DRLPARTO: A Machine Learning Based Partograph for Fetal

Monitoring System

Deepa J., Nandhyala Geetha Reddy, P. Sai Kiran and P. V. Sai Ram Reddy

Department of Information Technology, Vel Tech Rangarajan Dr. Sagunthala R&D Institute of Science and Technology,

Avadi, Chennai, Tamil Nadu, India

Keywords: Partograph, Labor, Fetal, Delivery, Prediction, Amniotic Fluid, Cervix Dilation.

Abstract: The review focuses on predicting the mode of delivery using machine learning techniques. Our approach

involves developing a machine learning model that evaluates partograph data to anticipate possible

complications or the necessity for medical interventions. Although various perspectives exist on the

application of partographs, a comprehensive understanding of their implementation remains unclear. The

proposed model assesses multiple parameters, considering the mother’s health status, fetal condition, and the

ongoing labor progression. The primary goal is to identify the most effective algorithms for predicting

delivery outcomes, specifically distinguishing the types of delivery i.e, normal or cesarean deliveries using

machine learning techniques. Supervised learning algorithms such as Decision Trees, Random Forest, and

Logistic Regression were employed with the proposed method DRLPARTO achieving 93% accuracy and

consistently high precision, recall, and F1 score (92-93%), demonstrating its robustness and effectiveness in

the given task.

1 INTRODUCTION

A partograph is a tool used to track labor progress and

monitor the health of both the mother and baby. It

records data in a graphical format, helping doctors

make decisions if complications arise during

childbirth. To predict delivery outcomes, we develop

machine learning models using parto- graph data.

According to the Government of India, around 1.3

million Indian women have died during childbirth in

the past two decades due to various reasons. The

partograph plays a crucial role in understanding the

health of the mother and baby, identifying

complications early, and making labor safer. It also

helps healthcare workers better understand labor

patterns.

Partographs have been the subject of extensive

research, but because of a lack of standardized

implementation, their practical application is still

restricted. The goal is to close the gap between data-

driven insights and practical clinical decision-making

by incorporating machine learning into partograph

analysis. The results demonstrate artificial

intelligence's potential.

If we can predict the type of delivery (normal or

C-section), we can prevent unnecessary medical

complications, ensuring the safety of both the mother

and baby. This also reduces risks during childbirth

and helps lower death rates. The World Health

Organization (WHO) supports monitoring active

labor and encourages new approaches during this

phase. The main goal of this project is to analyze

health conditions during labor and predict the

delivery type using machine learning. To understand

why partographs are important, we must look at their

history. The idea of recording labor progress in a

graph started in 1952 when Dr. Emanuel Friedman

couldn’t be present for his first child’s birth due to

work. He contacted the hospital frequently and

recorded his wife’s cervical dilation on a graph. He

continued observing all women in the maternity ward

throughout the night. Though contraction frequency

wasn’t very informative, he noticed that cervical

dilation followed an S-shaped curve, now called the

“sigmoid curve.” Since his first recording happened

on the day his child was born, the concept of the curve

is believed to have started in 1952.

Friedman later decided to divide labor into two

phases: latent and active, published a study analyzing

100 women, recording cervical dilation and

contractions per centimeter. This graph became

known as the “Friedman Curve” or “cardiograph”. In

J., D., Reddy, N. G., Kiran, P. S. and Reddy, P. V. S. R.

DRLPARTO: A Machine Learning Based Partograph for Fetal Monitoring System.

DOI: 10.5220/0013932800004919

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 5, pages

553-558

ISBN: 978-989-758-777-1

553

1955, this study expanded to 500 women and

published another research paper. Figure 1 shows

Phases of Labour representation.

Friedman divided labor into four phases based on

cervical dilation as in Figure 1:

• Latent phase - Slow dilation (up to 2.5 cm).

• Acceleration phase - Rapid dilation with a

change in speed.

• Maximum slope - The cervix dilates steadily

in a straight- line pattern.

• Deceleration phase - The cervix is fully

dilated, and the rate of dilation slows down.

Later, Dr. Philpott developed a partograph

based on Friedman’s curve. He introduced

two lines:

Figure 1: Phases of labor representation.

