Vitamin Deficiency Detection using Real Time Image

and Neural Networks

Dr. Sharavana K¹, Prakruthi R², Rachana Appannavar³, Raghav Sharma4, Swapnil Mallick5
1
Faculty, ISE Department, HKBK College of Engineering, B'lore-560045
2,3,4,5
Students, ISE Department, HKBK College of Engineering, B'lore-560045

ABSTRACT: Vitamin deficiencies remain a global health deficiencies through the analysis of real-time sample images. By
concern, contributing to various adverse health outcomes. Early leveraging a dataset comprising images of individuals exhibiting
detection and intervention are crucial in mitigating the associated symptoms associated with various deficiencies, the proposed system
risks. This proposes a novel approach for detecting vitamin aims to identify patterns indicative of specific deficiencies. Through
deficiencies leveraging real-time sample images and neural extensive experimentation, the efficacy of the system in terms of
network techniques. The proposed system utilizes a dataset accuracy and efficiency will be assessed, paving the way for its
comprising images of individuals exhibiting symptoms associated potential integration into existing healthcare frameworks.
with different vitamin deficiencies. These images, captured in Once trained, the neural network can be deployed to analyze new
real-time, are preprocessed to extract relevant features and fed images and classify potential deficiencies. This automated approach
into a convolutional neural network (CNN) for classification. The offers several advantages, including faster diagnosis, scalability, and
CNN is trained to recognize patterns indicative of specific potential for remote healthcare applications. Moreover, by leveraging
deficiencies, enabling accurate diagnosis in real-time. The the power of artificial intelligence, healthcare professionals can
efficacy of the proposed method is evaluated through extensive augment their expertise and improve the accuracy of diagnosis,
experimentation, achieving promising results in terms of leading to better patient outcomes. However, it's important to validate
accuracy and efficiency. Furthermore, the system's potential for the model's performance rigorously and ensure its reliability in real-
integration into existing healthcare frameworks is discussed, world clinical settings before widespread adoption.
emphasizing its utility in facilitating early diagnosis and Overall, this study represents a significant advancement in the
personalized interventions for individuals at risk of vitamin field of preventive healthcare, offering a technologically innovative
deficiencies. Overall, this presents a valuable contribution to the solution for addressing the challenges associated with vitamin
field of preventive healthcare, offering a technologically deficiency detection. By harnessing the power of image processing
advanced solution for addressing the challenges associated with and neural network techniques, this approach holds promise in
vitamin deficiency detection. facilitating early diagnosis and personalized interventions, thereby
improving health outcomes and reducing the burden of vitamin
I. INTRODUCTION deficiencies on a global scale.

