Automated Analysis of Paper-Based Immunoassay Tests on Mobile Devices Krittika D’Silva, Nicola Dell, Gaetano Borriello and Paul Yager Department of Computer Science and Engineering and Department of Bioengineering University of Washington {kdsilva, nixdell, gaetano}@cs.washington.edu and yagerp@uw.edu ABSTRACT Disease detection in rural areas of low-resource settings is often hampered by a lack of accurate, convenient and af- fordable diagnostic tests. This paper presents research fo- cused on the design and analysis of software used to diag- nose immunoassay tests on a mobile device. The tests are paper-based and therefore cheap and appropriate for use in developing regions. Further, the software is simple and in- tuitive to use and would not require a health worker to have prior training. The developed Android application allows a user to first take a photo of an immunoassay test. The image of the test results is then processed using computer vision algorithms run on the mobile device. The mobile applica- tion then produces an objective and automated diagnosis of the immunoassay test at the point of care. General Terms Measurement Keywords Point-of-care diagnostics, computer vision, mobile phone 1. INTRODUCTION Infectious diseases account for more than 50% of deaths in developing countries worldwide [1] . Detection of these dis- eases in rural areas, however, is often hampered by a lack of diagnostics tests that are accurate, affordable and con- venient [2]. Tests routinely administered in well-equipped clinical laboratories are often expensive, complicated and require trained professionals; consequently, these tests are not appropriate for low-resource settings [3, 4]. The Yager laboratory at the University of Washington has developed a device to be used at the point of care to diag- nose infectious diseases. This device is shown in Figure 1. Termed MAD NAAT, short for multiplexed autonomous dis- posable nucleic acid amplification test, the device uses a combination of cell lysis, isothermal amplification and lat- eral flow to determine the presence of a particular pathogen. These tests will be paper-based (therefore inexpensive to produce), portable, and reliable. Further, in developing countries where clinics often have low patient return rates, this point-of-care device will provide an immediate diagno- sis to health workers. The result of a test is indicated by the presence, absence and intensity of a series of colored lines. To produce a diagnosis, these lines must be inter- preted. This critical task, however, requires an evaluation of the intensity profile, location, and color of numerous lines. Left to the human eye, a subjective or inaccurate diagnosis could be developed. Therefore, a health worker should not visually interpret the test results. Mobile phones can, in this situation, be appropriate tools to automate the interpreta- tion of these test results. Phones are portable, durable, and generally have built-in cameras; as the price of smart phones continues to decrease, they are increasingly becoming a vi- able tool in low-resource environments [5]. This paper presents a mobile application as an accurate and reliable means of disease diagnosis. It describes the devel- opment process and implementation of the software as well as its evaluation. 2. PREVIOUS WORK This research presented in this paper combines image pro- cessing on mobile devices with the analysis of diagnostics tests. In the commercial sector, software with a similar di- agnostic purpose has been developed by FIO Corporation. The Deki Reader, a solution they have created, interprets rapid diagnostic tests and manages the data for multiple pa- tients at a time [6]. The company, however, requires all users to purchase their Deki Phone, Deki Reader, Deki Tablet or Spiki (a device similar to a computer) [6]. This would mean that health workers would be restricted in the devices they could use. Additionally, the FIO Corporation requires use of their cloud-based services that have recurring fees; this model would not be ideal in developing regions. The soft- ware we are developing works on any device with an Android