Medical Image Computing and Computer-Assisted Intervention -- MICCAI'99. Second International Conference, Cambridge, UK, September 19-22, 1999, Proceedings. Lecture Notes in Computer Science, Volume 1679/1999, pp. 245-252, Springer-Verlag, Berlin. Abstract. pdf.
J. P. Ko and M. Betke, "Chest CT: Automated Nodule Detection and Assessment of Change over Time-Preliminary Experience." Radiology, 218, 267-273, January 2001. Link to publisher, pdf (restricted access).

and received the US patent

"Method and system for the detection, comparison and volumetric quantification of pulmonary nodules on medical computed tomography scans." Inventors Margrit Betke and Jane P. Ko., US Patent 7,206,462, issued on April 17, 2007.

In the last decade, Betke and her students at Boston University continued the collaboration with Dr. Ko, who moved to New York University Medical Center, The group developed detection, segmentation, and registration methods for the chest, lungs, fissures, and blood vessels, e.g.,

W. Mullally, M. Betke, J. Wang and J. P. Ko, "Segmentation of Nodules on Chest Computed Tomography for Growth Assessment," Medical Physics, 31:4, pp. 839-848, April 2004. pdf.
M. Betke, H. Hong, D. Thomas, C. Prince, J. P. Ko, "Landmark Detection in the Chest and Registration of Lung Surfaces with an Application to Nodule Registration." Medical Image Analysis, 7:3, pp. 265-281, September 2003. Link to publisher and pdf, PubMed Entry.

Additional information on the lung imaging project.

Margrit Betke, Associate Professor

Computer Science Department

Boston University

Last updated: August 3, 2010