The alert line, which acts as a warning. The action

line, which signals the need for medical intervention

if crossed. This system was later introduced in

England, where British gynecologist John Studd

made modifications. He replaced the alert and action

lines with a nomogram that used cervical dilation at

the time of admission as a reference.

In 1994, the World Health Organization (WHO)

officially approved the partograph. They

recommended its use in all labor wards to ensure safer

deliveries. This recommendation was based on a

study involving over 35,000 women. The use of

partographs significantly reduced maternal and infant

deaths and lowered the risks of prolonged labor. The

WHO also introduced an updated version called the

labor scale, which helps study labor conditions in

more detail. This is how the partograph has evolved

and developed over the years, becoming a vital tool

in maternal healthcare.

Section 2 provides an overview of related work,

while Section 3 outlines the proposed methodology.

The results and findings are discussed in Section 4,

and the conclusion is presented in Section 5.

2 RELATED WORKS

Khali et al. (2022) evaluated a digital version of the

partograph for labor management. The goal was to

enhance labor quality and prevent complications by

ensuring accurate monitoring. The study, which

involved 800 women, found that the average delivery

duration was 3.5 hours for first-time mothers and 3.3

hours for those who had given birth before. The

average participant age was 25.6 years, with an

average parity of 2. The findings showed that the

digital partograph effectively improved labor

management and newborn health outcomes.

Ashour E.S. et al. (2023) conducted a study in

Egypt to compare digital and paper-based

partographs. They found that digital partographs

improved maternity nurses’ performance, ensuring

better documentation and timely interventions.

Additionally, digital tools led to better maternal and

neonatal outcomes, reducing cesarean sections and

newborn intensive care admissions. The study

suggested that adopting digital partographs in labor

wards could enhance maternal and newborn health in

other regions.

Bedwell et al. (2017) analyzed the effectiveness

of the partograph in labor monitoring. Their review

followed five steps: defining the research scope,

gathering evidence, evaluating primary studies,

synthesizing findings, and sharing results. They

reviewed 95 sources, including research papers,

policies, audits, and expert opinions. The findings

showed that while the partograph is useful for

tracking labor, its effectiveness depends on factors

like healthcare worker competency and acceptance.

GJ Hofmeyr et al. (2021) introduced a new

approach to labor monitoring called the next-

generation partograph. This tool aims to offer

personalized care, focusing on respectful maternity

services. Key features include considering individual

differences in labor progression and setting

intervention triggers based on the health of the mother

and baby. The model promotes better communication

between healthcare providers and laboring women.

The study argues that adopting this advanced

partograph could improve health outcomes and

patient satisfaction.

Melese et al. (2020) studied the use of the

partograph in Southern Ethiopia. The research

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

COMMUNICATION, AND COMPUTING TECHNOLOGIES

554

assessed healthcare workers’ knowledge, attitudes,

and usage of the tool in obstetric care. While most

participants were aware of the partograph, their

willingness to use it was inconsistent. The study

found that factors such as equipment availability,

workload, and training influenced its use. The authors

emphasized the need to promote partograph use in

healthcare facilities to improve childbirth outcomes.

Lavender T et al. (2020) examined the role of the

partograph in labor management. Their study

discussed the history of the partograph, its challenges,

and its effectiveness in improving childbirth

outcomes. They explored different versions of the

partograph and the factors that affect its success. The

paper also suggested best practices and future

research directions to enhance labor monitoring tools.

The authors supported their conclusions with

references to multiple studies.

Xiaoqing He et al. (2023) reviewed 62 studies to

gain new insights into labor patterns. Their research

challenged traditional labor curves and definitions of

abnormal labor. The review emphasized that recent

studies suggest abnormal labor cannot always be

determined based on a standard labor curve. This has

significant implications for global childbirth care

practices.

Yeliz Dog˘an Merih et al. (2023) developed the

Electronic Touch and Partograph Device, an

innovation that combines vaginal examination with

digital labor monitoring. This device can track up to

50 patients simultaneously, ensuring real-time data

collection. Developed between 2016 and 2020, the

device underwent extensive research and ethical

reviews. It improves safety for both healthcare

workers and pregnant women, offering a promising

advancement in labor monitoring technology.

