Lens-Free Imaging
Project Summary
This project utilizes a microscopic imaging modality known as lens-free imaging (LFI) for a variety of biomedical applications. Lens-free imaging is based on the principle of illuminating a specimen object with a coherent light source (such as from a laser or laser diode) and then recording the resulting diffraction pattern with an image sensor. One then reconstructs an image of the object by solving an inverse model of the light diffraction process. As the name implies, LFI does not use lenses or other costly optical components, which allows for very compact and low-cost imaging systems. Additionally, LFI systems typically have a larger field of view than traditional microscopes, and the fact the image is reconstructed in software allows for images to be generated at arbitrary focal distances, eliminating the need for manual focusing of the microscopy. In this project we exploit these advantages of LFI combined with computer vision techniques for a variety of biomedical point-of-care applications.
Point-of-Care Blood Counting
One of the most widely ordered blood tests in the world is the complete blood count (CBC). A CBC consists of obtaining a blood sample from a patient and then quantifying the concentrations of the various types of blood cells (e.g., red blood cells, platelets, white blood cells and various subtypes of white blood cells) in the patient’s blood. Typical testing for a CBC consists of sending a blood sample to a centralized laboratory for analysis by a hematology analyzer machine. In this project we use LFI and computer vision techniques combined with microfluidics to perform a CBC with a miniaturized device that can be done directly at the point-of-care from a very small volume of blood (e.g., a single drop). Several goals of this project include developing techniques to reconstruct LFI images from the recorded holograms and detecting and classifying various blood cell types from the reconstructed images or from the hologram directly.
Automated Monitoring for Infections
Urinary tract infections (UTIs) are one of the most common hospital acquired infection due to the use of urinary catheters for hospitalized patients or for individuals in nursing homes. The goal of this project is to develop a compact and low-cost LFI device which is capable of automatically monitoring a urinary catheter line for a developing UTI by detecting various abnormalities in the urine stream that might be indicative of a developing UTI, such as bacteria or white blood cells, via the use of LFI and computer vision techniques. By enabling such early detection, this approach could potentially lead to a reduction in overall UTI prevalence and antibiotic usage by allowing clinicians to remove urinary catheters in patients who show initial signs of a developing infection before the infection reaches a point of requiring more significant interventions.
Reconstruction methods for lensless images
Lensless digital holography is achieved by illuminating a specimen with partially coherent light and recording its diffraction pattern (or hologram). Since the diffraction pattern is measured at the sensor plane, reconstruction algorithms are needed to retrieve relevant data at the plane of interest. In principle, back-propagating the measured signal to the specimen plane by deconvolving with the point spread function (PSF) of the system would lead to perfect reconstruction if the measured signal was the full complex-valued diffraction pattern and the PSF of the system was known. However, the forward model is in principle not invertible due to the loss of phase information at the sensor plane, which makes reconstruction algorithms highly prone to artifacts. To alleviate this issue, we designed an artifact removal method based on techniques from sparse dictionary learning and coding [1], which was published in the 2017 IEEE 14th International Symposium on Biomedical Imaging (ISBI 2017). We then continued leveraging the sparse prior, as specimens such as blood or urine correspond to reconstructed images in which most of their pixels represent background as opposed to objects, by proposing an efficient reconstruction algorithm that directly attempts to retrieve the (unmeasured) phase of the hologram [3]. This work was published in the International Conference on Medical Imaging Computing and Computer-Assisted Intervention (MICCAI 2017). The proposed method generates high-quality reconstructions in a fully unsupervised way, however itrequires accurate knowledge about the imaging conditions and the optical parameters of the system. Recently, we have proposed a more robust version of this reconstruction method that is able to adapt the PSF of the system in a data-driven fashion [8]. This work was published in Optics Express.
Acknowledgement
This work supported by NIH National Institute on Aging grant 5R01AG067396-02 and miDiagnostics.
Publications
[1]
B.D. Haeffele, S. Roth, L. Zhou, and R. Vidal.
Removal of the twin image artifact in holographic lens-free imaging by sparse dictionary learning and coding
IEEE Int. Smyp. on Biomedical Imaging (ISBI) 2017.
[2]
F. Yellin, B.D. Haeffele, and R. Vidal.
Blood cell detection and counting in holographic lens-free imaging by convolutional sparse dictionary learning and coding
IEEE Int. Smyp. on Biomedical Imaging (ISBI) 2017.
[3]
B.D. Haeffele, R. Stahl, G. Vanmeerbeeck, and R. Vidal.
Efficient reconstruction of holographic lens-free images by sparse phase recovery
Medical Image Computing and Computer-Assisted Intervention (MICCAI) 2017.
[4]
F. Yellin, B.D. Haeffele, and R. Vidal.
Multi-cell detection and classification using a generative convolutional model
IEEE Computer Vision and Pattern Recognition (CVPR) (*oral presentation - top 2.5% of submissions) 2018.
[5]
B.D. Haeffele, C. Pick, Z. Lin, E. Mathieu, S.C. Ray, and R. Vidal.
An optical model of whole blood for detecting platelets in lens-free images
MICCAI Simulation and Synthesis in Medical Imaging (SASHIMI) 2019.
[6]
F. Yellin, B. Bejar, B.D. Haeffele, E. Mathieu, C. Pick, S.C. Ray, and R. Vidal.
Joint holographic detection and reconstruction
MICCAI Machine Learning in Medical Imaging (MLMI) 2019.
[7]
B.D. Haeffele, C. Pick, Z. Lin, E. Mathieu, S.C. Ray, and R. Vidal.
Generative optical modeling of whole blood for detecting platelets in lens-free images
Biomedical Optics Express 2020.
[8]
C. Pacheco, G.N. McKay, A. Oommen, N.J. Durr, R. Vidal, and B.D. Haeffele.
Adaptive sparse reconstruction for lensless digital holography via PSF estimation and phase retrieval
Optics Express 2022.