Page 70 - AIH-1-1

Artificial Intelligence in Health Advancing fetal health classification

6. Limitations, challenges, and future work 6.3. Future works
in the field of fetal health classification 6.3.1. Data collection

6.1. Limitations Collecting more data is the prime direction to advance
6.1.1. Data availability ML-based fetal health classification in future. This data
should be of high quality and should be representative of
One of the main limitations of ML-based fetal health the population of pregnant women.
classification is the availability of data. Fetal health data
are often difficult to obtain, as specialized equipment and 6.3.2. Data cleaning and preprocessing
expertise are required in the data extraction process. This Future work should also focus on cleaning and
significantly restricts the training and evaluation of ML preprocessing the data. This will help to improve the
models for fetal health classification. accuracy of ML models.
6.1.2. Data quality 6.3.3. Model development
Another limitation of ML-based fetal health classification Future work should focus on developing more accurate
is the quality of data. Fetal health data can be fraught with and interpretable ML models. This could be done by using
noises and incompleteness issue, which render the training different ML algorithms or by incorporating additional
of accurate ML models difficult. features into the models.

6.1.3. Bias 6.3.4. Evaluation
ML models can be biased, which means that they can Evaluating the performance of ML models in clinical
make inaccurate predictions for certain groups of people. settings should be one of the key future directions in this
This is a particular concern for ML-based fetal health area. This will help to ensure that the models are safe and
classification, as it could lead to inaccurate predictions for effective.
pregnant women from minority groups.
6.3.5. Regulations
6.2. Challenges
Future work should also focus on developing regulations
6.2.1. Interpretability governing the use of ML models in fetal health
One of the challenges of ML-based fetal health classification classification. This will help to ensure that the models are
is interpretability. ML models are often regarded as black safe and effective.
boxes, meaning that it is to understand why they make the Overall, ML has the potential to revolutionize the way
predictions that they do. Therefore, this leaves users having that fetal health is assessed and managed. However, there
low level of trust in ML models and discourages them to are a number of limitations, challenges, and future work
use these models in clinical practice. that need to be addressed before ML can be widely adopted
in clinical practice.
6.2.2. Regulations
ML-based fetal health classification is a relatively new area, 7. Implementation and dataset
and there are currently no regulations governing the use of The source of the implementation code and dataset used
ML models for this purpose. This could lead to concerns for the experiments described in this paper are disclosed in
about the safety and efficacy of ML models for fetal health the availability of data section (in back matter). This allows
classification. researchers and practitioners to reproduce results and
validate methodology presented in this paper, and further
6.2.3. Ethical considerations
explore the proposed approach. The code and dataset
Ethical concerns are an inevitable aspect of the integration used in this study are available at and , respectively.
[49]
[48]
of ML in obstetrics for fetal health classification. Issues such The code is written in Python, and the dataset is in CSV
as data privacy, consent, and potential biases in training format. The code is well-documented and easy to follow.
data may arise, underscoring the importance of responsible The dataset is large and diverse, and it includes a variety of
and fair use of this technology in the rather sensitive health- features that can be used to train and evaluate ML models
care domain. Addressing ethical considerations becomes for fetal health classification. These resources are valuable
crucial to ensure the equitable and respectful application to the research community and will help to advance the
of ML models in the field of obstetrics. field of ML-based fetal health classification.

Volume 1 Issue 1 (2024) 64 https://doi.org/10.36922/aih.2121