The optical fiber taper coupled with CMOS has advantages of high sensitivity, compact structure and low distortion in the imaging platform. So it is widely used in low light, high speed and X-ray imaging systems. In the meanwhile, the peculiarity of the coupled structure can meet the needs of the demand in microscopy imaging.

Toward this end, we developed a microscopic imaging platform based on the coupling of cellphone camera module and fiber optic taper for the measurement of the human blood samples and ascaris lumbricoides. The platform, weighing 70 grams, is based on the existing camera module of the smartphone and a fiber-optic array which providing a magnification factor of ~6x.The top facet of the taper, on which samples are placed, serves as an irregular sampling grid for contact imaging.

The magnified images of the sample, located on the bottom facet of the fiber, are then projected onto the CMOS sensor. This paper introduces the portable medical imaging system based on the optical fiber coupling with CMOS, and theoretically analyzes the feasibility of the system. The image data and process results either can be stored on the memory or transmitted to the remote medical institutions for the telemedicine. We validate the performance of this cell-phone based microscopy platform using human blood samples and test target, achieving comparable results to a standard bench-top microscope.

In recent years, along with progress about low light level image technology, the optical fiber coupling technique has been developed rapidly. The optical fiber taper coupled with CMOS has advantage of high sensitivity, compact structure and low distortion. So it is widely used in low light, high speed and X-ray imaging systems.
The research work has been used in optical fiber coupling technique and CMOS imaging performance test to make virtual progress in resolution, construction, power consuming, and integrating features, ultimately to provide a low light optical fiber image systems to be equipped in defense weapons and law enforcement equipment.

There has been much effort to develop a mobile phone microscope. The system of Breslauer et al. achieves high resolution by using the smartphone without the algorithm optimization, which limited the resolution improvement and further application. The smart-phone based computational microscopy with the attachment of a fiber-optic array used by Ozcan, et al. can image dense or connected samples without holographic reconstruction. The structure is of small size at the expense of low magnification and restriction by the sophisticated optical lens group.

This paper introduces a microscopic imaging platform using the CMOS sensor in smartphone coupling with the optical fiber, and theoretically analyzes optical coupling feasibility between the optical fiber taper and CMOS. Then we test the resolution of the platform by USAF 1951 resolution target.

Theories and experiments have been extended and the coupling techniques have been developed in the paper. The performance and utility of microscopic device with optical fiber coupled CMOS sensor have been improved during the research work.

Before the coupling of the taper and CMOS, the protective surface of the CMOS photosensitive surface should be removed. In the optical imaging system, coupling medium need to be added between the optical imaging transmission devices to improve the image quality, and to effectively reduce or avoid the imaging loss due to the energy loss during the light transmission. Since the optical images are projected onto the CMOS sensor after passing through the optical taper, reflection and refraction will damage the imaging quality if there is an extra layer of space between the two optical devices. Therefore, a coupling material (photosensitive adhesive) with a refractive index of~1.5 (close to the refractive index of glass) between the imaging devices need to be utilized to avoid the above situation (Fig.1).

Fig 1. (A) Schematic diagram of the cellphone attachment of the Contact Scope is demonstrated. (B)The micro-objects are placed in contact with the small facet of taper

The microscopic imaging platform based on fiber-optic array coupled CMOS module (Fig.2) utilizes a coupling medium to form a closed optical waveguide between the CMOS photosensitive surface and the large facet of a tapered fiber-optic array. The mechanism reduce the scattering loss of light, and make the pixels on the CMOS photosensitive surface correspond to the pixel of the imaging device connected to the large end of fiber-optic array to provide a high image resolution and clarity.

Fig 2. (A, B, C) Photographs of the portable microscope platform are shown. (D) Resolution test of the microscope platform

In this microscope platform, the imaging system is designed using aluminum to strengthen the robustness and reduce the weight. The mobile phone microscopy utilizes Nokia Lumia 1020 (f_built-in=7.2mm, N_{realtive aperture}=f/2.2) for the image acquisition, while the length is 110mm, and the weight is ~160g. The CMOS sensor of the smartphone equipped with 41 megapixel (7728×5368 pixels), yielding a ～1.12um pixel spacing is chosen. A commercially available broad-spectrum LED flashlight powered by watch batteries provides illumination. The micro-objects are placed in contact with the small facet of taper, and illuminated by an incoherent light source (LED). A focusing lens(OP005) is mounted to the LED to create uniform illumination over the entire FOV of the sample. The filter is used between focusing lens. A transmission image from LED through the focusing lens and the filter is magnified by the objective lens, and then delivering the image of the sample onto the CMOS sensor.

