A low-cost, high-throughput whole slide imaging system
Whole slide imaging (WSI) system is one important tool for biomedical research and clinical diagnosis. In particular, the advances of computer and image sensor technologies in recent years have significantly accelerated the development of WSI systems for high-content screening, telemedicine, and digital pathology.
We report the development of a high-throughput WSI system by adapting a low-cost add-on kit to existing microscopes. In the prototype setup, we can acquire a 2 gigapixel image with 14 mm by 8 mm field of view in 1.4 minutes. The reported platform may find applications in biomedical research, telemedicine, and clinical digital pathology. It may also provide new insights for the development of high-content screening instruments. The operation of this platform is shown in the following video:
This platform contains two modules: 1) autofocusing module and 2) mechanical scanning module. The autofocusing module is shown in Fig. 1 (a), where we attached two pinhole-modulated cameras at the eyepiece ports of a microscope platform. When the two pinholes are placed at off-axis positions (see Fig. 1(b)), the view angles are different for CCD1 and CCD2. As shown in Fig. 1(b2), when the sample is placed at the in-focus position, the two captured images will be identical. On the other hand, if the sample is placed at an out-of-focus position (Fig. 1(b1) and (b3)), the sample will be projected at two different view angles, causing a translational shift in the two captured images. The translation shift is proportional to the defocus distance of the sample, and thus, we can directly recover the optimal focal position by identifying the translational shift of the two captured images; no z-scan is needed.
Fig. 1: Pinhole-modulated cameras for instant focal plane detection. (a) By adjusting the positions of the two pinholes, we can effectively change the sample view angle. The optimal focal position can be recovered from the translational shift of the two corresponding images. (b) The schematic of the proposed autofocusing scheme. The translational shift of the images can be used to identify the optimal focal position. (c) An off-axis pinhole was inserted to the Fourier plane of the reduction lens in the eyepiece adapter. The pinhole was punched by a needle on a paper. A 3D-printed plastic case was used to assemble the autofocusing module. (d) The autofocusing module attached at the eyepiece ports.
The design of the pinhole-modulated camera is shown in Fig. 1(c1), where the off-axis pinhole is inserted into an eyepiece adapter (Amazon). This pinhole was created by piercing a hole through a black printing paper, as shown in Fig. 1(c2). We use a 3D-printed plastic case to assemble the two eyepiece adapters and two cameras, as shown in Fig. 1(c3). To use this autofocusing module, we can simply insert it into the eyepiece port of a regular microscope platform, as shown in Fig. 1(d). In the proposed platform, there are three cameras; two of them (pinhole-modulated cameras) are low-cost cameras attached to the eyepiece port for focal plane detection, and the main camera (Fig. 1(d)) is used to capture the high-resolution image of the sample without losing photons. For the pinhole-modulated cameras, the only requirement is to capture image in high speed; the size of image sensor and the total number of pixels are not important in the design. Therefore, low-cost, small-pixel-number webcams can be used in the proposed autofocusing module.
The mechanical scanning kit is important to achieve high-throughput performance in WSI. We note that the mechanical stages has been greatly optimized in conventional microscope platforms. We can easily move the sample with sub-micron accuracy using our hands. In our design, we will use stepper motors (NEMA-17, Amazon) to drive the knobs of the microscope stages. Depending on the gear-coupling systems, the positional accuracy can reach 1 um x- y scan and ~200 nm for z scan. As shown in Fig. 2(a), we used 3D-printed plastic gears to control the focus knob for autofocusing and the smallest z-step is 200 nm. If needed, one can change the size ratio of the two mechcentimeter range, and it can be further improved with better mechanical design. In Fig. 2(c), we demonstrate the use of an open-source, programmable robotic arm (uArm, Kickstarter, $300) for sample loading. The use of robotic arm is not a new idea. However, low-cost and open-source robotic arm is only available very recently. We can expand its capability for handling different samples and integrate other image recognition strategies for better and affordable laboratory automation. In the preliminary setup, we used Arduino microcontroller to control the mechanical scanning kit as well as the robotic arm.
Fig. 2: Sample loading and mechanical scanning in the InstantScope platform. (a) 3D-printed plastic gear for controlling the focus knob. (b) x-y scanning kit. (c) Sample loading using an open-source robotic arm. (d-e) Gigapixel images captured by using the reported WSI platform. (d) The captured image of a pathology slide using a 9 megapixel CCD. This image has a field of view of 14 mm by 8 mm and the acquisition time is 1.4 minutes. (e) A captured image of a blood smear using a color CMOS sensor. See also: http://gigapan.com/profiles/SmartImagingLab.
Fig. 2(d) and 2(e) show the gigapixel images captured using the reported platform. In Fig. 2(d), we used a 9 megapixel monochromatic CCD camera (at the main camera port) to capture the image of a pathology slide. Using a 20X, 0.75 numerical aperture objective lens, it took 1.4 minutes to acquire a 2 gigapixel image with 14 mm by 8 mm field of view. This image contains 340 segments, and the image acquisition of each segment takes 0.24 second using a regular desktop computer with an Intel i5 processor. The main speed limitation is located at the data readout from pinhole-modulated cameras. In this early prototype, we used an old camera model (31AU03, IC Capture, 1024 by 768 pixels). A better camera can reduce the acquisition time of single segment to 0.15 second (~40% improvement). In Fig. 2(e), we show a gigapixel color image of a blood smear captured by the proposed WSI platform.
The proposed multimodal WSI kit
Conventional WSI systems
|Autofocusing method||Instant autofocusing based on the pinhole-modulated cameras||Image-contrast methods; need to acquire images at different z positions to determine the optimal focal position|
|Autofocusing range||~1 mm based on the pinhole-modulated images||~10 um based on high-resolution image captured through the main camera port (surveying 3 point along the z-axis)|
|Adaptability for different samples||Phase correlation information can be used to infer the sample’s 3D structure without a z-scan; Scanning strategies can be adaptively changed during the scanning process||Scanning strategies are assigned prior to the scanning process|
|Imaging throughput||~ 1.4 gigapixel / min (40% improvement is expected)||1.0-4.0 gigapixel / min (with high-end automatic stages and highly synchronized electronics)|
|Cost||$100-$500 kit (depending on the cameras) as an add-on for regular microscopes, including three parts: 1) autofocusing module, 2) mechanical scanning module, and 3) illumination module using the LCD; If needed, additional $300 for the robotic arm platform||$20,000-$100,000|
|Imaging modalities||InstantScope: brightfield, fluorescenceBy adding the $15 LCD module, we can further achieve: darkfield, phase contrast, quantitative phase (based on Fourier ptychography), 3D tomography, polarization, and super-resolution Fourier ptychographic microscopy||Brightfield and fluorescence|