Article Text


Original article
Mobile Whole Slide Imaging (mWSI): a low resource acquisition and transport technique for microscopic pathological specimens
  1. Louis Auguste,
  2. Dhaval Palsana
  1. NYU Polytechnic School of Engineering Incubator, Brooklyn, New York, USA
  1. Correspondence to Louis Auguste, NYU Polytechnic School of Engineering Incubator, 20 Jay Street, Suite 312, Brooklyn, NY 11201, USA; lou.aug{at}


The Open Mobile Telepathology System (OMT) is a combination of two components, the Pocket Electronic Health Record (pEHR) and the Mobile Whole Slide Imaging (mWSI) app. This system was created over the course of this study to help reduce the cost of telepathology in developing countries. The affordable system as described will help to expedite the diagnostic process in low-resource environments and provide more patients with better health outcomes. The OMT system offers a number of advantages to the standard Whole Slide Imaging machines. It can be deployed at a fraction of the cost, the images are more easily transportable at an average image size of less than 500 MBs and any worker, even someone without any knowledge of pathology, can perform the scans; also, components can be replaced and upgraded at a minimal cost, offering a large advantage to low resource and rural environments. The OMT system utilises a standard light microscope with a custom built 3D adaptor and the iPhone 5s. Acquired mWSIs can be transported through the cloud using the pEHR database and accessed through web and mobile platforms. Therefore it can be used in any part of the world with an Internet connection. The study follows the step-by-step process used to acquire mWSIs of various stains including H&E stains, and thin preps. The results of the tests have been found to be of a diagnostic quality and the imaging process as described has been optimised and standardised over the course of the 2-year study.

  • Cancer
  • Diagnostics
  • mHealth
  • Imaging
  • Global Health

Statistics from


Laboratory testing plays an integral role in modern health care. The College of American Pathologists (CAP) estimates that between 60–70% of all diagnosis depends on pathology, including the early detection and diagnosis of cancer. Developing nations often lack the resources and medical personnel necessary to provide patients with quick diagnosis and prompt medical treatment. This shortage of pathologists creates massive disparities in health outcomes between nations.

Converting specimen glass slides into digital images, which then are accessible for remote analysis using viewing software, offers a substantial benefit in these settings.1 ,2 However, the systems used for digital Whole Slide Imaging (WSI) and TelePathology cost on average between $30 000 and $300 000. A cost that is unaffordable in low resource environments. Additionally, WSI files are often 2 GBs or larger, requiring a substantial amount of memory to store and discouraging transfer of data due to bandwidth limitations, thereby restricting the uptake of current WSI scanners and the deployment of TelePathology services in developing nations, and smaller healthcare facilities.

Mobile apps are already providing a significant reduction in the cost of teleconsultation.3 However, the current apps use static telepathology, which only provides a limited view of the specimen. Our hypothesis was that expensive WSI systems could be replaced with a suite of mobile applications that could accomplish the same tasks in a repeatable manner, with diagnostic quality, small file sizes and at a fraction of the cost. In researching this paper, we consulted sources on using mobile phones for static imaging,4 and accuracy of WSI and TelePathology for distance diagnosis,5 but we have found no information on using mobile phones to acquire and transmit WSIs for diagnosis.

Materials and methods

Traditionally, WSI's have been created using two patented techniques, Grid6 or Linear.7 This paper documents a new technique—panorama based Mobile WSI (mWSI).8

The mWSI system used for acquisition in this experiment seen in figure 1, consists of a standard light microscope for holding a specimen slide (101), an iPhone 5s for imaging the slide through the microscope (102), and our own 3D printed adaptor configured to receive and position the phone's camera on the eyepiece of the microscope 103.

Figure 1

Image acquisition hardware. 101=Light microscope with standard glass slide. 102=iPhone 5s. 103=3D printed adaptor.

We tested a number of different cell phones including the iPhone 4, iPhone 5s, Sony Z2 and the Samsung Galaxy S3, and have determined the optimal phone for panoramic image acquisition is the iPhone 5s and higher. This is thanks to Apple's use of full-resolution image capture in the panorama software and 8 megapixel camera, while other mobile devices offer more megapixels all other devices offered scaled down panorama image capture (note: Smart Phone recommendation represented best models available at time of research project).

The following is a method of performing manual whole slide imaging by acquiring multiple panoramic images to obtain a single composite image. Attach the adaptor to the microscope eyepiece and position the mobile imaging and communication device in the adaptor. Adjust the microscope to the desired objective lens. We determined that the optimal objective was either 10× or 20×, and the optimal ocular was either ×10 or 15×. After the desired magnification is set, the user may start to acquire panoramas of the sample. Acquisition may be performed by translating the sample stage so that it follows a zig-zag pattern. It should be noted that the sample viewable in the display is only a portion/subset of the entire sample.

