AI‑Driven Heart Attack and Blood Flow Restriction Prediction Using

A Smart Band Integrated with Wearable Sensors

Gobinath R.

, Shahana K.

, Rajeshwari R.

, Fahumitha Afrose

Fathima Fazlina M.

and Durga Devi S.

Department of Artificial Intelligence and Data Science, E.G.S. Pillay Engineering College,

Nagapattinam, Tamil Nadu, India

E. G. S. Pillay Engineering College, Nagapattinam, Tamil Nadu, India

Keywords: Smart Wearable Band, Heart Attack Prediction, Blood Flow Restriction, Machine Learning, Gradient

Boosting, LSTM.

Abstract: Cardiovascular diseases remain a leading cause of mortality worldwide, often due to the lack of timely

detection and intervention. Wearable health monitoring devices have emerged as crucial tools for continuous

health tracking and early risk detection. This study introduces Aura Wear, a smart wearable designed to

predict heart attack risk and detect blood flow restriction in real-time. The aim is to provide users with early

warnings, empowering them to take preventive measures and improve overall well-being. The device

integrates multi-sensor technology, including Near-Infrared Spectroscopy, Tissue Perfusion, Heart Rate,

Oxygen Saturation, Blood Pressure, Stress Level, and Physical Activity sensors. Gradient Boosting is

employed for current risk detection while the wearable is worn whereas LSTM networks predict future risk

when the device is not worn. The platform analyses sensor data collected over a 3-4-hour period to predict

heart attack risk for the next 20 hours. For blood flow restriction detection, a Gradient Boosting classifier

evaluates instantaneous variations in heart rate, oxygen saturation, and perfusion index, ensuring accurate

identification. The model demonstrated 80% accuracy for cardiac attack risk forecasting by testing medical

datasets. With our application, it successfully notified instantaneous alerts when health parameters deviated

from normal thresholds, enabling timely intervention. Aura Wear bridges the gap between conventional

wellness tracking and proactive support through offering continuous well-being insights. The device

empowers users to manage their health more effectively, potentially reducing the risk of life-threatening

conditions.

1 INTRODUCTION

In contemporary society, cardiovascular diseases

(CVDs) remain a significant global health concern,

with myocardial infarctions and circulatory disorders

being major causes of mortality. Early detection and

timely intervention are crucial in preventing these

life-threatening conditions. However, existing

diagnostic methods primarily rely on clinical

assessments, which may not be accessible at critical

moments. Wearable health monitoring technology has

emerged as a transformative innovation in real-time

health tracking, enabling individuals to monitor their

vital signs continuously. Wearable technology has

undergone significant advancements in recent years,

revolutionizing healthcare by providing continuous,

non-invasive monitoring of physiological parameters.

These devices utilize sensor fusion, artificial

intelligence-driven analytics, and wireless

connectivity to deliver real-time health insights.

Wearables are increasingly being employed not only

for fitness tracking but also for disease prevention,

early diagnostics, and remote patient management.

The capacity to collect large volumes of real-time

data facilitates early detection of critical health

conditions, reducing dependence on periodic clinical

evaluations. With ongoing advancements in sensor

technology and artificial intelligence integration,

wearable health devices are playing a pivotal role in

personalized medicine and preventive healthcare.

Aura Wear is a smart wearable band designed to

predict myocardial infarction risks and detect blood

flow restrictions using artificial intelligence-driven

analytics and sensor integration. This project

716

R., G., K., S., R., R., Afrose, F., M., F. F. and S., D. D.

AI-Driven Heart Attack and Blood Flow Restriction Prediction Using A Smart Band Integrated with Wearable Sensors.

DOI: 10.5220/0013888900004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 2, pages

716-722

ISBN: 978-989-758-777-1

leverages advanced machine learning techniques to

process real-time physiological data collected

through multiple sensors, including Near-Infrared

Spectroscopy (NIRS), Heart Rate, Blood Pressure,

Oxygen Saturation (SpO2), Stress Level, and

Physical Activity sensors. Unlike traditional

wearables that provide only basic health statistics,

Aura Wear utilizes intelligent data processing and

predictive modeling to assess cardiovascular health

comprehensively. The distinguishing feature of Aura

Wear is its dual-functionality. Real-time health

monitoring ensures that while the user wears the

device, the system continuously analyzes vital signs

to detect instantaneous abnormalities, ensuring

immediate intervention when critical thresholds are

exceeded. Future risk prediction using Long Short-

Term Memory (LSTM) networks allows the system

to predict myocardial infarction risks for up to 20

hours in advance based on data collected over 3 to 4

hours, providing proactive healthcare solutions. For

blood flow restriction detection, the system employs

Gradient Boosting classifiers to evaluate fluctuations

in key parameters such as heart rate, oxygen

saturation, and perfusion index. If abnormalities are

detected, the user receives instant alerts, allowing for

immediate medical attention and preventive action.

