Disintegration testing augmented by computer Vision technology
Abstract
Oral solid dosage forms, specifically immediate release tablets, are prevalent in the pharmaceutical industry. Disintegration testing is often the first step of commercialization and large-scale production of these dosage forms. Current disintegration testing in the pharmaceutical industry, according to United States Pharmacopeia (USP) chapter 〈701〉, only gives information about the duration of the tablet disintegration process. This information is subjective, variable, and prone to human error due to manual or physical data collection methods via the human eye or contact disks. To lessen the data integrity risk associated with this process, efforts have been made to automate the analysis of the disintegration process using digital lens and other imaging technologies. This would provide a non-invasive method to quantitatively determine disintegration time through computer algorithms. The main challenges associated with developing such a system involve visualization of tablet pieces through cloudy and turbid liquid. The Computer Vision for Disintegration (CVD) system has been developed to be used along with traditional pharmaceutical disintegration testing devices to monitor tablet pieces and distinguish them from the surrounding liquid. The software written for CVD utilizes data captured by cameras or other lenses then uses mobile SSD and CNN, with an OpenCV and FRCNN machine learning model, to analyze and interpret the data. This technology is capable of consistently identifying tablets with ≥ 99.6% accuracy. Not only is the data produced by CVD more reliable, but it opens the possibility of a deeper understanding of disintegration rates and mechanisms in addition to duration.
Additional Information
© 2022 Elsevier. Received 30 June 2021, Revised 8 March 2022, Accepted 11 March 2022, Available online 15 March 2022. The authors would like to acknowledge the internship program of Merck & Co., Inc., with headquarters in Kenilworth, NJ, USA through which some work was completed. Additionally, we thank Mr. Randolph Crawford, Dr. Antong Chen, RNDr. Jindrich Soukup, and Mr. Janakiraman Gopinath for their advice on the machine learning concepts and neural network architecture, Mr. William Blincoe for his support with the probe-based particle imaging technology, Mr. David Rossi for his experiments testing additional imaging technologies, and Ms. Joanna Everitt and Mr. Wilfredo Maldonado for assisting in gathering disintegration testing videos that were used to help train the model. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.Additional details
- Eprint ID
- 114001
- Resolver ID
- CaltechAUTHORS:20220322-742272000
- Created
-
2022-03-23Created from EPrint's datestamp field
- Updated
-
2022-04-18Created from EPrint's last_modified field