Page 65 - AIH-1-1

Artificial Intelligence in Health Advancing fetal health classification

et al. conducted a comparative study, evaluating various learning frameworks, fuzzy neural networks, and hybrid
[4]
machine learning algorithms to enhance fetal health feature selection approaches, respectively, highlighting
assessment. Kim et al. focused on ensemble methods, the continuous advancements in the field. The remaining
[5]
providing a comprehensive analysis of their application studies by Ma et al. , Xie et al. , Kumar et al. , Zhang
[38]
[39]
[37]
in fetal health prediction. Gupta et al. delved into fetal et al. , Liu et al. , Li et al. , Zhang et al. , Chen et al. ,
[6]
[42]
[43]
[41]
[40]
[44]
health monitoring using neural networks, offering valuable Zhao et al. , Yu et al. , and Liu et al. covered a range
[46]
[45]
[47]
insights into the comparative effectiveness of different of topics, including hybrid deep learning models, multi-
approaches. The research conducted by Cheng et al. , Li layer perceptron neural networks, adaptive neuro-fuzzy
[7]
[8]
et al. , and Wu et al. explored the application of genetic inference systems, radial basis function neural networks,
[9]
algorithms, support vector machines, and k-nearest self-organizing maps with particle swarm optimization,
neighbor approaches, respectively, emphasizing the recurrent neural networks with feature selection, deep
importance of optimizing features for fetal health diagnosis. convolutional neural networks, and various hybrid
Zhou et al. , Tan et al. , and Liu et al. contributed to intelligent systems, collectively contributing to the ongoing
[12]
[11]
[10]
the field by investigating the significance of random forest, advancements in maternal-fetal medicine.
hybrid intelligent systems, and gradient boosting machines
in fetal health prediction. Additionally, studies by Rao The results of these studies suggest that ML-based
et al. , Jiang et al. , and Xu et al. provided insights approaches can be used to improve the accuracy of
[13]
[14]
[15]
into feature extraction using wavelet transform, principal fetal health classification. However, there are still some
component analysis, and fuzzy systems, contributing challenges that need to be addressed before ML-based
to the diverse methodologies employed in fetal health approaches can be widely used in clinical practice. One
assessment. These studies collectively showcase the challenge is that ML models can be sensitive to the quality
expansive landscape of research in ML-based fetal health of the data used to train them. Another challenge lies in the
classification, underscoring its potential across various difficulty in interpreting ML models, making it difficult to
algorithmic approaches. understand why they make the predictions that they do.
The research conducted by Huang et al. , Guo et al. , Despite these challenges, ML-based approaches have
[16]
[17]
and Yang et al. offered a comprehensive understanding of the potential to revolutionize the way that fetal health is
[18]
fetal health classification using ensemble learning, adaptive assessed and managed. By providing a more objective and
neuro-fuzzy inference systems, and hybrid approaches accurate assessment of fetal health, ML-based approaches
with feature selection and deep learning, respectively. could help to improve outcomes for both mothers and
Furthermore, studies by Wang et al. , Chen et al. , and babies.
[19]
[20]
Zhang et al. explored fuzzy logic, genetic algorithms, 3. Proposed methodology
[21]
and deep learning models, providing diverse perspectives
on fetal health prediction. Liu et al. investigated the 3.1. Dataset
[22]
application of convolutional neural networks, while Wu In this research, the Ayres-de-Campos dataset , titled
[48]
et al. focused on a hybrid intelligent system for fetal “Sisporto 2.0: A program for automated analysis of
[23]
health classification. Ma et al. introduced the concept cardiotocograms” was utilized. This dataset, published
[24]
of adaptive resonance theory networks in fetal health in the Journal of Maternal-Fetal Medicine, is a valuable
prediction. The comparative study of deep learning resource for studying fetal health. It contains a collection
[25]
models by Yang et al. , recurrent neural networks by of cardiotocograms (CTGs) that provide information on
Li et al. , and self-organizing maps by Zhang et al. various fetal health indicators. The dataset’s availability and its
[26]
[27]
enriched the field’s understanding of the applications of relevance to this study make it an ideal choice for this research.
these techniques in fetal health assessment. Additionally,
Cheng et al. , Li et al. , and Li et al. contributed to 3.2. Feature selection
[28]
[30]
[28]
fetal health prediction using radial basis function neural A set of 22 features, each playing a distinct role in the
networks, particle swarm optimization, and hybrid genetic
algorithm-support vector machine models, respectively. comprehensive assessment of fetal health, was meticulously
Tang et al. , Zhao et al. , and Yu et al. investigated curated from the Ayres-de-Campos dataset. The selected
[31]
[33]
[32]
multi-kernel support vector machines, convolutional features, represented as float64 values for precise numerical
neural networks, and stacked denoising autoencoders, analysis, include:
showcasing the versatility of advanced machine learning 1. Baseline value refers to the baseline fetal heart rate, a
models in fetal health assessment. Zhang et al. , Wang fundamental indicator reflecting the average rate over
[34]
et al. , and Wu et al. explored hybrid ensemble a specific time period.
[36]
[35]
Volume 1 Issue 1 (2024) 59 https://doi.org/10.36922/aih.2121