Continuous monitoring ensures uninterrupted

tracking of key cardiovascular parameters, while

predictive healthcare models anticipate risks before

symptoms manifest. Seamless integration with

mobile applications provides facile data access and

remote health monitoring. The device is designed

with a lightweight, ergonomic, and energy-efficient

build for extended usage. Enhanced preventive care

reduces reliance on hospital visits and enables self-

management of cardiovascular health. This project

aims to provide an innovative artificial intelligence-

powered health monitoring solution, reducing

dependence on hospital visits while enhancing

proactive healthcare. By integrating real-time

monitoring with predictive analytics, Aura Wear

bridges the gap between traditional health tracking

and intelligent, personalized healthcare solutions. The

future of wearable healthcare technology lies in

artificial intelligence-driven, real-time health

diagnostics, and Aura Wear is at the forefront of this

evolution.

2 RELATED WORK

The landscape of wearable health monitoring has

expanded rapidly in recent years, offering continuous

and non-invasive methods for tracking physiological

parameters. These devices often leverage sensor

fusion, AI-driven analytics, and wireless

communication to deliver real-time health insights,

moving beyond basic fitness tracking to encompass

disease prevention, early diagnostics, and remote

patient management. The increasing availability of

high-resolution, real-time data has enabled more

proactive approaches to healthcare, potentially

reducing reliance on periodic clinical evaluations.

Several wearable health monitoring systems have

emerged, focusing on parameters such as heart rate,

oxygen saturation, and physical activity. While these

systems offer basic cardiovascular monitoring

capabilities, their ability to provide early warnings for

acute cardiovascular events, such as myocardial

infarction, or detect localized issues like blood flow

restriction remains limited. Although some existing

wearables can detect atrial fibrillation, their

functionality does not extend to continuous risk

assessment for myocardial infarction or direct

detection of peripheral blood flow limitations.

Significant research has explored the utilization of

artificial intelligence to enhance the predictive

capabilities of health monitoring systems. Machine

learning techniques such as Gradient Boosting have

been employed for their ability to handle complex,

non-linear relationships between physiological

parameters and cardiovascular risk. These models

have demonstrated efficacy in analyzing real-time

health data to improve cardiovascular risk

assessment. However, many existing

implementations are constrained to clinical settings

and rely on retrospective data rather than real-time

monitoring. Current methods for detecting blood flow

restriction often rely on invasive techniques such as

angiography or Doppler ultrasound, which are costly,

require specialized equipment, and are not suitable

for continuous monitoring. Wearable sensors, such as

those based on Near-Infrared Spectroscopy (NIRS),

offer a non-invasive alternative for assessing

peripheral blood flow. However, few wearable

systems integrate NIRS with other physiological

sensors and advanced machine learning algorithms to

provide real-time detection of blood flow restriction.

Aura Wear addresses these limitations by offering a

smart wearable band designed for real-time

myocardial infarction risk prediction and blood flow

restriction detection through AI-driven analytics and

sensor integration. The system leverages real-time

data collected from multiple sensors, including Near-

Infrared Spectroscopy (NIRS), heart rate, blood

pressure, oxygen saturation (SpO2), stress level, and

physical activity sensors. Unlike traditional

wearables that primarily provide basic health

AI-Driven Heart Attack and Blood Flow Restriction Prediction Using A Smart Band Integrated with Wearable Sensors

717

statistics, Aura Wear utilizes intelligent data

processing and predictive modelling to assess

cardiovascular health comprehensively. A key

differentiator of Aura Wear is its dual-functionality.

