diff --git a/_pages/projects.md b/_pages/projects.md index 4ca239e..eaf3d8b 100644 --- a/_pages/projects.md +++ b/_pages/projects.md @@ -163,7 +163,6 @@ horizontal: false - @@ -171,8 +170,7 @@ horizontal: false
- Fourier ptychography (FP) is a powerful computational imaging technique that provides super-resolution and quantitative phase imaging capabilities by scanning samples in Fourier space with angle-varying illuminations. However, the image reconstruction in FP is inherently ill-posed, particularly when the measurements are noisy and have insufficient data redundancy in the Fourier space. To improve FP reconstruction in high-throughput imaging scenarios, we propose a regularized FP reconstruction algorithm utilizing anisotropic total variation (TV) and Tikhonov regularizations for the object and pupil functions, respectively. To solve this regularized FP problem, we formulate a reconstruction algorithm using the alternating direction method of multipliers and show that our approach successfully recovers high-quality images with sparsely sampled and/or noisy measurements. The results are quantitatively and qualitatively compared against various FP reconstruction algorithms to analyze the effect of regularization under harsh imaging conditions. In particular, we demonstrate the effectiveness of our method on the real experimental FP microscopy images, where the TV regularizer effectively suppresses the measurement noise while maintaining the edge information in the biological specimen and helps retrieve the correct amplitude and phase images even under insufficient sampling. Our proposed algorithm achieves high-quality FP reconstruction along with the improvement of the speed and the cost of the FP data acquisition process without sacrificing image quality. We believe that our FP reconstruction algorithm will be of great interest to the optics and imaging community, particularly as it further impacts various extensions of FP as well as real-space ptychography, which share a common algorithmic -foundation. + Fourier ptychography (FP) is a powerful computational imaging technique that provides super-resolution and quantitative phase imaging capabilities by scanning samples in Fourier space with angle-varying illuminations. However, the image reconstruction in FP is inherently ill-posed, particularly when the measurements are noisy and have insufficient data redundancy in the Fourier space. To improve FP reconstruction in high-throughput imaging scenarios, we propose a regularized FP reconstruction algorithm utilizing anisotropic total variation (TV) and Tikhonov regularizations for the object and pupil functions, respectively. We have demonstrated the effectiveness of our method on the real experimental FP microscopy images, where the TV regularizer effectively suppresses the measurement noise while maintaining the edge information in the biological specimen and helps retrieve the correct amplitude and phase images even under insufficient sampling.
+ The reduction of the size and cost of microscopes creates new possibilities for disease screening and health care in resource-limited settings. Smartphones and mobile devices can be useful for the construction of compact and portable imaging devices, by replacing expensive and bulky optical instruments with state-of-the-art cameras, high-performance processors, and network connectivity in a compact footprint. We propose a smartphone-mountable computational microscope that achieves wide-field and high-resolution imaging, based on the Fourier ptychographic (FP) microscopy technique. Our device uses the smartphone’s built-in camera for microscopy imaging, the display screen for programmable illumination, and the computational power of the smartphone processors for FP image reconstruction. Our device achieves a half-pitch resolution of 870 nm over a wide field-of-view of 2.1 × 1.6 mm² with amplitude, phase, and color imaging capabilities. As global health issues continue to persist and new epidemic challenges arise, we believe that our smartphone microscopy technology can effectively address the clinical needs in point-of-care environments. +
+- The reduction of the size and cost of microscopes creates new possibilities for disease screening and health care in resource-limited settings. Smartphones and mobile devices can be useful for the construction of compact and portable imaging devices, by replacing expensive and bulky optical instruments with state-of-the-art cameras, high-performance processors, and network connectivity in a compact footprint. We propose a smartphone-mountable computational microscope that achieves wide-field and high-resolution imaging, based on the Fourier ptychographic (FP) microscopy technique. Our device uses the smartphone’s built-in camera for microscopy imaging, the display screen for programmable illumination, and the computational power of the smartphone processors for FP image reconstruction. With a compact and lightweight optics module and a custom-built application, we can transform a regular smartphone into a high-performance microscope device with a space-bandwidth product exceeding that of regular benchtop microscopes. Our device achieves a half-pitch resolution of 870 nm over a wide field-of-view of 2.1 × 1.6 mm² with amplitude, phase, and color imaging capabilities. As global health issues continue to persist and new epidemic challenges arise, we believe that our smartphone microscopy technology can effectively address the clinical needs in point-of-care environments. -
-- Virtual staining techniques now enable the transformation of label-free images into clinically standard stained images; however, the extensive costs and time requirements of labelled dataset generation impede its widespread use in clinical settings. We aim to implement semi-supervised learning for virtual staining to enable high-fidelity results while reducing the need for labor-intensive dataset generation. To achieve this, we designed a neural network architecture that digitally stains label-free images of unstained tissue by minimizing style feature differences between the outputs of stained and unstained raw images. We have compared the output from the virtual staining network with various training strategy including supervised training with end to end and cascaded architecture, unsupervised training, and our semi-supervised training. Additionally, for consistency in comparison, we prepared a unstained slide and a well-stained slide from the same tissue block. While minor differences in cellular features may exist between the slides, our methods exhibits the best performance among the methods as well as their overall tissue structures remain comparable. Our approach is particularly valuable for computational microscopy techniques that enable various label-free imaging applications, where generating paired datasets is challenging, such as in 3D pathology and immunohistochemistry staining. + Virtual staining techniques now enable the transformation of label-free images into clinically standard stained images; however, the extensive costs and time requirements of labelled dataset generation impede its widespread use in clinical settings. We aim to implement semi-supervised learning for virtual staining to enable high-fidelity results while reducing the need for labor-intensive dataset generation. To achieve this, we designed a neural network architecture that digitally stains label-free images of unstained tissue by minimizing style feature differences between the outputs of stained and unstained raw images. Our approach is particularly valuable for computational microscopy techniques that enable various label-free imaging applications, where generating paired datasets is challenging, such as in 3D pathology and immunohistochemistry staining.