As we know, the resolution of the images is proportional to a high density of grid of the small pixels. Continuous scene will degrade by some factors of down-sample, warping and blurring.

Although we have chosen a CMOS sensor with small pixel size of 1.12 um, a shift-and-add reconstruction algorithm is still necessary to create a high-resolution image. Effectively, the reconstruction approaches are based on over-sampling in the time domain (obtaining a series of images) to compensate for under-sampling in the spatial domain, where the resolution is limited by the image sensor pixel size. Multiframe pixel super-resolution approaches rely on combing information from multiple low-resolution images to reconstruct a high-resolution image. We acquire several raw images by the smartphone (e.g, 50).

The algorithm obtains the under-sampled original images and estimates the relative motion between the reference image and each frame. Then, the low-resolution images are rearranged to reconstruct into a single high-resolution matrix. Finally, deburring process (As we know, the resolution of the images is proportional to a high density of grid of the small pixels. Continuous scene will degrade by some factors of down-sample, warping and blurring.

Finally, deburring process (depending on the observation model)is preformed to reduce the effects of blurring and noise caused by the system. It takes less than 3 minutes to obtain the raw images (including the time of capturing 50 images) (Fig 3).

Fig. 3 Process flow for Shift-and-add reconstruction performed on personal computer with an MATLAB program

The full-field high-resolution images (1280×1024) are still too large to be restored by the algorithms. However, we can zoom to an interested area to reconstruct the images. We reset the inherent white balance to obtain the raw images to dispose the background details.

To test the Resolution of the microscope platform, we directly quantified resolution of the system using images of a 1951 USAF resolution target taken in monochrome light. 100 images were used to perform 8×8 enhancement by rounding the sub-pixels shifts to integer multiples of 1/8 of a low resolution pixel. The images are taken by tilting the glass slide with the samples, and then converted into a binary image to highlight the dark region and to remove the noises in the images. Next, we reconstructed the images using the shift-and-add algorithm. The measured intensity profile across the bars of the group7, element 6 on the portable microscopy is shown in Fig. 4.

Fig. 4 Resolution test of the microscope platform. Where the (A, B) Measurements of resolution achieved by the platform. The resolution measurements are based on the smallest resolvable group of a 1951 USAF resolution target. (A) Thee raw test images of the resolution target. (B) The images after Shift-and-add reconstruction. The measured intensity profile across the bars of the group 7, element 6 on the portable sleeve microscopy

We carried out the experiment with the prototype of blood cell and ascaris lumbricoides (Fig. 5). Ascariasis is a common parasitic disease epidemically in the developing countries that infect up to several billion people especially for the children at risk from inappetence, nausea and emesis. It is of great inconvenience for the people in the remote area where medical level cannot meet the requirements. The portable microscopy is proved to be useful for the routine testing ascaris eggs.

Fig. 5 Performance of the microscope platform. Where the (b, d) Measurements of blood cell and ascaris lumbricoides achieved by the platform. (a, c) The raw test images of blood cell and ascaris lumbricoides by a 1000 Digital Microscope（Keyence VH-Z100RRZ100). All the scale bar is 10μm

We have reported on a portable microscopic imaging platform based on the coupling of cellphone camera module and fiber optic taper. The microscope can initially image the blood cell and ascaris lumbricoides from the patients and the green algae from the red tide. Moreover , we have demonstrated that the light platform can improve the resolution of 1951 USAF resolution target based on the shift-and-add reconstruction algorithm.

The main advantages of the system are its simplicity and stability, which will create a significant improvement in the remote areas and developing countries. Furthermore, using smartphone with a custom-developed Android application will enhance the portability. Moreover, some other algorithms such as shift-and-add algorithm can be used to extend to higher resolution in telemedicine.