Posted Date : 07th Mar, 2025
Peer-Reviewed Journals List: A Guide to Quality Research Publications ...
Posted Date : 07th Mar, 2025
Choosing the right journal is crucial for successful publication. Cons...
Posted Date : 27th Feb, 2025
Why Peer-Reviewed Journals Matter Quality Control: The peer revie...
Posted Date : 27th Feb, 2025
The Peer Review Process The peer review process typically follows sev...
Posted Date : 27th Feb, 2025
What Are Peer-Reviewed Journals? A peer-reviewed journal is a publica...
Patient Care Portal for Health Care Industry using Machine Learning
Author Name : M. Mohamed Rafi, M. Sabari Ramachandran, M. A. Abubucker Siddik
ABSTRACT Medication safety is a critical issue in daily patient care, with medication errors being one of the most preventable forms of medical error. Among these, 'look-alike and sound-alike' (LASA) medications pose significant challenges. Accurate identification of prescription pills based on their visual appearance is essential for patient safety, particularly for the elderly, visually impaired, and those with other accessibility needs. While several research efforts have addressed pill identification using content-based image retrieval (CBIR) and image classification techniques, challenges persist, notably due to the few-shot learning problem. This project aims to advance the state of pill recognition by developing an innovative Web Dashboard for automatic pill identification and reminder. We leverage a Masked Region Convolutional Neural Network (MR-CNN) to improve upon existing methods. Our approach involves two key stages: First, we use a blob-detection Region Proposal Network (RPN) for image segmentation, isolating the pill from its background and cropping a focused image around it. Second, we apply an MR-CNN classifier to analyze the cropped image and generate a ranked list of potential drug codes based on the pill's appearance. This dual-stage deep-learning model has demonstrated superior performance compared to traditional computer vision methods, achieving an accuracy of over 90%. The system's effectiveness in real-world scenarios highlights its potential to aid users in accurately identifying pills and reducing medication errors associated with LASA drugs.