The Application of Deep Learning in thePrevention and Diagnosis of Type 2 Diabetes Mellitusof the Specialization in Health Informatics

Abstract

This study analyzed the application of deep neural networks in the early diagnosis and prediction of complications of Type 2 Diabetes Mellitus (T2DM). The main objective of the study was to investigate how these technologies contribute to improving diagnostic and predictive accuracy in relation to traditional methods. An integrative literature review was used, covering articles published between 2019 and 2024 in databases such as PubMed, IEEE Xplore, ScienceDirect and Google Scholar. The analysis included models such as LSTM, CNN and RNN, highlighting their superior performances in metrics such as accuracy (up to 85%) and area under the curve (AUC) (up to 0.98). The results showed that deep neural networks allow more personalized predictions, identifying glycemic control trajectories and categorizing patients into risk groups. Thus, the integration of genomic, tabular and clinical data proved to be essential for the personalization of clinical management. However, limitations were identified, such as the dependence on large volumes of data and the high computational cost, which may restrict large-scale adoption. It was concluded that deep learning is a promising tool in the management of T2DM, promoting significant advances in precision medicine. Future research should focus on optimizing models and expanding data sources for greater accessibility and practical applicability.