PHARMA_INSIGHTS_Abkqh3geuyq3rihk3jrh218y493813re3hr,i23yr

Pharma Insights

1. PHARMA INSIGHTS AI-POWERED DRUG DETECTION SYSTEM

2. TITLE PHARMA INSIGHTS THROUGH AI-POWERED DRUG DETECTION SYSTEM PROJECT GUIDE: SRI Y.KUMAR SHEKHAR MTech(PhD) ASST.PROFESSOR DEPARTMENT OF CSM PROJECT ASSOCIATES: K.MONIKA S.DHANALAXMI SAI SRINIVAS PANDA I.JOSHI KUMAR

3. ABSTRACT Medication errors, especially incorrect dosages, pose significant health risks, leading to adverse drug reactions and increased medical costs. This project proposes a machine learning-based system that analyzes medicine compositions to assess risk levels and potential side effects. Users provide images of medicine labels, from which OCR/NLP technology extracts ingredient names and their compositions. The extracted data is compared with a medical reference database to determine if any ingredient exceeds recommended limits. Machine learning models classify risk levels as Safe, Moderate, or High Risk and predict possible side effects based on historical pharmacological data. A web-based interface allows users to interact with the system, upload images, and receive real-time analysis results. By integrating ML-based classification and predictive analysis, this tool enhances medication safety, reduces human errors, and helps users make informed drug consumption decisions. This project aims to provide an accessible and automated risk assessment system for safer medication use.

4. INTRODUCTION With the growing availability of pharmaceutical drugs, ensuring medication safety is crucial, as improper dosages or harmful ingredient combinations can pose serious health risks. This project introduces a machine learning-based drug detection system that analyzes medicine labels to assess risk levels and predict possible side effects. Using Optical Character Recognition (OCR), the system extracts text from labels, which is then refined using Natural Language Processing (NLP) to validate drug names. Machine learning models (CNN & RNN) classify drugs into Safe, Moderate and Risk categories and predict adverse effects. A web-based interface enables real-time analysis, helping users make informed decisions about their medication and reducing health risks.

5. Title Authors Year Key Contributions Algorithms/Methods Advantages Limitations Online Analysis of Ingredient Safety, Leveraging OCR and Machine Learning for Enhanced Consumer Product Safety S. S. S. R. Depuru 2022 Developed a smartphone-based tool that analyzes product ingredients for safety using OCR and machine learning. OCR, Machine Learning Empowers consumers with real-time safety information; Enhances transparency. Dependent on the accuracy of OCR; Requires comprehensive and up-to-date ingredient databases. Product Ingredient Toxicity Analyzer and Recommender Mrs. Likitha R, Arun S, B K Pramila, Chandan G 2021 Developed a system to assess ingredient toxicity and suggest safer alternatives using NLP and machine learning. NLP, Machine Learning Enhances consumer safety by identifying harmful ingredients; Provides safer alternatives. Dependent on the quality and comprehensiveness of the ingredient database. LITERATURE SURVEY

6. Title Authors Year Key Contributions Algorithms/Methods Advantages Limitations Using Machine Learning to Identify Toxic Ingredients in Products Eric et al. 2021 Created a mobile app that reads product ingredient lists to identify banned or restricted substances. Machine Learning, OCR Empowers consumers with real-time information on product safety; Enhances transparency. Accuracy depends on OCR performance; Requires up-to-date regulatory databases AI-Based Medical Diagnosis with Medicine Recommendation Janvi Patel 2024 Developed a system for disease prediction and corresponding medicine recommendations using AI. KNN, SVC, GridSearchCV, Sentiment Analysis High accuracy in predictions; Incorporates patient feedback for improved recommendations. Limited to a predefined set of diseases; Relies on the quality of input data.

7. EXISTING SYSTEM Manual Drug Label Inspection – Users manually read medicine labels, which can lead to human errors and misinterpretation. Online Drug Databases (WebMD, Drugs.com) – Users enter drug names to check side effects, but these databases require accurate input and may not include newly released medicines. Mobile Apps (Medscape, Epocrates) – Some apps provide drug interaction checks but lack real-time ingredient analysis and don’t detect excessive composition levels. AI-Based Systems (IBM Watson Health) – Used in hospitals for drug interactions but are not accessible for public use and don’t analyse ingredient risks.

