Researchers from Rensselaer Polytechnic Institute developed a new image reconstruction method to directly reconstruct the intensity and lifetime images from raw time-resolved compressed sensed data
Macroscopic fluorescence lifetime imaging (MFLI) via compressed sensed (CS) measurements can be used to offer efficient and accurate quantification of molecular interactions in vivo over a large field of view (FOV). However, the approach is limited by slow data-processing workflow, which also is complex and performs poorly under photon-starved conditions. Now, a team of researchers from Rensselaer Polytechnic Institute developed a new technique using convolutional neural network (CNN). The new process can offer ultra-fast speed image reconstruction. The approach was described in the journal Light: Science and Applications on March 06, 2019.
CS-based imaging is a signal processing technique and can be used to create images based on a limited set of point measurements. In June 2017, the team reported developing an imaging platform based on time-resolved structured light and hyperspectral single-pixel detection. The platform can perform quantitative MFLI over a large FOV and multiple spectral bands simultaneously. This technique can be used to gain inclusive molecular data sets. Although the approach produced more complete images, it consumed more time in processing the data and forming an image. The current advancement in the study is based on the previous research and is capable of producing real-time images. It can also enhance the image quality.
According to the researchers, the breakthrough can promote further development of clinical diagnostics and personalized drugs. The technique can offer a complete snapshot of organs or tumors and can calculate the decay rate of the fluorescence to provide information about the effective intracellular delivery of drugs. The team developed a CNN architecture known as Net-FLICS: Fluorescence Lifetime Imaging with Compressed Sensing that promotes real-time visualization of molecular events. According to the researchers, further development may facilitate the use of this approach in a clinical setting.