Detection of Type 2 Diabetes Mellitus In Individuals With Normal Fasting Glucose Using A Multilayer Perceptron Neural Network

Type 2 Diabetes Mellitus (T2DM) is a chronic metabolic disease that can cause severe, life-threatening complications. Therefore, its early detection is crucial for timely prevention and treatment. Commonly, fasting blood glucose through laboratory testing (A1C) is used for large-scale screening of T2DM; however, some individuals with normal fasting glucose (NFG) levels have undiagnosed diabetes that is not detected through these tests.

The use of artificial intelligence algorithms for informed decision-making, based on the analysis of clinical data, offers an alternative solution to optimize medical care in healthcare centers where it is essential to minimize the time healthcare professionals spend on searching, retrieving, subjectively and objectively reviewing, analyzing, and planning in line with available resources.

This work proposes the implementation of a multilayer perceptron (MLP) neural network to support informed decision-making based on previously captured data and established clinical guidelines for clinical-diagnostic consistency, enabling the early prediction of Type 2 Diabetes Mellitus (T2DM).

The importance of analyzing the use of machine learning (ML) models as a feasible tool to predict T2DM—a chronic metabolic disorder characterized by abnormal blood glucose levels caused by ineffective utilization or insufficient production of insulin—allows for assessing its viability as an alternative to support informed medical decision-making. This is particularly critical in the public health sector, where resources are limited, and vulnerable populations are served.