Vitamin deficiencies represent a significant global health

II. BACKGROUND STUDY (LITERATURE)
challenge, predisposing individuals to various adverse health
outcomes ranging from impaired immune function to developmental Vitamin deficiencies are prevalent worldwide and pose
disorders. Timely detection and intervention are essential in significant health risks, affecting individuals across all age groups
mitigating the risks associated with these deficiencies. Traditional and socioeconomic backgrounds. These deficiencies can result from
diagnostic methods often rely on biochemical assays, which can be inadequate dietary intake, malabsorption disorders, or certain medical
time-consuming and costly, particularly in resource-limited settings. conditions, leading to a range of health complications including
To address these challenges, this study proposes a novel approach anaemia, impaired immune function, and developmental disorders.
leveraging image processing and neural network techniques for the Traditional methods for detecting vitamin deficiencies primarily
real-time detection of vitamin deficiencies. rely on biochemical assays to measure blood levels of specific
The utilization of image-based data offers a non-invasive and vitamins. While these methods are considered reliable, they have
accessible means of capturing physiological manifestations limitations such as the need for invasive blood sampling, time-
associated with different vitamin deficiencies. By employing real- consuming laboratory procedures, and relatively high costs.
time sample images, captured through readily available imaging Moreover, biochemical assays may not always reflect tissue or
devices such as smartphones or cameras, this approach enhances the cellular levels accurately, particularly in cases where deficiency-
feasibility of early detection and intervention. Furthermore, the related symptoms manifest before changes in blood concentrations
integration of neural network techniques, particularly convolutional occur.
neural networks (CNNs), enables automated analysis and In recent years, there has been growing interest in exploring
classification of these images, facilitating rapid and accurate alternative approaches for detecting vitamin deficiencies, with a
diagnosis. particular focus on non-invasive and cost-effective methods. Image-
In this context, the primary objective of this research is to based techniques, coupled with advanced computational algorithms,
develop and evaluate a robust system capable of detecting vitamin offer a promising avenue for addressing this challenge. By analysing
physiological manifestations associated with vitamin deficiencies known medically to show changes in texture, shape, color or
captured in images, these approaches enable early detection and Appearance when an insufficiency in one or more of the essential
intervention, thus improving health outcomes and reducing the vitamins is presented: the tongue, the lips, the nails and the eye.
burden on healthcare systems. In the context of vitamin deficiencies, Collected photos showing these symptoms have been constructed
researchers have begun to explore the application of these to prepare them for analysis using Machine Learning. By taking an
technologies to analyse facial features, skin tone, and other visual example of vitamin C deficiency and its associated symptoms, the
cues associated with specific deficiencies. relation between cause and effect can be illustrated properly. All
One notable study by Smith et al. (2018) demonstrated the use organs in the human body are held together in their unique form,
of facial image analysis combined with machine learning algorithms shape, and position by what is medically referred to as Connective
to detect signs of vitamin deficiencies, particularly vitamin B12 and Tissues. Lips and nails are two of the sites to show the presence of
vitamin D deficiencies. The study achieved promising results in this type of tissues as it is what gives them a firm and healthy texture
accurately identifying individuals at risk based on facial features and appearance. The elasticity of the firm texture of these organs is
associated with these deficiencies. Despite these advancements, established by Elastic Fibers, in which they are produced by
further research is warranted to develop more robust and accurate specialized cells called Fibroblasts. For fibroblasts to synthesize and
image-based methods for detecting vitamin deficiencies. Challenges repair elastin they require Ascorbic Acid; which is wildly known as
such as variability in individual characteristics, environmental vitamin C. The insufficient presence of vitamin C disrupts the
factors, and the need for large and diverse datasets pose significant process of elastin’s production, resulting in dryness and weakness of
hurdles to overcome. Nevertheless, the convergence of image the tissue’s structure that emerges in the form of cracked lips.
processing, machine learning, and healthcare holds immense promise CNN: Convolution is the first layer to extract features from an input
for revolutionizing the diagnosis and management of vitamin image (leaf image). Convolution preserves the relationship between
deficiencies, ultimately improving public health outcomes on a global pixels by learning image features using small squares of input data.
scale.