8. PROPOSED SYSTEM Image-Based Drug Composition Extraction - Uses OCR to extract drug names and compositions from medicine labels. NLP corrects errors and validates drug names against a pharmaceutical database. Risk Level Classification Using Machine Learning - CNN processes the extracted image data to identify drug information. RNN/XGBoost predicts potential risks by analysing ingredient patterns and dosages. Categorizes drugs into Safe, Moderate Risk, or High Risk. Side Effect Prediction - Matches detected ingredients with a medical database to provide possible side effects. Web-Based Interactive Platform - Users can upload medicine images and receive instant risk assessments.

9. SOFTWARE REQUIREMENTS Programming Languages: Python (ML), HTML, CSS, JavaScript (Web). Frameworks & Libraries: ML & OCR : TensorFlow, OpenCV, Tesseract, OCR.NLP: NLTK / SpaCy . Web & Backend: Flask/Django, React.js. Database: MySQL / PostgreSQL (Stores drug data). Development & Deployment: VS Code, Postman, Docker, AWS/GCP. HARDWARE REQUIREMENTS Processor: Intel i5/i7 or AMD Ryzen 5/7 (for ML model training). RAM: Minimum 8GB (Recommended: 16GB for faster processing). Storage: SSD (256GB or more) for efficient data handling. GPU (Optional but Recommended): NVIDIA RTX 2060 or higher for ML acceleration.

10. Methodology for AI/ML-Based Drug Detection System Image Processing: Users upload a medicine label image via a web interface. OCR (Tesseract, OpenCV) extracts drug names and compositions. Drug Composition Analysis: Extracted data is validated using NLP and compared with a medical database to check for excessive ingredient levels. Risk Classification (CNN & RNN): CNN enhances text clarity, while RNN predicts risk levels (Safe, Moderate, High) based on ingredient concentration. Side Effects Prediction: The system matches compositions with medical records to provide possible side effects. Web-Based Output: Displays risk levels and warnings in an interactive web interface for users.

11. EXPECTED OUTPUT

12. ADVANTAGES Automated Drug Analysis – No manual input required; extracts data from images. Real-Time Risk Assessment – Detects harmful drugs instantly. Accurate Composition Validation – Uses ML & NLP to verify drug ingredients. User-Friendly Web Interface – Easy access for non-experts. Reduces Human Error – Eliminates misinterpretation of drug labels. DISADVANTAGES OCR Errors – Poor image quality may affect text extraction accuracy. Database Dependency – Requires an updated drug composition database. Computational Requirements – ML models need high processing power. Limited to Text-Based Analysis – Cannot analyze unknown or missing drug data. Internet Dependency – Web-based system needs an active connection.

13. BIBLIOGRAPHY 1.Depuru, S. S. S. R. (2022). Online Analysis of Ingredient Safety, Leveraging OCR and Machine Learning for Enhanced Consumer Product Safety. IEEE Xplore. This research introduces a smartphone-based system that utilizes OCR and machine learning to extract and analyze ingredient information from product labels, enabling real-time consumer safety insights. 2.Likitha, R., Arun, S., Pramila, B. K., & Chandan, G. (2021). Product Ingredient Toxicity Analyzer and Recommender. IEEE Xplore. This paper presents an NLP-based system that evaluates ingredient toxicity and suggests safer alternatives using machine learning, improving consumer awareness and safety. 3.Eric et al. (2021). Using Machine Learning to Identify Toxic Ingredients in Products. IEEE Xplore. A mobile application that employs machine learning and OCR to scan product labels and identify harmful or restricted substances, providing real-time safety analysis. 4.Patel, J. (2024). AI-Based Medical Diagnosis with Medicine Recommendation. Proceedings of the International Conference on Artificial Intelligence and Healthcare (ICAIH). This study develops an AI-powered medical diagnosis system that predicts diseases and recommends corresponding medicines using KNN, SVC, and sentiment analysis, enhancing healthcare decision-making.

14. QUESTIONS AND ANSWERS

15. T H A N K Y O U