Volume 10, Issue 1 (laser in medicine 2013)
lmj 2013, 10(1): 13-20 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Resolution improvement of multispectral fluorescent molecular tomography images using linear unmixing algorithm. lmj 2013; 10 (1) :13-20
URL: http://icml.ir/article-1-279-en.html
URL: http://icml.ir/article-1-279-en.html
Abstract: (10039 Views)
Background and Aim: Fluorscence Tomography is Emerging as an important alternative to molecular imaging. Simultaneous detection of several biological processes in vivo is a common requirement in molecular imaging. Application of multiple fluorophores and multispectral approaches have been widely used for distinguishing fluorescence signals. But, in case of spectral overlapping isolation of signals requires using unmixing algorithms. In this study, the researchers used blind source unmixing method, Non-Negative Matrix Factorization (NMF), to decompose contributions of different fluorescent probes in Fluorescence Molecular Tomography (FMT) images.
Methods: A cylinder tissue-like phantom containing 2 transparent tube with 0.5 millimeter inner diameter filled with different fluorophores (fluorescine and rhodamine). It was imaged by using experimental FMT setup at three different wavelengths. We performed NMF algorithm on multispectral measurement to resolve the relative contributions of fluorescent dyes. Multispectral projection was simulated and used as input of unmixing algorithms in simulation part of study. The projection images of each probes was simulated and used as reference data to evaluate performance of unmixing algorithm for simulated images.
Results: Two set of data, each of them corresponded to isolated contribution of each fluorophore, were obtained as the result of performing the algorithm on multispectral measurements. The "root mean square error" (RMSE) was calculated between original reference and related unmixed images. The difference between " Peak signal to noise ratio" (PSNR) values was statistically significant for each reference image (p<0.01). It shows that the proposed algorithm works in an accurate manner. The correlation coefficient between unmixed images and related reference image was more than 97% in simulation study. Unmixing method based on NMF can be used as an efficient technique to unmix contribution of different florophores. The 3-D location of each florophore can be obtained separately by reconstructing of these unmixed image.
Methods: A cylinder tissue-like phantom containing 2 transparent tube with 0.5 millimeter inner diameter filled with different fluorophores (fluorescine and rhodamine). It was imaged by using experimental FMT setup at three different wavelengths. We performed NMF algorithm on multispectral measurement to resolve the relative contributions of fluorescent dyes. Multispectral projection was simulated and used as input of unmixing algorithms in simulation part of study. The projection images of each probes was simulated and used as reference data to evaluate performance of unmixing algorithm for simulated images.
Results: Two set of data, each of them corresponded to isolated contribution of each fluorophore, were obtained as the result of performing the algorithm on multispectral measurements. The "root mean square error" (RMSE) was calculated between original reference and related unmixed images. The difference between " Peak signal to noise ratio" (PSNR) values was statistically significant for each reference image (p<0.01). It shows that the proposed algorithm works in an accurate manner. The correlation coefficient between unmixed images and related reference image was more than 97% in simulation study. Unmixing method based on NMF can be used as an efficient technique to unmix contribution of different florophores. The 3-D location of each florophore can be obtained separately by reconstructing of these unmixed image.
Keywords: Linear Unmixing, Fluorescence Molecular Tomography, Multispectral Imaging, Blind Source Separation, Nonnegative Matrix Factorization
Educational: Research |
Subject:
General
Received: 2013/09/25 | Accepted: 2015/02/15 | Published: 2014/05/5
Received: 2013/09/25 | Accepted: 2015/02/15 | Published: 2014/05/5
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