Singh et al. (2022) conducted a quality

improvement initiative to increase the use of the

modified WHO partograph in labor rooms. The aim

was to improve maternal and neonatal outcomes by

ensuring regular labor monitoring. The team

identified reasons for low partograph usage and

introduced solutions such as allocating triage rooms,

training staff, involving interns and nurses, setting

clear policies, and designating supervisory roles. As

a result, partograph usage increased from 25 Percent

to 95 Percent. Despite challenges like printer

malfunctions and misplaced documents, the initiative

successfully improved labor care quality.

Shivani Sharma et al. (2022) focused on

improving maternal health through a new partograph

model. The study aimed to reduce preventable

maternal deaths by identifying abnormal labor early

and providing timely interventions. Researchers

developed and validated a reliable partograph, which

proved effective in improving childbirth outcomes.

The study recommended widespread adoption of

partographs in labor monitoring to enhance maternal

and newborn safety.

3 PROPOSED SYSTEM

The traditional approach to labor monitoring

primarily relies on manual partograph recording to

track the progress of labor and ensure maternal and

fetal well-being. A partograph is a standardized tool

used by healthcare providers to document key

parameters such as cervical dilation, fetal heart rate,

uterine contractions, and maternal vital signs at

regular intervals. In conventional systems, paper-

based partographs are widely used in hospitals and

maternity wards. Healthcare professionals manually

record observations and assess labor progression

based on predefined thresholds. If abnormal labor

patterns are detected, appropriate clinical

interventions are initiated. However, this manual

process is prone to human errors, inconsistencies in

data recording, and delays in decision making.

Despite its importance, the adoption and effective

utilization of partographs remain limited due to

several challenges. These include insufficient

training, high workload on healthcare staff, and a lack

of standardized monitoring protocols. Additionally,

paper-based partographs lack real-time analysis and

do not provide predictive insights that could help

anticipate complications in labor.

Furthermore, while some digital partograph

solutions have been introduced, their adoption is still

in early stages. Many healthcare facilities continue to

rely on traditional manual methods, which may not

provide the level of efficiency required for modern

obstetric care. Thus, the existing labor monitoring

system heavily depends on manual data entry and

fixed threshold-based decision making, making it less

adaptable to variations in labor progression and real

time clinical needs. There is a growing necessity for

intelligent, automated, and predictive solutions that

can enhance the accuracy and efficiency of labor

monitoring, ultimately improving maternal and

neonatal outcomes.

The primary objective of this work is to develop a

machine learning model that analyzes partograph data

to predict potential complications during labor and

determine the need for medical interventions. The

model evaluates various maternal and fetal

parameters, including maternal status, fetal condition,

and labor progression, to classify the delivery type as

DRLPARTO: A Machine Learning Based Partograph for Fetal Monitoring System

555

either normal or cesarean. To achieve this, the study

leverages advanced machine learning techniques to

process and interpret partograph data effectively. The

system is designed to identify crucial patterns that

indicate deviations from normal labor, enabling

timely medical decisions. Various machine learning

algorithms are assessed to determine the most

accurate model for delivery type prediction.

This research aims to bridge the gap between

theoretical understanding and practical

implementation of partographs in clinical settings. By

integrating machine learning into obstetric care, the

system enhances decision-making, improves

maternal and fetal health outcomes, and reduces risks

associated with labor. Through extensive

experimentation and validation, this study contributes

to the advancement of intelligent healthcare solutions

for maternity care.

3.1 Methodology

Upon analyzing the reviewed research papers, it is

evident that multiple methodologies and techniques

have been applied to understand and implement the

partograph. However, despite numerous studies and

recorded data, the actual implementation remains

limited. To improve the effectiveness and accuracy of

labor monitoring, certain machine learning models

have been identified as optimal solutions. The

algorithms employed in this study include Decision

Trees, Random Forest, and Logistic Regression.

These models fall under the category of supervised

machine learning algorithms, which are utilized to

make accurate predictions based on historical data.

The functioning of each algorithm is elaborated

below:

3.1.1 Supervised Machine Learning

Algorithms

Supervised learning is a branch of machine learning

where models are trained on labeled datasets,

meaning the input data is associated with known

output values. By leveraging these pre-labeled

examples, the model learns patterns and makes

predictions when new, unseen data is introduced.