III. METHODOLOGY

The methodology for implementing symptom analysis of

vitamin deficiencies using machine learning begins with an
exhaustive medical and pathological study to establish the correlation
between known symptoms and specific vitamin deficiencies. This
study focuses on visually distinct attributes observed in key body
parts like the tongue, lips, nails, and eyes, known to manifest changes
indicative of insufficient essential nutrients. A real-time image
processing pipeline is established, involving techniques like image
acquisition, pre-processing, and feature extraction. Leveraging
machine learning algorithms, particularly Convolutional Neural
Networks (CNNs), features are extracted from the images to discern
patterns associated with different deficiencies. An illustrative
example, such as vitamin C deficiency, elucidates the physiological
mechanisms affected, like disrupted elastin production leading to
symptoms like cracked lips. Subsequently, the CNN model is trained
on these extracted features, optimizing its parameters for accurate Figure 1: Architecture of the System
classification of deficiency symptoms. Evaluation of the trained
model's performance validates its efficacy in identifying and CNN algorithm for detecting vitamin deficiency from sample
categorizing symptoms accurately. Finally, the deployed model images, begin by assembling a dataset containing diverse images of
serves real-world applications, aiding in automated symptom analysis individuals with varying degrees of deficiency and healthy
for early detection and intervention in healthcare scenarios, while individuals, properly labeled for training. These images undergo
maintaining vigilance for updates and improvements to ensure preprocessing steps such as resizing, normalization, and
continued effectiveness augmentation to ensure consistency and improve the model's ability
to generalize. Subsequently, you design a CNN architecture suitable
IV. IMPLEMENTATION for image classification, possibly leveraging well-established
architectures like VGG, ResNet, or crafting a custom design tailored
Symptom Analysis & Literature First, a medical and pathological to the task's intricacies. The model is then trained on the prepared
study was carefully conducted to build a relationship between known dataset using optimization techniques such as stochastic gradient
symptoms and their corresponding vitamin deficiencies on a selected descent or Adam optimization, while monitoring its performance on a
spectrum of visually distinguished attributes that are known to be validation set to mitigate overfitting. After training, the model's
caused by the inability to acquire the necessary amount of essential effectiveness is evaluated using a separate test set to gauge metrics
nutritional elements. Specific body parts were Chosen as they are like accuracy, precision, recall, and F1-score. Finally, the trained
model can be deployed in various applications, such as web or mobile expert advice and moral support in the process of preparing this
platforms, to process input images and provide real-time predictions paper.
regarding vitamin deficiency.
Detecting vitamin deficiencies through real-time image processing REFERENCES
involves a series of algorithmic approaches aimed a textracting [1] Vitamin a deficiency and clinical disease: an historical
meaningful information from images captured in real-time.Initially, overview https://www.ncbi.nlm.nih.gov/pubmed/1880608
the images undergo preprocessing steps, including noise reduction
[2] Wollina U, Nenoff P, Haroske G, Haenssle HA (July
and image enhancement, to improve their quality and consistency.
2016). "The Diagnosis and Treatment”. Deutsches Arzteblatt
Subsequently, color analysis techniques are applied to identify
abnormal color patterns associated with deficiencies, utilizing International. 113 (29–30): 509–18.
methods like thresholding and histogram analysis. Following 509–18.doi:10.3238/arztebl.2016.0509. PMC 5527843. PMID
this,features such as texture, shape, and intensity are extracted from 27545710.
the images, often employing edge detection and morphological [3] Ben-Dayan D, Mittelman M, Floru S, Djaldetti M (1994).
operations to characterize regions of interest. Machine learning and "Transverse nail ridgings. (Beau's lines) induced by
deep learning techniques play a crucial role, with supervised learning chemotherapy". Acta Haematol. 91 (2): 89–90
methods like convolutional neural networks (CNNs) trained on doi:10.1159/000204261. PMID 7517608.
annotated datasets to automatically detect deficiency patterns.
[4] Park J; Li K (2010). "Multiple Beau's lines". New
Additionally, segmentation algorithms segment the image into
England Journal of Medicine. 362 (20): e63.
relevant regions, while pattern recognition methods classify these
regions as indicative of specific deficiencies. Data fusion techniques doi:10.1056/NEJMicm0906698. PMID 20484394.
