Non-Invasive Anemia Detection Using Conjunctiva and Nailbed Image Analysis with a Fusion Convolutional Neural Network

Bhuvan Gowda D L; Supreetha Gowda H D

doi:10.70528/IJLRP.v7.i7.2292

Non-Invasive Anemia Detection Using Conjunctiva and Nailbed Image Analysis with a Fusion Convolutional Neural Network

Author(s)	Mr. Bhuvan Gowda D L, Dr. Supreetha Gowda H D
Country	India
Abstract	Anemia, a condition characterized by abnormally low hemoglobin levels, remains one of the most prevalent health disorders worldwide, disproportionately affecting women, children, and elderly individuals. Conventional diagnosis relies on invasive blood tests that, while accurate, require laboratory infrastructure, trained personnel, and processing time that are not always available in rural or resource-limited settings. This paper presents a non-invasive anemia screening system that analyzes digital images of the palpebral conjunctiva and nailbed—anatomical regions whose coloration is known to vary with hemoglobin concentration—using two independently trained Convolutional Neural Network (CNN) models whose outputs are combined through a fusion strategy to produce a single Anemic-or-Normal classification. Each uploaded image is preprocessed through BGR-to-RGB color conversion, resizing to 224×224 pixels, and pixel-value normalization before being passed to its respective CNN model; the two resulting prediction scores are averaged, and the final classification is determined by thresholding the averaged score at 0.5. The system was developed and evaluated on a balanced dataset of 1,200 images (600 Anemic and 600 Normal) split 80:20 into training and testing subsets, and was deployed behind a Flask-based web interface that stores patient details and prediction records in both a MySQL database and an Excel file for future reference. Experimental results show per-class test performance of 82.5% accuracy, 81.8% precision, 83.2% recall, and an 82.5% F1-score for the Anemic class, and 83.7% accuracy, 84.1% precision, 82.9% recall, and an 83.5% F1-score for the Normal class, corresponding to an overall classification accuracy of approximately 83%. These results indicate that fusing conjunctiva- and nailbed-derived CNN predictions provides a practical, low-cost, and reasonably accurate preliminary screening tool that can support—rather than replace—laboratory-based anemia diagnosis, particularly in settings where blood testing is inconvenient or inaccessible.
Keywords	Anemia Detection; Non-Invasive Diagnosis; Conjunctiva Image Analysis; Nailbed Image Analysis; Convolutional Neural Network; Deep Learning; Medical Image Classification.
Field	Computer > Artificial Intelligence / Simulation / Virtual Reality
Published In	Volume 7, Issue 7, July 2026
Published On	2026-07-07
DOI	https://doi.org/10.70528/IJLRP.v7.i7.2292

View / Download PDF File

About IJLRP Fees & Payment Current Issue Publication Archive	Submit Research Paper Track Submission Status Publication Guidelines Publication Ethics	Join as a Reviewer Editors & Reviewers Get Reviewer Membership Certi.	Website/Journal Policies Usage Policy Content Policies Privacy Policy

Contact Us		+91-9687-828-838	editor@ijlrp.com

International Journal of Leading Research Publication

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Monthly Scholarly International Journal

Non-Invasive Anemia Detection Using Conjunctiva and Nailbed Image Analysis with a Fusion Convolutional Neural Network

Share this