“Sepsis Detection using Wireless Sensors and

Random Forest Algorithm”

PRESENTED BY :
AKHIL A : 4JK22CS005 Under the Guidance of
AMAL NAHAN N : 4JK22CS007 MR.SHARAD SHANDHI RAVI
DEEKSHITH V : 4JK22CS012 Asst. Prof., Dept. of CSE
NAVANEETH K : 4JK22CS031
 Contents
• Introduction
• Problem statement
• Objectives
• Literature Survey
• Methodology
• Expected outcome
• Timeline
• Conclusion
• References

02/04/2025 AJIET Mangaluru 2

 Introduction
• Sepsis is a life-threatening condition caused by the body’s
extreme response to infection, resulting in tissue damage and
organ failure.
• Early prediction of sepsis is critical for improving patient
outcomes and reducing mortality rates.
• According to the World Health Organization, sepsis affects
millions globally and accounts for approximately 20% of
deaths in hospitals, making it a leading cause of mortality in
Intensive Care Units.
• Our project aims to develop a predictive model for early
sepsis detection using machine learning techniques.
• By enabling healthcare providers to identify at-risk patients
quickly, we hope to enhance treatment effectiveness and
ultimately save lives.
02/04/2025 AJIET Mangaluru 3
 Problem Statement
Current sepsis prediction models have delays and do not monitor
vital signs in real-time, making it hard to act quickly. They also
deal with issues like data imbalance and complexity, which make
them difficult for healthcare providers to use effectively in
critical care situations.

02/04/2025 AJIET Mangaluru 4

 Objectives
• Develop a Random Forest-based machine learning model
that integrates patient vital signs, demographic details,
and health history to enhance early prediction of sepsis.

• Create a user-friendly interface for healthcare

professionals to easily enter patient lab details and other
critical information.

• Provide clear output regarding the patient's risk level for

sepsis, indicating whether the patient is at risk, to
facilitate timely intervention and improve patient
outcomes.
02/04/2025 AJIET Mangaluru 5
 Literature Survey
Author Title Methodology Features Challenges
[citation]
[1] Sepsis Prediction Model for Sepsis Prediction Model High accuracy at Slow response time, high false
Determining Sepsis vs. SIRS, (SPM) using hospital high thresholds, negatives, complex decision-making
qSOFA, and SOFA admission data EHR integration
(60,000+ patients)

[2] Machine Learning for Early Logistic Regression, Comprehensive Class imbalance, lower accuracy in
Prediction of Sepsis in ICU KNN, Naive Bayes, data, multiple certain models
Patients Survival Analysis algorithms

[3] Early Prediction of Sepsis Random Forest, LGBM, Early detection Imbalanced data, complex models
Using Machine Learning XGBoost on PhysioNet (up to 6 hours),
dataset better data
handling with
mixed filling
method
[4] Early Biomarker Signatures Random Forest with Effective Low interpretability, high false
for Sepsis data augmentation biomarker positives
detection,
improved
accuracy
[5] Random Forest Classifier for Random Forest on Good early Data imbalance, low specificity
Early Sepsis Detection MIMIC-III & eICU detection,
databases external validation

02/04/2025 AJIET Mangaluru 6

 Methodology
Hardware setup Google Cloud - Web
Server

MAX30102 ESP32 Random

(Heart rate Forest
Risk
and SpO2 Prediction Assessme
sensor) Model nt
Wi-Fi
DS18B20 Modul
(Temperatur e
e
Sensor) Clinical
Display
Data Entry
Results
Interface

Fig. 1 Block diagram proposed approach

02/04/2025 AJIET Mangaluru 7
• Sensor Module: This module consists of various sensors that
continuously monitor vital signs.
• ESP32: The ESP32 is a powerful SoC with built-in Wi-Fi
capabilities. It processes the data received from the sensor
module. This component facilitates the transmission of
processed data from the ESP32 to the web server.
• Web Server: A remote server that receives data from the
ESP32 and processes it using a machine learning model
(Random Forest). The server analyzes the incoming data to
predict the risk of sepsis
• User Interface for Clinical Experts: A web-based interface or
application that allows healthcare professionals to input
additional patient data, such as medical history and lab results.

02/04/2025 AJIET Mangaluru 8

 Expected Outcome
1. Early Detection of Sepsis: The system will enable timely
identification of patients at risk of developing sepsis by
continuously monitoring vital signs and applying machine
learning algorithms for prediction.
2. Improved Patient Management: Healthcare providers will
have access to real-time data and alerts regarding patients'
conditions, allowing for more informed decision-making
and prompt medical intervention.
3. Enhanced Predictive Accuracy: By integrating various
patient data, including vital signs and medical history, the
prediction model will improve its accuracy in assessing
sepsis risk, potentially leading to better clinical outcomes.

02/04/2025 AJIET Mangaluru 9

4. User-Friendly Interface: The web-based interface will
facilitate easy data entry and retrieval for clinical experts,
streamlining the process of monitoring patient conditions
and updating records.

02/04/2025 AJIET Mangaluru 10

 Time line
SI Description Time
No.
1. Project planning, research and proposal submission 15-10-2024
2. Initial hardware setup and dataset collection 30-10-2024
3. Random Forest model training and testing 20-11-2024
Web server integration and UI creation
4. Conference paper preparation and presentation 30-11-2024
5. Testing, validation and final adjustments 10-12-2024
Project completion

02/04/2025 AJIET Mangaluru 11

 Conclusion
• In conclusion, this real-time sepsis monitoring system has the
potential to greatly enhance patient care by enabling early
detection of sepsis through continuous vital sign monitoring.
• By utilizing machine learning for accurate predictions,
healthcare providers can intervene swiftly, reducing the risks
associated with sepsis.
• The user-friendly interface further supports clinicians in
entering essential patient data, improving the overall
effectiveness of the predictive model.
• Ultimately, this project aims to improve patient safety and
healthcare outcomes significantly.

02/04/2025 AJIET Mangaluru 12

 References
[1] Schertz, Lenoir, Bertoni, Levine, Mongraw-Chaffin, & T
homas(2023). Sepsis prediction model for determining
sepsis vs. SIRS, qSOFA, and SOFA.
[2] Alanazi, Aldakhil, Aldhoayan, & Aldosari,(2022). Machine
learning for early prediction of sepsis in intensive care
unit (ICU) patients. Department of Health Informatics,
College of Public Health and Health Informatics, King Saud
Ibn Abdulaziz University for Health Sciences, Riyadh, Saudi
Arabia.
[3] Shankar, Diwan, Singh, Nahrpurawala, & Bhowmick,(2021).
Early prediction of sepsis using machine learning. In
Proceedings of the 11th International Conference on Cloud
Computing, Data Science & Engineering (Confluence) (pp.
978-1-6654-1451-7). IEEE.
02/04/2025 AJIET Mangaluru 13
 [4] Zhou, Shao, Wang, & Wang, (2024). An early sepsis
prediction model utilizing machine learning and
unbalanced data processing in a clinical context. Volume
45.
[5] Lyra, Leonhardt, & Hoog Antink, (2019). Early
prediction of sepsis using random forest classification
for imbalanced clinical data. Medical Information
Technology, Helmholtz-Institute for Biomedical
Engineering, RWTH Aachen University, Germany.

02/04/2025 AJIET Mangaluru 14

 THANK YOU

02/04/2025 AJIET Mangaluru 15