Real-time health monitoring continuously analyzes

vital signs to detect instantaneous abnormalities,

ensuring immediate intervention when critical

thresholds are crossed. For blood flow restriction

detection, Aura Wear employs a Gradient Boosting

classifier to evaluate fluctuations in key parameters

such as heart rate, oxygen saturation, and perfusion

index. The model demonstrated 80% accuracy for

cardiac attack risk forecasting by testing medical

datasets. When abnormalities are detected, the user

receives instant alerts, allowing for immediate

medical attention and preventive action. The system's

capacity to identify minor fluctuations in blood

circulation, coupled with its ongoing monitoring and

forecasting functions, offers a notable improvement

over current technique. Aura Wear offers a novel

approach to cardiovascular health management by

integrating real-time monitoring with predictive

analytics. By bridging the gap between traditional

health tracking and intelligent, personalized

healthcare solutions, Aura Wear aims to empower

individuals to proactively manage their

cardiovascular health, reducing dependence on

hospital visits and enabling more timely

interventions.

3 METHODOLOGY

Aura Wear is a smart wearable band designed to

provide real-time heart attack risk prediction and

blood flow restriction detection. The system consists

of multiple integrated sensors that collect

physiological data, which is then processed using AI-

driven analytics. The wearable device is equipped

with Near-Infrared Spectroscopy (NIRS) for assessing

tissue perfusion and blood flow, a heart rate sensor for

monitoring beats per minute and heart rate variability,

an oxygen saturation (SpO2) sensor to measure

oxygen levels in the blood, a blood pressure sensor to

detect fluctuations in systolic and diastolic pressure, a

stress level sensor to evaluate physiological responses

to stress, and a physical activity sensor to track

movement patterns and exertion levels.

The collected sensor data is continuously analyzed

to detect deviations from normal physiological ranges.

If critical thresholds are exceeded, an alert is generated

for the user. A Gradient Boosting classifier is used to

analyse variations in heart rate, oxygen saturation, and

perfusion index, ensuring accurate detection of blood

flow abnormalities. Using historical data collected

over a 3-4-hour period, the system predicts heart

attack risks up to 20 hours in advance, allowing for

timely intervention and preventive measures. A

mobile application provides users with real-time

updates, notifications, and historical health data. If

critical risk levels are detected, the system

automatically generates alerts for the user and

designated emergency contacts. The wearable device

synchronizes with a cloud-based platform for

seamless data storage and remote access, and the

application allows users to personalize alert thresholds

and reporting preferences based on their health profile.

The figure 1 shows the Detailed Methodology of Aura

Wear.

Aura Wear combines advanced sensor technology

with machine learning techniques to enhance real-time

health monitoring and predictive analytics. This

system bridges the gap between conventional health

tracking and proactive cardiovascular risk

management, ultimately improving preventive care

and reducing medical emergencies.

Figure 1: Detailed Methodology of Aura Wear.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

718

4 MACHINE LEARNING

MODELS

4.1 Gradient Boosting

Gradient Boosting for Current Heart Risk Prediction

and Blood Flow Restriction Detection It is a powerful

ensemble learning technique used in Aura Wear for

both predicting current heart risk and detecting blood

flow restriction. By combining multiple weak

learners, this model enhances accuracy and

robustness in analyzing real-time sensor data. It

effectively identifies abnormalities in heart rate,

blood pressure, and oxygen saturation, providing

immediate risk assessments and ensuring timely

alerts for potential health concerns.

4.2 LSTM for Future Risk Prediction

Long Short-Term Memory (LSTM), a type of

recurrent neural network (RNN), is utilized to predict

future heart risk based on time-series data from

wearable sensors. LSTM captures long-term

dependencies in physiological patterns, allowing it to

forecast potential cardiovascular issues. This

predictive capability enables proactive health

management, giving users early warnings and

recommendations to mitigate future risks.

5 DATASET COLLECTION

This study utilizes two publicly available datasets to

develop a heart attack risk prediction model. The

Framingham Heart Study Dataset (sourced from

Kaggle) provides essential cardiovascular risk

factors, including age, blood pressure, cholesterol

levels, smoking status, diabetes, and other lifestyle-

related indicators. These features have been widely

used in predictive modeling for cardiovascular

diseases. Additionally, to incorporate physiological

parameters critical for heart health assessment, such

as heart rate and oxygen saturation (SpO2), the study

references the MIMIC-IV Clinical Database from

PhysioNet. This dataset includes real-world patient

monitoring data collected from intensive care units

(ICUs), making it highly relevant for medical

applications. The table 1 shows the Table 1: Sample

Dataset. By utilizing these datasets, the study ensures

a comprehensive approach to heart attack risk

prediction.

Table 1: Sample Dataset.

