You are on page 1of 7

NON-INVASIVE BLOOD

SUGAR MONITORING FOR


NEONATES
BY TEAM MEDICATRONS:
Ananya Lakshmi Ravi
Dhurika K
Srihitha Pulapa
Nuaim Mohammed Chemnad Lasiyath
PROBLEM STATEMENT-1
(PS number: PS-XYZ123)
▪ EXPLANATION OF THE PROBLEM:
Babies born to mothers with impaired glucose metabolism are at a higher risk of experiencing neonatal hypoglycemia.
This condition can lead to serious consequences such as seizures, brain damage, and developmental problems. The
current protocols for monitoring a neonate's blood sugar levels typically require routine blood draws, which can be
intrusive, uncomfortable, and may pose additional risks, including the potential for infection.

▪ SIGNIFICANCE:
The vulnerability of newborns underscores the importance of a secure, precise, and less stressful approach to blood
sugar monitoring. Minimizing the stress associated with routine invasive blood testing is crucial for the overall growth
and well-being of the neonate, as it can be distressing for both parents and the infant.

▪ NEED FOR A SOLUTION:


A non-invasive, trustworthy technique for keeping an eye on newborns' blood sugar levels is much needed. A solution
like this would make less intrusive blood draws necessary, lower the risk of infections, and give newborns and their
caregivers a more relaxed and comfortable environment.
SOLUTION OVERVIEW
CONCEPT OVERVIEW:
Utilizing state-of-the-art iris imaging technology paired with advanced machine learning algorithms, our innovative non-invasive device is
designed to monitor newborns' blood sugar levels. This cutting-edge gadget provides a swift, painless, and highly effective method for
continuous glucose level monitoring, aiming to render invasive blood tests unnecessary.

COMPONENTS OF THE SOLUTION :


▪ IRIS IMAGE CAPTURE: We suggest using a customized, non-intrusive camera system that can take fine-grained pictures of the neonate’s
iris. In order to protect the comfort of the newborn, this camera will be built to work in low light. It will also include characteristics that
allow it to automatically focus and stabilize in order to take high-quality pictures.

▪ MACHINE LEARNING IMPLEMENTATION: Our main component is a Python-based machine learning model. A dataset of iris pictures
connected with known blood sugar levels will be used to train this algorithm. Subtle iris changes, including patterns, hues, and other
characteristics that have been tentatively linked to glucose levels, will be examined by the algorithm.

▪ RASPBERRY PI INTEGRATION: During the first stage, the prototype will be controlled by a Raspberry Pi. It will perform the machine
learning algorithm, interpret the camera's collected images, and deliver real-time blood sugar level measurements. The Raspberry Pi is a
great option for the prototype stage due to its small size and processing power.
DESIGN FLOW

Capturing Passing the CNN model


CNN model
the iris iris image to giving
evaluating
START image using a CNN output to END
the class of
camera model using through
the image
module raspberry pie raspberry pie
METHODOLOGY
▪ DATA COLLECTION:
The process begins with the collection of data, which consists of non-invasive measurements related to blood sugar levels. This
could include data from sensors capturing physiological signals that correlate with blood sugar concentration, such as infrared
signals or other non-invasive biomarkers.

▪ DATA PROCESSING AND FEATURE EXTRACTION:


The collected data is processed to extract relevant features that could provide information about the blood sugar levels. This
involves signal processing techniques and feature extraction algorithms to identify patterns or trends in the acquired data.

▪ CNN MODEL TRAINING:


A Convolutional Neural Network (CNN) model is designed and trained using TensorFlow. The model is trained on a labeled dataset,
which includes pairs of input features (extracted from the non-invasive measurements) and corresponding blood sugar level labels.

▪ MODEL ARCHITECTURE:
.
The CNN architecture is chosen to effectively capture hierarchical patterns in the input data. Convolutional layers are utilized for
spatial feature learning, and fully connected layers are employed for decision-making based on the learned features.
▪ RASPBERRY PI INTEGRATION:

The trained CNN model is converted to a format suitable for deployment on the Raspberry Pi. This could involve using
tools like TensorFlow Lite to optimize the model for edge computing on the Raspberry Pi's hardware.

▪ SOFTWARE DEVELOPMENT FOR CAMERA SENSOR INTEGRATION:


. Software is developed to interface with the camera sensor connected to the Raspberry Pi. The camera captures images or
signals related to the physiological measurements, which are then fed into the CNN model for real-time processing.

▪ REAL-TIME INFERENCE ON RASPBERRY PI:

The integrated system on the Raspberry Pi performs real-time inference using the trained CNN model. The non-invasive
measurements from the camera sensor are input into the model, and the model outputs predictions for blood sugar levels.

▪ OUTPUT AND DISPLAY :

The prediction if a person is diabetic or non diabetic can be displayed on a screen connected to the Raspberry Pi.
Additionally, the system may provide alerts or notifications based on predefined thresholds or patterns detected in the
predictions.
▪ TESTING AND OPTIMIZATION:

The entire system is rigorously tested to ensure correct functionality and smooth integration. Optimization steps may be
undertaken to enhance the performance of the model on the Raspberry Pi, considering the resource constraints of the
hardware.

. ▪ USER INTERACTION :

User interfaces or displays may be developed to provide feedback to healthcare providers or individuals about the monitored
blood sugar levels. This could include visualizations, trends, or historical data.

▪ CONTINUOUS MONITORING :

The device is designed for continuous monitoring, allowing it to provide real-time information about blood sugar levels
without the need for invasive procedures.

▪ CONCLUSION:

The proposed non-invasive blood sugar monitoring device leverages a combination of data processing techniques, CNN
model training, and real-time inference on a Raspberry Pi to provide an innovative solution for monitoring blood sugar
levels without the need for traditional invasive methods. The integration of hardware and software components aims to
create a user-friendly and efficient system for improved healthcare monitoring.

You might also like