Quantitative phase microscopy (QPM) has been widely applied due to its high contrast, label-free, and non-invasive characteristics. In recent years, QPM has made significant progress and is extensively used in the biomedical field. These QPM systems typically feature integrated structures and high-precision phase imaging capabilities. However, their high costs and dependence on optical platforms limit their application in portable devices. Most existing portable multimodal imaging microscopes do not improve phase reconstruction accuracy by performing noise quantitative estimation and denoising on images. Higher phase precision can help more accurately identify pathological features. Noise in differential phase contrast (DPC) images, during the reconstruction process based on phase transfer function (PTF) deconvolution, causes phase distortion and high-frequency ringing artifacts, which greatly limits the phase reconstruction accuracy of portable QPM systems. To address the aforementioned issues, we propose a low-cost, portable, multimodal imaging microscope with high-precision phase imaging capabilities and apply noise quantitative estimation and elimination algorithms to the captured images for denoising, thereby enhancing the accuracy of quantitative differential phase contrast (qDPC) phase reconstruction.

The multimodal imaging microscope proposed in this study uses an LED array as the illumination source. By controlling the Arduino Mega 2560, the light-emitting status of each LED unit is adjusted to display various lighting patterns, such as circular illumination, annular illumination, and asymmetric semi-circular illumination at different angles. This simulates the effect of adjusting the illumination pattern through the aperture diaphragm in traditional microscopes, thereby constructing a multimodal imaging microscope that integrates bright field (BF), dark field (DF), traditional differential phase contrast (tDPC), and color-multiplexed differential phase contrast (cDPC) imaging. The light emitted from the LED array illuminates the sample, and the light is collected by the objective lens. After passing through the reflecting mirror and the achromatic lens, the light is imaged onto the CMOS camera to complete the image acquisition process. An algorithm based on noise quantitative estimation and removal is applied, which quantitatively estimates the standard deviation of image noise and performs denoising on the image using a three-dimensional block-matching algorithm based on this standard deviation.

We conduct qDPC experimental tests on the USAF1951 QPT (Fig. 3), and it is found through the experiment that the resolution of the multimodal imaging microscope designed in this research can reach 1.10 μm, which is close to the theoretical resolution limit of the system, 1.03 μm. The difference between the actual resolution and the theoretical value may stem from a combination of factors such as the machining tolerances of the multimodal imaging microscope components and the assembly tolerances of the overall structure. After denoising with a noise quantitative estimation and removal algorithm, the original resolution is retained, which effectively preserves the details of the original image. By comparing a blank area and performing a standard deviation analysis, it is evident that the phase fluctuation is significantly reduced after denoising, which effectively suppresses the effect of noise on the image. The phase reconstruction of the second to fourth line pairs of the sixth group of the USAF1951 QPT indicates that after denoising with the noise quantitative estimation and removal algorithm, the phase reconstruction accuracy of the image is effectively improved, thus enhancing the accuracy and fidelity of phase reconstruction. Experiments on a microlens array (Fig. 5) show that the phase feature curve obtained by qDPC almost completely coincides with the true phase feature curve, demonstrating good phase reconstruction accuracy. This result indicates that the multimodal imaging microscope, under the qDPC imaging mode, can perform high-fidelity reconstruction of the phase information of transparent samples with excellent accuracy, which validates the effectiveness of the microscope in phase imaging of transparent samples. In addition, biological experiments have verified the BF, DF, and cDPC imaging functions of the multimodal imaging microscope (Fig. 6), which shows that the microscope can meet the imaging needs of various samples. By comparing the blank areas of BF, DF, and cDPC images before and after processing with the noise quantitative estimation and removal algorithm (Fig. 7), the standard deviation in the blank areas of all three imaging modes is reduced after processing, which indicates that the noise quantitative estimation and removal algorithm has good denoising effects and wide applicability. Finally, qDPC imaging is performed on water flea antennae and onion epidermal cells (Fig. 8), and by comparing the phase reconstruction curves before and after denoising with the noise quantitative estimation and removal algorithm, the phase feature curves become smoother after denoising, which is more consistent with the actual morphology. This further indicates that the noise quantitative estimation and removal algorithm also shows good effects in qDPC imaging of biological samples.

In this study, we successfully develop a low-cost, portable, multimodal imaging microscope with high-precision phase reconstruction capabilities. The microscope measures 300 mm×240 mm×310 mm, which makes it easy to move close to the sample for observation and greatly enhances its portability. Furthermore, by using an LED array as the light source, various imaging modes such as BF, DF, cDPC, and qDPC can be achieved simply by changing the LED array’s illumination patterns. This design avoids the addition of extra optical components, effectively reducing production costs. Finally, introducing a noise quantitative estimation and removal algorithm significantly improves image quality and phase reconstruction accuracy.