Age Gender Heart Rate BP O

Activity Stress

45 Male 78 120/80 98 Low High

60 Female 85 130/85 95 Low Medium

6 IMPLEMENTATION AND

EVALUATION

The heart attack risk prediction model was developed

using machine learning techniques, specifically Long

Short-Term Memory (LSTM) and Gradient Boosting.

The dataset was preprocessed by handling missing

values, normalizing numerical features, and encoding

categorical variables where necessary. The

Framingham Heart Study Dataset provided

cardiovascular risk factors, while the MIMIC-IV

Clinical Database contributed physiological

parameters such as heart rate and oxygen saturation

(SpO2). The figure 2 shows the Confusion Matrix

Accuracy. The dataset was split into 80% training and

20% testing, ensuring a balanced evaluation. Model

training and evaluation were conducted in Jupyter

Notebook using Python and key libraries, including

TensorFlow, Scikit-learn, Pandas, and NumPy. The

figure 3 shows the Model Performance Evaluation.

Performance was measured using accuracy, precision,

recall, F1-score, and AUC-ROC, with the final model

achieving an accuracy of 80%, demonstrating its

effectiveness in predicting heart attack risk.

Figure 2: Confusion Matrix Accuracy.

AI-Driven Heart Attack and Blood Flow Restriction Prediction Using A Smart Band Integrated with Wearable Sensors

719

Figure 3: Model Performance Evaluation.

7 WORK FLOW OF PROPOSED

MODEL

The process begins with loading the dataset, followed

by dividing the data into holdout and training sets.

Once the data is accessed, preprocessing is performed

to clean and prepare it for model training. The

approach involves two distinct machine learning

models: Gradient Boosting for evaluating current

heart attack risk and blood flow restriction, and

LSTM for predicting future heart attack risk. In the

Gradient Boosting pathway, the model is initially

trained using default settings. Next, hyperparameter

tuning is applied to optimize its performance,

followed by feature engineering to enhance the

predictive capabilities. Further refinements are made

through advanced optimization techniques. Once the

model is fully trained, its performance is evaluated

based on heart attack risk and blood flow restriction

detection. Similarly, in the LSTM pathway, the model

starts with default settings and undergoes

hyperparameter adjustments. Feature engineering

techniques are applied to improve the input data

representation, followed by advanced model

optimization for better accuracy. The figure 4 shows

the Flow chart of Proposed model. The trained model

is then evaluated to assess its effectiveness in

predicting future heart attack risk. After both models

complete their evaluations, an overall performance

assessment is conducted to compare and validate the

results. The final prediction is generated based on the

models' insights, and the outcomes are interpreted for

further analysis. The process concludes by

summarizing the results, ensuring that the predictions

provide meaningful insights into heart attack risk and

blood flow restriction detection.

Figure 4: Flow Chart of Proposed Model.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

720

8 RESULTS AND COMPARISON

The heart attack risk prediction model, developed

using Long Short-Term Memory (LSTM) techniques,

demonstrated promising results when evaluated on

the test dataset, providing valuable insights into

cardiovascular risk factors. The model achieved an

accuracy of 80%. These results highlight the potential

of the Aura Wear system in proactively identifying

individuals at risk of heart attack, enabling timely

interventions. Additionally, the blood flow restriction

detection model, built using Gradient Boosting

techniques, monitors key physiological parameters

such as heart rate, oxygen saturation, and perfusion

index. The quantitative results for this model are

currently being finalized. Compared to other

wearable technology projects focused on health

monitoring, Aura Wear stands out due to its real-time

data utilization, leveraging live physiological data

instead of relying solely on historical healthcare

datasets. This enhances the relevance and timeliness

of predictions, making the system more adaptive.

Moreover, Aura Wear integrates advanced machine

learning techniques, including both Gradient

Boosting and LSTM models, to improve predictive

accuracy and efficiency. Unlike many wearables that

focus on a single health aspect, Aura Wear offers dual

functionality by incorporating both heart attack risk

prediction and blood flow restriction detection within

a single system, making it a more comprehensive and

innovative solution for health monitoring.

