A Deep Neural Network Approach with Pioneering Local Dataset to Recognize Doctor's Handwritten Prescription in Bangladesh

• Published in: 2024 International Conference on Advances in Computing, Communication, Electrical, and Smart Systems (iCACCESS) pp. 1 - 6
• Main Authors: Mia, Abdur Rahim, Chowdhury, Mohammad Abdullah-Al-Sajid, Mamun, Abdullah Al, Ruddra, Aurunave Mollik, Tanny, Nawshin Tabassum
• Format: Conference Proceeding
• Language: English
• Published: IEEE 08.03.2024
• Subjects: Adaptation models; Bangladesh; Convolutional Neural Networks; Custom Dataset; Data models; Handwriting recognition; Handwritten Prescription Recognition; Healthcare; Hospitals; Machine learning; Medical Terminologies; Medication Safety; Software; Terminology
• DOI: 10.1109/iCACCESS61735.2024.10499631

This thesis introduces an innovative machine learning solution to a prevalent healthcare challenge in Bangladesh: the legibility of handwritten prescriptions. Where 97.1% of Bangladeshi doctors relying on handwritten prescriptions, this study addresses a critical gap in patient safety and prescription accuracy. The motivation for this research stems from the urgent need to improve legibility in prescriptions, a factor that directly impacts patient care and medication errors. At the core of our study is the development of a specialized machine learning system, tailored to the nuances of Bangladeshi prescriptions. Our research is the first in Bangladesh to apply machine learning, specifically the VGG16 model, to interpret and digitize handwritten prescriptions. This innovative study is underpinned by a unique dataset, the first of its kind, comprising diverse hand-writing styles and medical terminologies collected from various Bangladeshi hospitals. A pivotal achievement of this research is the effective adaptation and application of the VGG16 model. This model demonstrated remarkable accuracy in recognizing and interpreting medical handwriting, thereby ensuring precise medication instructions and dosages. The success of this system represents a major advancement in medication safety and is a critical step toward mitigating the risks associated with misin-terpreted prescriptions. By enhancing the clarity and accuracy of prescription interpretation, this thesis contributes significantly to the improvement of healthcare delivery in Bangladesh. It not only exemplifies the transformative power of machine learning in refining healthcare practices but also establishes a foundational model for creating safer, more efficient healthcare systems in developing nations.