Non - Invasive Detection of Vitamin Deficiency Using Resnet and Image Processing
P. Srija Vagdevi1,T.Sai Sampath2, G.Sai Madhuri3,V.Sachin4
1,2,3,4Students, Department of Computer Science and Engineering, Lendi Institute of Engineering and Technology (Autonomous), Andhra Pradesh, India
A.V.D. N Murthy, M.Tech, Assistant Professor, Department of Computer Science and Engineering, Lendi Institute of Engineering and Technology (Autonomous), Andhra Pradesh, India
paidasrija@gmail.com1, saisampath672@gmail.com2, govardhanagirisaimadhuri717@gmail.com3, sachinvudiga14082003@gmail.com4
ABSTRACT - Vitamin Deficiencies are not identified and addressed in a timely manner, can cause major health problems. Blood tests, which can be intrusive, time-consuming, and resource-intensive, are frequently used in traditional techniques of diagnosing vitamin deficiencies. In our work, we suggest a non-invasive method for detecting possible vitamin deficiencies from facial photos or other biomarkers such changes in the texture of the skin and nails by utilizing deep learning and image processing techniques. Utilizing pre-trained models and convolutional neural networks (CNNs) such as RESNET50, we extract important features from these images and link them with recognized signs of inadequacies. The system is trained using a collection of tagged photos of healthy controls and people with different vitamin deficits. Using thorough picture analysis, the model detects visual indicators of inadequacies such as skin discoloration, pallor, anomalies of the eyes, and changes in hair. We are predicting deficiencies for vitamins A, B2, B3, B5, B6, B7, B12, C, D, E, and K. However, we are not detecting deficiencies for vitamins B1 and B9 because the textures we have used to extract features are not suitable for identifying these particular vitamins. Our initial findings indicate that we can anticipate deficiency levels with a promising degree of accuracy. This could provide a rapid, easily accessible, and non-invasive early screening tool that could support preventative healthcare initiatives and standard diagnostic techniques.
Key Words: : Convolutional Neural Networks (CNNs), ResNet50, Deep Learning, Image Processing, Feature Extraction