9 CONCLUSIONS

Aura Wear represents a groundbreaking advancement

in personalized, AI-driven healthcare, with the

potential to revolutionize cardiovascular health

management and promote independent living,

especially for vulnerable populations such as older

adults and individuals with pre-existing

cardiovascular conditions. While the study

acknowledges certain limitations, including a

restricted dataset size, limited population diversity,

and the ongoing refinement of quantitative results for

blood flow restriction detection, the heart attack risk

prediction model—achieving 80% accuracy—

demonstrates its potential for early intervention and

improved patient outcomes. As highlighted in our

comparative analysis, Aura Wear outperforms

existing wearable solutions by offering unique dual

functionality (simultaneously monitoring heart attack

risk and blood flow restriction), real-time data

processing via advanced sensor fusion, and proactive

intervention capabilities, making it a powerful tool

for continuous health monitoring and timely

responses to critical events. The system’s ability to

provide real-time alerts and remote health monitoring

through a user-friendly mobile application is

especially beneficial for individuals with limited

access to medical care, such as older adults or those

in rural areas, empowering self-management and

facilitating timely access to necessary care. Moving

forward, future research should focus on refining

algorithms with more diverse and representative

datasets, conducting clinical trials to validate the

system's effectiveness across various demographic

groups, and addressing privacy concerns. Ethical

considerations, such as ensuring robust protection of

sensitive health data and maintaining user privacy,

must remain a priority to prevent misuse and foster

trust. By bridging the gap between traditional health

tracking and proactive cardiovascular risk

management, Aura Wear has the potential to redefine

how individuals monitor and manage their health,

reducing reliance on hospital visits, enhancing overall

well-being, and paving the way for a more accessible,

equitable, and personalized healthcare future for all.

REFERENCES

A. A. Ahmad and H. Polat, “Prediction of heart disease

based on machine learning using jellyfish optimization

algorithm,” Int. J. Adv. Comput. Sci. Appl. (IJACSA),

vol. 15, no. 1, pp. 143–152, 2024, doi:

10.14569/IJACSA.2024.0150117.

A. Ande, J. A. Jenifer, S. Jayesh, and F. V. Jacob, "Smart

wrist band for fitness monitoring," IEEE, [Online].

[Accessed: Feb. 28, 2025].

B. Anbuselvam, B. M. Gunasekaran, S. Srinivasan, M.

Ezhilan, V. Rajagopal, and N. Nesakumar, “Wearable

biosensors in cardiovascular disease,” Clinica Chimica

Acta, vol. 551, p. 119766, Jun. 2024, doi:

10.1016/j.cca.2024.119766.

C. M. Bishop, Pattern Recognition and Machine Learning.

New York, NY, USA: Springer, 2006.

C. Shuwen, Q. Jiaming, F. Shicheng, Q. Zheng, Y. J.

Chuan, and L. C. Teck, "Flexible wearable sensors for

cardiovascular health monitoring," Adv. Healthc.

Mater., vol. 10, no. 13, p. 2100116, Jul. 2021, doi:

10.1002/adhm.202100116.

H. El-Sofany, B. Bouallegue, and Y. M. Abd El-Latif, “A

proposed technique for predicting heart disease using

machine learning algorithms and an explainable AI

method,” Sci. Rep., vol. 14, no. 23277, 2024.

IEEE, “Heart attack prediction using machine learning

approach,” 2024 IEEE International Conference on

Bioinformatics and Biomedicine (BIBM), pp. 1–6,

2024, doi: 10.1109/BIBM59803.2024.10381192.

AI-Driven Heart Attack and Blood Flow Restriction Prediction Using A Smart Band Integrated with Wearable Sensors

721

M. A. E. M. D. Badawy et al., “Evaluation of cardiovascular

diseases risk calculators for CVDs prevention and

management: scoping review,” BMC Public Health,

vol. 22, no. 1742, Sep. 2022, doi: 10.1186/s12889-022-

14111-z.

M. Rizwan, S. Arshad, H. Aijaz, R. A. Khan, and M. Z. U.

Haque, “Machine learning-based risk assessment for

cardiovascular diseases,” BMC Med. Inform. Decis.

Mak., vol. 22, no. 1, p. 279, 2022, doi: 10.1186/s12911-

022-02020-0.

S. Raschka and V. Mirjalili, Python Machine Learning:

Machine Learning and Deep Learning with Python,

scikit-learn, and TensorFlow 2, 3rd ed. Birmingham,

UK: Packt, 2019.

S. N. R. Vonteddu, P. K. Nunna, S. Jain, S. I. J., N. S., and

G. Diwakar, "Smart wearable wristband for patients'

health monitoring system through IoT," IEEE.

[Accessed: Feb. 28, 2025].

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

722