3.1.2 Decision Tree Classifier

A decision tree is a rule-based model that facilitates

decision-making by structuring the data into a

hierarchical tree format. It consists of decision nodes

and leaf nodes, where decision nodes represent

different choices based on feature values, and leaf

nodes signify the final output. This algorithm

effectively categorizes data through a sequence of

binary decisions, ultimately arriving at a conclusion

based on conditions such as yes/no or true/false

responses.

3.1.3 Random Forest

Random Forest is an ensemble learning technique

that enhances the performance of individual decision

trees by generating multiple trees and aggregating

their predictions. Each tree is built using a subset of

the training data, and the final prediction is obtained

by averaging (for regression) or voting (for

classification) among all trees. This method improves

accuracy, reduces overfitting, and ensures robust

predictions, making it highly effective in labor

monitoring applications.

3.1.4 Logistic Regression

Logistic regression is a statistical model used for

classification tasks, where the output is a discrete

value (such as 0 or 1, yes or no, or true or false). It

applies a sigmoid function to transform input values

into probability scores, determining the likelihood of

an event occurring. When the probability surpasses a

certain threshold, the outcome is classified as 1

(positive class); otherwise, it is classified as 0

(negative class). This algorithm is particularly useful

for predicting binary outcomes in medical and

healthcare applications.

By integrating these algorithms, we can achieve

reliable and efficient labor monitoring solutions,

overcoming some of the challenges and limitations

identified in prior research.

Figure 2: DRLPARTO: A partogragh tracer architecture.

The figure 2 shows the workflow of the machine

learning-based classification system. The process

starts with collecting datasets containing FHR,

contractions, pulse, amniotic fluid, BP, temperature,

and cervix length. The data undergoes processing and

cleaning to remove inconsistencies. The cleaned data

is then split into training data and testing data. The

training data is used to train machine learning

algorithms, including Decision Tree Classifier,

Random Forest, and Logistic Regression. After

training, a classification model is generated, which is

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

COMMUNICATION, AND COMPUTING TECHNOLOGIES

556

then used to make predictions based on new input

data.

4 RESULTS AND DISCUSSION

Figure 3 indicates that the cervix length increases

sharply at the beginning and then stabilizes over time.

This suggests that initial changes occur rapidly, but

further progression is minimal or steady.

Figure 3: Cervix dilation.

Phase-wise analysis of a process, where each

phase is categorized as either “Normal” or requiring

“Consult” as shown in figure 4. While most phases

are marked as “Normal,” Phase 4 and Phase 7

indicate” Consult,” suggesting potential concerns in

those stages. However, the final result is marked as

“Normal,” implying that despite some irregularities,

the overall outcome is within acceptable limits.

Figure 4: Delivery prediction.

Figure 5: Comparison of the proposed model with

individual models.

The table 1 presents a comparative analysis of

different machine learning algorithms based on four

key performance metrics: accuracy, precision, recall,

and F1 score. It is also shown graphically in figure 5.

The Decision Tree model achieved 85% accuracy,

with balanced precision, recall, and F1 scores around

83-84%. Random Forest performed better, attaining

89% accuracy and an F1 score of 91%, indicating

improved overall performance. Logistic Regression

showed similar effectiveness with 88% accuracy and

slightly lower recall. The proposed method

outperformed all models, achieving 93% accuracy

and consistently high precision, recall, and F1 score

(92-93%), demonstrating its robustness and

effectiveness in the given task.

Table 1: State of art comparison.

algorithm accuracy precision recall f1

score

decision

tree

85 83 84 83

random

forest

89 90 87 91

logistic

regression

88 87 86 86

proposed

metho

93 92 92 93

5 CONCLUSION AND FUTURE

WORK

Predicting the type of delivery a few hours before

labor begins can help ensure a safe childbirth.

Whether a woman delivers vaginally or through a

cesarean section can impact both her health and the

baby’s well-being. Therefore, making the right

decision is crucial. This can be done using a par-

tograph, a tool that is widely recognized and

DRLPARTO: A Machine Learning Based Partograph for Fetal Monitoring System

557

recommended by the World Health Organization

(WHO), highlighting its importance in labor

management. From our review of various studies, we

found that machine learning and artificial intelligence

can be used to implement this prediction system.

