Abstract
This research introduces a novel approach towards deciphering and understanding medical prescriptions by employing Handwritten Character Recognition (HCR) and Recurrent Neural Networks (RNN). Our method offers an innovative solution to the longstanding problem of interpreting physicians’ handwritten notes, shorthand, symbols, and abbreviations in prescriptions, thereby minimizing medication errors due to misinterpretation. The study presents a two-step process that initially uses image processing techniques for enhancing the quality of prescription images, and subsequently applies an RNN-based model to recognize and interpret both handwritten and printed text. A potential application strategy involves the development and release of an open-source platform with an initial version of the model, which is further fine-tuned using real-world user data. The upgraded licensed version could include value-added features such as medicine availability checks, personalized dosage recommendations, and incentives for medicine purchase. The proposed system targets a wide range of users, primarily patients and pharmacists, with the possibility of incorporating personalized medicine suggestions based on customer medical histories and local medicine store inventory information.
Recommended Citation
Maiti, A.; Podder, A.; Dutta, C.; and Saha, D.
(2024)
"Improved RNN-based System for Deciphering Doctors' Handwritten Prescriptions,"
American Journal of Advanced Computing (AJAC): Vol. 2:
Iss.
4, Article 13.
Available at:
https://research.smartsociety.org/ajac/vol2/iss4/13