НЕЙРОМЕРЕЖЕВА МОДЕЛЬ ПРОГНОЗУВАННЯ РІВНЯ РИЗИКУ ІМПЛАНТАЦІЙНОГО  ЛІКУВАННЯ ПАЦІЄНТІВ ІЗ ПАТОЛОГІЄЮ ШЛУНКОВО-СТРАВОХІДНОГО З’ЄДНАННЯ

O. I. Halushko

doi:10.32782/2521-6643-2025-2-70.1

O. I. Halushko Taras Shevchenko National University of Kyiv https://orcid.org/0009-0009-0275-6542

DOI: https://doi.org/10.32782/2521-6643-2025-2-70.1

Keywords: multilayer perceptron, artificial neural network, classification, clinical prediction, endoscopic indicators, implantation risk

Abstract

The article presents a mathematical model for predicting the risk level of potential therapeutic device implantation in patients with gastroesophageal junction pathology, based on a multilayer perceptron (MLP) artificial neural network. The model integrates 15 clinical, endoscopic, and morphometric indicators reflecting both the functional state of the esophagus and struc- tural changes in the mucosa. The model was developed using a dataset of 558 observations (401 for training, 157 for testing), ensuring a statistically reliable assessment of its predictive performance. The network architecture includes an input layer with 50 parameters (after variable encoding), a single hidden layer with 7 neurons, and an output layer with 5 nodes corresponding to risk levels (0–4). Training was performed using the backpropaga- tion algorithm with the Scaled Conjugate Gradient optimizer. The model demonstrated extremely high classification accuracy – 99.8 % for the training set and 99.4 % for the test set, with the area under the ROC curves (AUC) for each class equal to 1. Such stability indicates the absence of overfitting and high generalization ability of the model. The greatest contribution to the prediction comes from morphometric parameters characterizing the degree of gastric prolapse into the esophagus, diaphragmatic constriction indices, and the position of the Z-line, as well as endoscopic features of mucosal damage, including the presence of erosions and ulcers covered with hematin. The results demonstrate the potential of the neural network approach for automated patient stratification by risk level and informed selection of interventional strategies (including the implantation of Bravo, JSPH-1, BEST Capsule, EndoStim systems, etc.). The proposed model could serve as a foundation for the development of an intelligent clinical decision support system, improving diagnostic accuracy, enabling personalized treatment, and reducing the risks of unjustified invasive interventions in gastroenterological practice.

References

1. Al Bataineh A., Manacek S. MLP-PSO hybrid algorithm for heart disease prediction. Journal of Personalized Medicine. 2022. Vol. 12, No. 8. P. 1208. DOI: https://doi.org/10.3390/jpm12081208.
2. Pan L., Liu X., Zhu L., Yu Z., Wang J., Li X., Zhang W., Li R., Wang Z., Lu H., Yang S., Li P., Su Y., Hua W., Liang Y. Machine learning-based prognostic models for mortality in patients receiving implantable cardioverter defibrillators. Pacing and Clinical Electrophysiology. 2025. Vol. 48, No. 8. P. 906–916. DOI: https://doi.org/10.1111/pace.70008.
3. Liu S. W., Li P., Li X. Q., Wang Q., Duan J. Y., Chen J., Li R. H., Guo Y. F. Recent advances in machine learning for precision diagnosis and treatment of esophageal disorders. World Journal of Gastroenterology. 2025. Vol. 31, No. 23. P. 105076. DOI: https://doi.org/10.3748/wjg.v31.i23.105076.
4. Pace F., Buscema M., Dominici P., Intraligi M., Baldi F., Cestari R., Passaretti S., Bianchi Porro G., Grossi E. Artificial neural networks are able to recognize gastro-oesophageal reflux disease patients solely on the basis of clinical data. European Journal of Gastroenterology & Hepatology. 2005. Vol. 17, No. 6. P. 605–610. DOI: https://doi.org/10.1097/00042737-200506000-00003.
5. Wang C.-C., Chiu Y.-C., Chen W.-L., Yang T.-W., Tsai M.-C., Tseng M.-H. A deep learning model for classification of endoscopic gastroesophageal reflux disease. International Journal of Environmental Research and Public Health. 2021. Vol. 18, No. 5. P. 2428. DOI: https://doi.org/10.3390/ijerph18052428.
6. Amiri Z., Heidari A., Navimipour N. J. та ін. The deep learning applications in IoT-based bio- and medical informatics: A systematic literature review. Neural Computing and Applications. 2024. Vol. 36. P. 5757–5797. DOI: https://doi.org/10.1007/s00521-023-09366-3.
7. Shirwaikar R. D., Acharya U. D., Makkithaya K., M. S., Srivastava S., Lewis U. L. E. S. Optimizing neural networks for medical data sets: A case study on neonatal apnea prediction. Artificial Intelligence in Medicine. 2019. Vol. 98. P. 59–76. DOI: https://doi.org/10.1016/j.artmed.2019.07.008.
8. Nazari Y., Lngeroodi P. F., Maddahi M., Kobravi S., Amin M. R., Bargrizaneh A. A., Fouladi S. Artificial intelligence models and predicting implant success. Biomedical Research and Therapy. 2025. Vol. 12, No. 1. P. 7029–7038. DOI: https://doi.org/10.15419/bmrat.v12i1.949.
9. Lawenko R. M. A., Lee Y. Y. Evaluation of gastroesophageal reflux disease using the Bravo capsule pH system. Journal of Neurogastroenterology and Motility. 2016. Vol. 22, No. 1. P. 25–30. DOI: https://doi.org/10.5056/ jnm15130.
10. IBM Corp. IBM SPSS neural networks 26 documentation. IBM Corporation, 2019. URL: https://www.ibm.com/docs/en/spss-statistics/26.0.0?topic=features-neural-networks.

NEURAL NETWORK MODEL FOR PREDICTING RISK LEVEL OF IMPLANTATION TREATMENT IN PATIENTS WITH GASTROESOPHAGEAL JUNCTION PATHOLOGY

Abstract

References