Many researchers have explored different ways to

apply these technologies to improve labor

monitoring. In our approach, we focus on supervised

machine learning algorithms such as decision trees,

random forests, and logistic regression. Using these

methods can help predict delivery outcomes with high

accuracy, ultimately improving maternal and

newborn care. The partograph plays a crucial role in

monitoring health conditions during labor. Since its

introduction, it has undergone several modifications

to improve its effectiveness. Various factors help

assess the mother’s health and the environment

necessary for a safe delivery, and these have been

refined over time. As seen from the analysis of

different studies, the partograph is highly beneficial

for healthcare providers, mothers, and newborns.

While many researchers have explored its theoretical

aspects and case studies, only a few have discussed

using machine learning algorithms for predicting

delivery outcomes. Some studies have highlighted

important parameters for labor monitoring, but

practical implementation is still lacking.In the future,

further advancements can be made by integrating

machine learning techniques and improving

implementation strategies. Through continuous

research, experimentation, and technological

innovation, the medical field can enhance labor

monitoring and improve maternal and neonatal health

outcomes.

REFERENCES

Bedwell, Carol, Karen Levin, Celia Pett, and Dame Tina

Lavender.” A realist review of the partograph: when

and how does it work for labor monitoring” BMC

pregnancy and childbirth 17 (2017): 1-11.

ES, Ashour, Tahany El-Sayed El-Sayed Amr, Gehan

Ahmed Mohmed Elbahlowan, and Samah Mohamed

Elhomosy.” Effect of Implementing Intrapartum

Digital versus Paper Partographs on Maternity Nurses’

Performance and Birth Outcomes.” Egyptian Journal of

Health Care 14, no. 1 (2023): 430-445.

He, Xiaoqing, Xiaojing Zeng, James Troendle, Maria

Ahlberg, Ellen L. Tilden, Joa˜o Paulo Souza, Stine

Bernitz et al.” New insights on labor progression: a

systematic review.” American Journal of Obstetrics and

Gynecology (2023).

Hofmeyr, G. Justus, Stine Bernitz, Mercedes Bonet,

Maurice Bucagu, Blami Dao, Soo Downe, Hadiza

Galadanci et al.” WHO next-generation partograph:

revolutionary steps towards individualized labor care.”

Bjog 128, no. 10 (2021): 1658.

Khalil, Abu Bakr, Sayed A. Mostafa, and Ayman D.

Mohamed.” Revaluation of Paperless Partograph in the

Management of Labor.” Suez Canal University Medical

Journal 25, no. 1 (2022): 35- 41.

Lavender, tina, and stine bernitz.” use of the partograph-

current thinking.” best practice research clinical

obstetrics gynaecology 67 (2020): 33-43.

Melese, Kidest Getu, Bedilu Girma Weji, Tezera Moshago

Berheto, and Eyasu Tamru Berkus.” Utilization of

partograph during labour: A case of Wolaita Zone,

Southern Ethiopia.” Heliyon 6, no. 12 (2020).

Merih, Yeliz Dog˘an.” An Artificial Intelligence-Assisted

Innovative Product in Labor Follow-Up: The Electronic

Touch and Partograph Device.” Journal of Education

Research in Nursing/Hems¸irelikte Eg˘itim ve

Aras¸tırma Dergisi 20, no. 2 (2023).

Sharma, Shivani, Saroj Parwez, Kiran Batra, and Bharat

Pareek.” Enhancing safe motherhood: Effect of novel

partograph on labor outcomes and its utility: An Indian

perspective.” Journal of Family Medicine and Primary

Care 11, no. 11 (2022): 7226.

Singh, Ritu, Mukta Agarwal, Sudwita Sinha, Hemali H.

Sinha, Monika Anant,” Modified WHO Partograph in

Labour Room: A Quality Improvement Initiative to

Find Out Concerns, Challenges and Solutions.” Cureus

14, no. 10 (2022).

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

COMMUNICATION, AND COMPUTING TECHNOLOGIES

558