doi: 10.18178/ijml.2024.14.3.1165
Prediction of CD4 T-Lymphocyte Count via Machine Learning for HIV-positive Patients
Saad Lamjadli1,*
,
Oumayma Ouedrhiri2 ,
Ikram Souli1 ,
Zouhair Elamrani Abou Elassad4 ,
Oumayma Banouar2 ,
Safa Machraoui1 ,
Moulay Yassine Belghali1 ,
Raja Hazime1 ,
Noura Tassi3 ,
Said Raghay2 ,
Brahim Admou1,5
- 1Laboratory of Immunology, Center of Clinical Research, Mohammed VI University Hospital, Marrakech, Morocco
- 2Laboratory of Computer Science Engineering. Marrakech, Morocco
- 3Infectious Diseases Department, CHU Mohamed VI. Marrakech, Morocco
- 4SARS Research Team, Computer Science Department, ENSAS, Cadi Ayyad University, Marrakech, Morocco
- 5Biosciences Research Laboratory, Faculty of Medicine and Pharmacy, Cadi Ayyad University, Marrakech, Morocco
- Manuscript receivedMay 12, 2024
- revisedJune 24, 2024
- acceptedJuly 4, 2024
- publishedSeptember 25, 2024
Abstract
The World Health Organization recommends routine immunological and virologic monitoring for all patients with Human Immunodeficiency Virus (HIV) infection. However, viral load and lymphocyte T CD4 (LTCD4) count analysis requires sophisticated equipment and qualified human resources. This creates a financial burden, especially in limited resource settings. Thus, there is a need for alternative approaches. One such alternative is machine learning (ML), which offers a more cost-effective solution. In this study, five highly optimized data-driven models for LTCD4 prediction were designed based on popular ML techniques: support vector machine (SVM), random forest (RF), logistic regression (LR), artificial neural networks (ANNs), and naive Bayes (NB). To guarantee the robust performance of the proposed algorithms, we meticulously scrutinized the optimal approach for constructing models. Furthermore, we analyzed the predictive capabilities of LTCD4 according to multiple thresholds of the total lymphocyte count. Moreover, an imbalance-aware strategy to overcome the aforementioned issue was adopted using the synthetic minority oversampling technique. The cutoff points for the number of lymphocytes “1100” had the best performance in predicting the LTCD4 count. SVMs, RF, NB, LR, and ANNs provided an area under the curves of 97%, 93.2%, 90%, 92.01%, and 93%, respectively. SVMs achieved better results in predicting LTCD4 in all metrics. The results offer novel perspectives on LTCD4 forecasting, presenting opportunities to enhance initiatives aimed at developing web-based systems. These systems could alleviate the financial burden associated with measuring LTCD4 in patients with HIV infection, particularly in resource-constrained settings.
Keywords
- HIV
- CD4 T lymphocyte count
- total lymphocyte count
- machine learning
How to Cite
Saad Lamjadli, Oumayma Ouedrhiri, Ikram Souli, Zouhair Elamrani Abou Elassad, Oumayma Banouar, Safa Machraoui, Moulay Yassine Belghali, Raja Hazime, Noura Tassi, Said Raghay, and Brahim Admou, "Prediction of CD4 T-Lymphocyte Count via Machine Learning for HIV-positive Patients," International Journal of Machine Learning, vol. 14, no. 3, pp. 97-106, 2024. https://doi.org/10.18178/ijml.2024.14.3.1165
Copyright & License
Copyright © 2024 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).