Medical Formulation Recognition (MFR) using Deep Feature Learning and One Class SVM

• Published in: Proceedings of ... International Joint Conference on Neural Networks pp. 1 - 7
• Main Authors: Kawi, Omar, Clawson, Kathy, Dunn, Paul, Knight, Daniel, Hodgson, Jonathan, Peng, Yonghong
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2020
• Subjects: Deep Learning; Drugs; Feature extraction; NLP; One-Class Learning; Principal component analysis; Stability analysis; Support Vector Machine; Support vector machines; Text Recognition; Training
• ISSN: 2161-4407
• DOI: 10.1109/IJCNN48605.2020.9206955

Specials medications are personalized formulations manufactured on demand for patients with unique prescription requirements and constitute an essential component of patient treatment. Specials are becoming increasingly in demand due to the need for personalized and precision medicine. The timely provision of optimal personalized medicine, however, is challenging, subject to strict regulatory processes, and is expert intensive. In this paper, we propose a new medical formulation engine (MFE) that performs semantic search across multiple disparate formulations archives to enable data driven formulation intelligence. We develop a new platform for medical formulations recognition (MFR) that curates a new dataset comprising formulations and non-formulations (clinical) text and uses a novel pipeline encompassing deep feature extraction and one-class support vector machine learning. The proposed MFR framework demonstrates promising performance and can be used as a benchmark for future research in formulations recognition.