Correct and timely diagnosis of an early-stage diabetes is important in order to ensure proper patient care and correct treatment regimen while avoiding possible serious complications. For this reason, a lot of research is performed with aim to support the process of medical decision making in this area, including application of data processing models based on machine learning.
Woo Seok Lee, Junghyo Jo, and Taegeun Song have discussed this particular issue in their research paper titled “Machine learning for the diagnosis of early-stage diabetes using temporal glucose profiles” that forms the basis of the following text and is aimed to introduce the machine learning algorithm to analysis of blood glucose profiles.
Machine learning could effectively facilitate the process of early and correct diagnosis for diabetes patients. Credit: Pixabay, free licence
Importance of this research
Diabetes is a chronic disease that causes long-term damage, dysfunction, and failure of diverse organs resulting in complications. The chronic nature and the long latent period of the disease makes it difficult to identify during the early stages. The researchers have proposed a Machine Learning Model to identify early-stage diabetes with an accuracy above 85%. The proposed ML model could be an effective way to identify diabetes earlier and manage it much effectively.
In our body, blood glucose levels (BGL’s) are tightly regulated by two counter-regulatory hormones, insulin, and glucagon. The endocrine pancreas releases insulin that helps with glucose homeostasis, which helps to maintain BGL’s.
How can we decide if a person is Diabetic?
Normal fasting glucose concentration is about 4 mmol/L. The American Diabetes Association Guideline defines hyperglycemia as 5.6 < BGL < 7 mmol/L. Severe hyperglycemic (BGL > 7.8 mM average at 2 hours fasting) is defined as diabetes mellitus (DM)
Types of Diabetes
There are three types of diabetes
- Type1 Diabetes: Type1 Diabetes refers to a condition where the pancreas does not produce enough insulin. Artificial pancreas can help patients with Type1 Diabetes.
- Type2 Diabetes: Most common (~90% of the cases) type of Diabetes. Type2 Diabetes occurs due to insulin resistance, which refers to a condition where the body is producing enough insulin, but it cannot reach cells, causing the glucose levels in the blood to rise.
- Gestational Diabetes: Temporary condition where BGL’s are elevated during pregnancy.
The Proposed Machine Learning Model
The researchers have proposed a Machine Learning Model that predicts diabetes by considering factors such as age, gender, BMI, waist circumference, smoking, job, hypertension, residential region (rural/ urban), physical activity, and family history of Diabetes. The researchers have monitored the increment of insulin resistance from the time trend of BGL to predict Type-2 Diabetes.
Results
The accuracy of the proposed model ranged from 70% to 90%
Future Work
Wearables provide for a non-invasive method for Continuous-glucose-monitoring. This monitoring that instructs the artificial pancreas to pump insulin as needed is very effective for Type1 diabetes patients. As more accurate diagnostic data becomes available for researchesr, the ML models should be improved accordingly. The abundance of rich data will help the medical specialists to detect diabetes much earlier and manage it much more effectively.
Conclusion
In the words of the researchers,
We checked whether machine learning could detect the patterns of BGL under insulin resistance. The temporal change of BGL results from the balanced response to the counter-regulatory hormones, insulin, and glucagon. Thus the ineffective action of insulin, called insulin resistance, should affect the BGL profile. Therefore, we simulated the glucose profiles under insulin resistance by using a biophysical model for the glucose regulation, and confirmed that the subtle change of glucose profiles under insulin resistance could be recognized by various machine-learning methods. This demonstrates a great potential of the machine learning approach for the diagnosis of early-stage Diabetes.
Source: Woo Seok Lee, Junghyo Jo and Taegeun Song’s “Machine learning for the diagnosis of early-stage diabetes using temporal glucose profiles”
Leave a Reply