Starting from the selected corner, the panorama image acquisition programme is activated and the stage is translated along the X-direction (eg, from left to right) with no motion along the Y-direction until the end of the specimen or an entire panorama image with dimensions of 10 800×2590 pixels is acquired. The panorama image acquisition programme used in this study was Apple's Camera App with the PANO setting selected. Once the panorama image is acquired, the user ceases translating the stage and the image acquisition programme stores the acquired image in memory. If the length of the sample to be imaged along the X-direction is longer than the length that can be captured in a single panorama image, the user may again activate the image acquisition and then continue translating the stage in the X-direction. This process may be repeated multiple times until the entire desired sample along the X-direction has been imaged. To improve the accuracy of the image-stitching programme that will later combine the acquired panorama images, each panorama image should have some overlap with an adjacent image. That is, for two adjacent panorama images obtained while translating the stage along the X-direction, there should be at least a 15% overlap on all sides.

After the scan along the X-direction is completed, the sample stage is translated along the Y-direction (eg, up or down) to a new row for a new series of image acquisitions. The next series of image acquisitions may begin from this new starting position.

Eventually, the user will have obtained one or more panorama images for multiple rows across the sample. Figure 2 is an arrangement of panoramic images obtained using the foregoing process. In particular, the images in figure 2 were obtained by starting a scan along a first row (201) to obtain a first image (202) of the sample. Then, the stage was translated to a second row (203), where another image (204) was obtained. Subsequently, the stage was translated again to a third row (205), where further images (206 and 207) were obtained.

Figure 2

Individual panoramas ready for image to image registration.

After the registration, the resulting images are added to the vignette programme that we created using Python and OpenCV. This programme removes the black matte that surrounds each image (vignette program will be available for free download at The vignetted images are then combined into a single composite image using PTGUI.

Subsequently, the composite image is compressed and wrapped with metadata. An example of wrapping the composite image with metadata includes saving the image file according to the Digital Imaging and Communications in Medicine (DICOM) standard. DICOM is a known standard for handling, storing, printing and transmitting information in medical imaging, and includes a file format definition and a network communications protocol.

Once a composite image is obtained and wrapped with metadata, a user may store the composite image on the smart phone and then transport it to another user using the Pocket Electronic Health Record (pEHR) application displayed in figure 3. For instance, the user operating the Mobile WSI (mWSI) system may be a technician in an isolated part of a country where there are few or no pathologists available for analysing the specimen. The technician may send the composite image of the specimen to a pathologist in another part of the country or in a different country, to obtain analysis of the imaged specimen. As a result, the amount of time that it takes to receive a diagnosis and analysis of the specimen may be substantially reduced. Such a reduction in diagnosis time is crucial when determining how to treat a patient with an unknown ailment or disease. Reporting may be communicated back to the technician from the analysing pathologist through the pEHR application as well.

Figure 3

Pocket Electronic Health Record (pEHR) mobile application. Horizontal image=slide viewer. Vertical image=patient record viewer.


The preliminary research was conducted at University College of London Hospitals (UCLH) using slides, which had been scrubbed of patient data, provided by BioBank. The first mWSI was created using an objective of 4× and an ocular of 15×. This failed to produce an image of diagnostic quality due to the low objective strength. In the second attempt, the objective was increased to 10× with the same 15× eyepiece. Figure 4 is the first successful mobile whole slide image. It is an H&E stain of a sample of microvilli. The mWSI contains an image pyramid and the screener is able to zoom in on mobile devices by using the pinch gesture and on computers by using the standard zoom in programmes such as Photoshop and Preview. When presented to members of the UCLH pathology team, the general consensus was that the image was of a diagnostic quality and offered a maximum magnification of 150× or better.

Figure 4

First successful Mobile Whole Slide Imaging (mWSI) scan. Insert image represents actual image pyramid magnification of main image.

Figure 5 is a thin prep pap test captured with a 10× objective and 15× eyepiece, which was our first successful cytology specimen. The digital image contains a similar image pyramid structure to figure 4 and the final tiff file was 496.5 MBs. It is important to note that the thin prep slide contained only a single z layer. This image was shared with members of the cytology staff at UCLH and, again, was determined to be of a diagnostic quality. In this first study, all 10 mWSI were created using Photoshop to stitch the images with no effort made to vignette the individual panoramas before importing them into Photoshop.

Figure 5

First Successful Mobile Whole Slide Imaging (mWSI) cytology scan—Insert image represents actual image pyramid magnification resolution of main image.

A year later, in September 2014, we continued our research across the pond at the North Shore/Long Island Jewish Medical Center in New York. In the interim, the workflow was refined, our 3D printed adaptor was optimised and the vignette removal programme was coded. Photoshop proved to be both unreliable and unstable. It frequently crashed or froze during everyday use, and so it was abandoned for the more robust image-stitching programme PTGUI. The result was a marked improvement from the previous workflow, including an increase of stitching speed of approximately 8×. Large specimens of 1.5×1.5 cm could be acquired and stitched on average in 45 min and the resulting images were between 400 and 500 MBs. Smaller specimens of 0.5×0.5 cm would take less than 15 min. All specimens were acquired using the iPhone 5s, 3D printed adaptor and a light microscope with an objective set to 20× and an eyepiece of 10×. Giving the digital images a 200× magnification or better. All specimens were of H&E stains, chosen because it is the most widely used stain in medical diagnosis (see figure 6). When presented to pathologists, all specimens were deemed of a diagnostic quality and a full validation study of the 60 H&E specimens is now underway.