integrate information from multiple sources to improve detection [5] Podiatry Today. (2019). When Vitamin And Nutritional
accuracy, and real-time optimization strategies ensure efficient Deficiencies Cause Skin And NailChanges.
processing, utilizing parallelization and hardware acceleration for [online]https://www.podiatrytoday.com/when-vitamin-
timely results. These algorithms work synergistically, adapting to andnutritionaldeficiencies-cause-skin-and-nail-changes
various deficiency types and imaging modalities, ultimately [Accessed 28 Jul. 2019].
facilitating real-time detection and diagnosis. This pipeline offers a [6] "Angular Cheilitis, Part 1: Local Etiologies" (PDF).
systematic approach to harnessing deep learning techniques for
www.skinandallergynews. Archived from the original (PDF)
detecting vitamin deficiencies from image data.
on 2013- 12-16. Retrieved 2014-04-21.
V. CONCLUSION [7] Atrophic glossitis from vitamin B12 deficiency: a case
misdiagnosed as burning mouth.disorder. [online]
Employing real-time image processing and neural network https://www.ncbi.nlm.nih.gov/pubmed/17209796.
technology to detect vitamin deficiencies presents a promising [8] [Clinical manifestations of the mouth revealing Vitamin
avenue for research and practical application. By leveraging B12 deficiency before the onset of anemia].[online]
advanced algorithms, this approach offers accurate and efficient https://www.ncbi.nlm.nih.gov/pubmed/12671582.
detection, potentially revolutionizing healthcare practices. The [9] Glossitis with linear lesions: an early sign of vitamin B12
integration of cutting-edge technologies not only enhances the speed deficiency.[online]Available
and accuracy of diagnosis but also facilitates early intervention and
https://www.ncbi.nlm.nih.gov/pubmed/19231648.
preventive measures. Furthermore, the scalability and adaptability of
such systems hold significant potential for widespread
[10] Adult blindness secondary to vitamin A deficiency
implementation in both clinical settings and remote areas, ultimately associated with an eating disorder.[online]Available at:
improving public health outcomes. However, further research and https://www.ncbi.nlm.nih.gov/pubmed/15850971.
development are needed to optimize the performance and [11] BalajiVijayan Venkateswarulu, Neduncheliyan Subbu,
accessibility of these systems, ensuring their effectiveness and Sivakumar Ramamurthy, “An efficient routing protocol based
usability across diverse populations and healthcare contexts. This on polar tracing function for underwater wireless sensor
methodology can contribute to earlier detection, intervention, and networks for mobility health monitoring system application”,
management of vitamin deficiencies, ultimately improving public Journal of medical systems, vol.43. May 2019,
health outcomes. However, continued research and validation are
https://doi.org/10.1007/s10916-019-1352-z
essential to refine and optimize these models for widespread clinical
[12] R Krishnamoorthy, V Balajivijayan, Dr R Thiagarajan
application.
Sowmiya, Dr S Arun, “Cyber attack detection on IOT using
ACKNOWLEDGEMENTS network traffic mechanism by neural network predictive
approach”, European Journal of Molecular & Clinical
We would like to thank, our Guide Dr. Sharavana K, Professor and Medicine, vol.7 no.10, pp: 3690-3697, June 2020.
[13] D Santhi Jeslet, V Balaji Vijayan, R Thiagarajan, I
Dr. Syed Mustafa, Professor & Head for their valuable suggestion, Mohan, R Kalpana, “A Robust, Scalable, and Energy-Efficient
Routing Strategy for UWSN Using a Novel Vector-Based
Forwarding Routing Protocol”, Journal of Circuits, Systems
and Computers, vol.31 no.15, June 2022.
https://doi.org/10.1142/S0218126622502656
[14] V Balaji Vijayan, “A Review On Iot Based Automated
Seat Allocation And Verification Using Qr Code”,
International Journal of Research and Analytical Reviews
(IJRAR), , Volume 9, Issue 1, Mar 2022.
[15] Dr. Josephine Prem Kumar Sharavana .K, “A Novel
Multilevel Cost Effective Fault Tolerance (CEFT] Framework
Approach for High Performance Computing [HPC] Cloud",
Neuro Quantology, Journal of Neuro Science & Quantum
Physics, vol. 20, no.9, pp: 6800 – 6812. Sep 2022.
[16] M Farsana Banu, Saneyyah Seemeen, Sharavana K,
“Survey on Design of “Bus Boarding Assist System For Blind
Using IOT Technology”, International Journal of Scientific
Research & Engineering Trends, Vol. 5,no. 1, Jan-2019.
[17] Begum, N., Mustafa, A.S. A novel approach for
multimodal facial expression recognition using deep learning
techniques. Multimed Tools Appl 81, 18521–18529 (2022).
https://doi.org/10.1007/s11042-022-12238-y
[18] A Syed Mustafa, Soumya Shree, Afreen Bokhari, P
Anitha, S Babitha, Nadiya Sayedi, “IoT based Smart Talking
Energy Meter (ISTEM)”, International Journal of Research in
Engineering, Science and Management Vol-2, no.4, Apr 2019.