Open-source software for automatic segmentation of lung in CT images. The software is an segmentation tool for the segmentation of lung from CT images. The sofware can be run in two modes: (a) fully automatic and (b) semi-automatic with manual seeding by the user. The software also allows the user perform basic filtering operations and manual correction to the segmentation, if needed. The VTK-based rendering implemntation along with option to view in axial, coronal, and saggital provide user with better visualization of the segmented lung. Please cite the following study when you use this software:
A Generic Approach to Pathological Lung Segmentation.
A.Mansoor,U. Bagci, Z. Xu, B.Foster, K. Olivier, J. Elinoff, A.F. Suffredini, J.K. Udupa, and D. J. Mollura
IEEE Transactions on Medical Imaging, 33(12):2293-2310, 2014.
Presented here is a GUI for segmenting and quantifying PET images with multi-focal and diffuse uptakes as commonly seen in pulmonary infections. The segmentation algorithm was presented at 2013 IEEE International Symposium on Biomedical Imaging (ISBI) and published in the IEEE Transactions on Biomedical Engineering (highlighted in cover of the journal). For validated the software, we have used rabbit models infected with Tuberculosis (TB), where diffuse and multi-focal uptake were common to observe.
The GUI imports a PET image (either Analyze format or Matlab format) and allows the user to draw region of interests (ROIs) in 2D or 3D to roughly separate the object of interest from the background PET image. Then, once the ROI or multiple ROIs have been selected, the areas are segmented using a PET image segmentation method based on Affinity Propagation clustering to cluster the image intensities into meaningful groups.
For quantification, the software calculates the Standardized Uptake Value (SUV) of the binary or ROI which is the standard quantification metric widely used in the clinical and research environment. The parameters needed for accurate SUV quantification are inputted by the user and clearly shown in the GUI. The SUVmax, SUVmean, and Volume (mm^3) of the pathologies are calculated and can be exported into an excel sheet. In this excel sheet, the quantification metrics are also split by the groups found by the segmentation method. We believe that there is meaningful information in the secondary groups, not just the highest uptake group which is the current standard. Renderings with the functional information overlaid can also be made using the GUI for visualization purposes. A detailed instruction pdf is provided in the zip folder. DOWNLOAD
The software is based on the following two publications. Please consider citing them when you use the software:
Segmentation of PET Images for Computer Aided Functional Quantification of Tuberculosis in Small Animal Models.
B. Foster, U. Bagci, Z. Xu, B. Dey, B. Luna, W. R. Bishai, S. Jain, and D. J. Mollura
IEEE Transactions on Biomedical Engineering, 61(3):711-724, 2014.
Robust Segmentation and Accurate Target Definition for Positron Emission Tomography Images Using Aﬃnity Propagation.
Foster, B., U. Bagci, B. Luna, B. Dey, W. Bishai, S. Jain, Z. Xu, and D. J. Mollura
In: IEEE 10th International Symposium on Biomedical Imaging (ISBI), 1461-1464, 2013.
Quantitative Analysis and Visualization of PET Images (QAV-PET) is an open-source software implemented in the popular MATLAB coding environment that allows easy, intuitive, and efficient visualization and quantification of multi-modal medical images.In particular, the software is well suited for PET-CT and MRI-PET images. It allows multi-modal images to be viewed simultaneously which allows the research to incorporate information that is not available on a single modality image for improved quantification and analysis. Additionally, the software includes a robust framework for quantification and analysis which can be easily manipulated and tuned to fit any medical imaging application. Notably, this software includes a novel auto-reporting feature which, in an automated fashion, produces a report which includes the most important information needed for quantifying the disease and high uptake regions. It is meant to complement, not replace, the exported quantification as a .CSV file.
Please refer to the included QAV-PET User Guide.pdf for detailed instructions and potential applications for this software. DOWNLOAD.
The software is based on the following publication. Please consider citing them when you use the software:
QAV-PET: Quantitative Analysis and Visualization of PET Images.
B.Foster, U. Bagci, G.Papadakis, D.J. Mollura
In: Conf Proc IEEE Eng Med Biol Soc, 1909-1912, 2014.
A Gaussian mixture model is a probabilistic model that assumes all the data points are generated from a mixture of a finite number of Gaussian distributions with unknown parameters. One can think of mixture models as generalizing k-means clustering to incorporate information about the covariance structure of the data as well as the centers of the latent Gaussians. Below find the C/C++ code for GMM implementation. I have used this implementation in many different tasks including image classification and musical genre recognition. Cite the following article when you use this code:
Automatic Classiﬁcation of Musical Genres Using Inter-Genre Similarity.
U. Bagci, E. Erzin
Signal Processing Letters, 8(14):521-524, 2007.
[PDF] [gmm_train.c] [test_gmm.c]
| Mailing address:
Dr. Ulas Bagci
Center for Research in Computer Vision (CRCV)
4328 Scorpius Street, HEC 221, UCF
Orlando, Florida 32816, USA.
Last updated September, 2015 by Ulas Bagci.