| Summary: | Tablet misidentification possesses hazardous threats to medication safety. Incorrect medications could potentially result in adverse drug effect (AVD), which could even lead to death in long-term consumption. Research has shown that nine out of ten US citizens over the age of 65, who take more than one prescription tablets, are likely to misidentify their tablets [1]. As there is an increasing type of medicines and pharmaceutical brands being marketed in addition to the thousands already available, many of these medications may have a very similar appearance to each other. The look-alike tablet presents a challenging task for people to identify them correctly, especially for the elderly. Under the worst-case scenario, clinical testing is required to identify the components of the unknown tablet to know its actual identity. This project aims to develop an image-based tablet identification algorithm using computer vision approaches. Four attributes being utilized are shape, colour, size and imprint. The algorithm developed will identify and show the top five “visually similar” tablets together with the medical name of these tablets and the corresponding similarity scores. A collection of approximately 300 tablet images has been provided by Tan Tock Seng hospital to support the development of this project. Matric Laboratory (MATLAB) is the software platform used to develop the algorithm.
|
|---|