Figure 6

Optimised Mobile Whole Slide Imaging (mWSI) scan of large pathology specimen with H&E stain. Insert image represents image pyramid magnification of main image.


According to the WHO, over 60% of the world's cancer deaths occur in Africa, Asia, and Central and South America. Among these deaths, cervical cancer ranks as the leading cause of mortality among women of reproductive age because more than 80% of these women have not been adequately screened for this disease.9

Haiti has the highest incidence of mortality from cervical cancer in the world, 94 in 100 000.10 A rate that is 50 times higher than the USA. In Haiti, there are approximately five pathologists for every 10 million persons, whereas in the USA, there are about five pathologists for every 90 000 people. Owing to the limited number of pathologists in places such as Haiti, analysis and reporting of test results from pathologists to a patient or the patient's doctor may take many weeks or months, if it even happens at all. This is the type of low resource environment where we envision the OMT system can make an immediate difference to health outcomes. For this reason, the Justinien Hospital in Cap Haitien was chosen for the pilot TELL programme (Test Early Live Longer), which started in 2013.

This is only fitting because the idea for OMT began in January of 2010, when Mr Auguste was deployed to Haiti in response to a massive 7.0 magnitude earthquake. Each of the emergency medical organisations at the main hospital created their own set of paper notes, and doctors rotated weekly, causing redundancies to abound. The first element of OMT, pEHR, is a cloud based Electronic Health Record (EHR) that can be easily deployed and accessed in multiple languages. During subsequent trips, the substantial lack of pathology services in the country became clear, giving rise to the low resource mWSI system. mWSI pairs with pEHR and puts the power of the laboratory in your pocket.

To deploy the mWSI system in low-resource environments such as Haiti, the process needed to be low cost, of a diagnostic quality, transportable, standardised and repeatable. During this study, we determined that the mWSI system achieves many of the stated goals. If none of the components were to exist in a potential laboratory setting, the materials needed to set up an OMT system would cost less than $3000. However, non-availability of materials is rarely the case, because there is nearly one smart phone for every five people on earth,11 and many labs already contain a light microscope and a computer, which brings the cost of the system down to less than $300.

In regard to having a diagnostic quality, the slides were presented to 10 different pathology and/or cytology specialists. Each specialist agreed that by looking at the digital mWSIs they could attain a diagnosis. Based on this conclusion, we plan to continue our research with a series of robust validation studies. These studies will help prove the accuracy of diagnosis that can be attained from using mWSIs with different stains.

This study also proved that the slides could be transported and the process was repeatable. In order to transport images, the team built a web database with a simple and secure upload screen. Once uploaded to the web server, all images could be viewed on either iPads or iPhones. Pathologists can view WSIs with the patient history in the app, request a more detailed view if needed, and then report results in the pEHR app.

As described, the OMT platform offers a number of clear advantages compared to traditional WSI machines. The mWSI system has a vastly simplified workflow, extremely low cost, is easily upgradable, and, in addition, accommodates static and video conferencing of telepathology. When paired with the pEHR application, it provides a powerful method for remote consultation and reporting of pathological diagnosis.


The authors of this paper would like to thank the following individuals for their help with this research: Ashley Moore from UCL for introducing us to the BioBank at UCLH; Dr Manuel Rodriquez-Justo for reviewing our research and giving us access to the BioBank; Dr Chiara Sugrue for donating the microscope we used for this research; Dr Tawfiqul Bhuiya and Dr Michael Esposito for reviewing the slides and providing support in New York; Dr Kristen Jacobs and Julie O'Keefe for developing the monolayer preparation; Dr Sharon Deans for travelling to Haiti every year and performing pap tests; Dr Louis-Joseph Auguste for reviewing the drafts of this paper and starting the TELL program.


View Abstract


  • Twitter Follow Louis Auguste Jr at @teamlivelonger and Dhaval Palsana at @dhavalpalsana

  • Contributors LA is the lead researcher and corresponding author of this project. He initiated this research in 2013. DP joined the research in 2014. He developed the software used for vignette removal. They collaborated on the authorship of this paper and both can attest to validity of the content of this paper.

  • Funding In 2013, we received funding and support from the British Medical Journal (BMJ), and University College London (UCL), by winning the IC Tomorrow Digital Innovation Contest in the Clinical Content Tools category.

  • Disclaimer Funding was provided by IC Tomorrow in association with the BMJ and UCL. However, the funding partners had no say in any of the research conducted in this project. They provided initial funds that were used to hire the developers who coded the iOS mobile application. LA Jr has filed for a patent based on this research.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; internally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.