NLM Announces Pill Image Recognition ChallengeJanuary 20, 2016
The National Library of Medicine (NLM) announced its Pill Image Recognition Challenge January 19, 2016 in the Federal Register at https://www.federalregister.gov/articles/2016/01/19/2016-00777/announcement-of-requirements-and-registration-for-pill-image-recognition-challenge. The Pill Image Recognition Challenge will also be posted on Challenge.gov. The submission period for the Challenge is April 4, 2016 to May 31, 2016, with winners announced August 1, 2016. More information about the Challenge itself can be found on the Web site at //pir.nlm.nih.gov/challenge.
The Pill Image Recognition Challenge is a National Institutes of Health (NIH) Challenge under the America COMPETES (Creating Opportunities to Meaningfully Promote Excellence in Technology, Education, and Science) Reauthorization Act of 2010 (Public Law 111-358). Through this Challenge the National Library of Medicine (NLM) seeks algorithms and software to match images of prescription oral solid-dose pharmaceutical medications (pills, including capsules and tablets). The objective of the Challenge is the development and discovery of high-quality algorithms and software that rank how well consumer images of prescription pills match reference images of pills in the authoritative NLM RxIMAGE database. NLM will use the Challenge entries (i.e., algorithm and software) to create a future API (Application Programming Interface) and a future software system for pill image recognition; the API will be freely accessible and the system will be freely usable.
Based on responses to a Request for Information in February 2015, NLM has developed evaluation software and created directories of images to use in selecting the Challenge winners. The software calculates the mean average precision (MAP) of submissions. The eligible submission with the highest score of the MAP will be recommended as the first place winner, with the second, third, fourth, and fifth best MAPs, respectively being recommended to earn second place, third place, and two honorable mentions. In the event of a tie score, the tied submission will be tested against additional datasets (DRJ and DCJ directories) until a winner is determined.
Unidentified and misidentified prescription pills present challenges for patients and professionals. Unidentified pills can be found by family members, health professionals, educators, and law enforcement. The nine out of 10 US citizens over age 65 who take more than one prescription pill can be prone to misidentifying those pills. Taking such pills can result in adverse drug events that affect health or cause death. To reduce such errors, any person should easily be able to confirm that a prescription pill or a refill is correct. For example, a person should be able to easily verify – or not – that a refill that has a different color, shape, or text imprinted on the pill is a different generic version of equivalent drugs he or she was already taking.
To help address these problems, the NLM Computational Photography Project for Pill Identification (C3PI) is developing infrastructure and tools for identifying prescription pills. The infrastructure includes photographs of such pills taken under laboratory lighting conditions, from a camera directly above the front and the back faces of the pill, and at high resolution. Specialized digital macro-photography techniques were then used to capture JPEG pill images. The NLM RxIMAGE database contains these high-quality images and associated pill data such as appearance (color, shape, size, text imprinted on the pill, etc.), ingredients, and identifiers such as its National Drug Code (NDC). RxIMAGE images and data are freely available. The freely accessible RxIMAGE API provides text-based search and retrieval of images and data from the RxIMAGE database. By contributing their algorithm and software, Challenge participants will take part in a broader NLM effort to develop a freely usable software system and a freely accessible API for image-based search and retrieval from a mobile device.