MD. Tho Nguyen

03/06/2026

I am happy to share this beginner-friendly material on Logistic Regression for Public Health Students.

I wrote it with care for students and early-career researchers who want to learn statistics and machine learning without feeling overwhelmed.

Logistic regression may look like a technical method, but at its heart, it helps us answer very human questions:

Who is at higher risk?
Which factors matter?
How can data support better prevention and care?

In this document, I explain logistic regression using simple formulas, public health examples, and a practical learning style. My hope is that students can read it not with fear, but with curiosity and confidence.

Learning data science in public health is not only about models. It is about using evidence to protect people, families, and communities.

You can find documents at link below:
https://iad.donga.edu.vn/detail/new-learning-resource-logistic-regression-for-international-students-43926

With love for learning and public health,
Tho Nguyen, MD, MPH
International Research Institute for Artificial Intelligence and Digital Transformation, D**g A University, Danang, 50000, Vietnam

30/05/2026

Stroke prevention is not only about controlling risk factors. It is also about preventing delay.

A recent article in Scientific Reports proposed a real-time stroke monitoring system that combines smart cameras, edge/fog computing, YOLOv8 for facial recognition, and a CNN model to detect abnormal changes in a patient’s face. When the system identifies a possible problem, an alert can be sent to physicians and healthcare staff.

The key question is simple but important:

What if facial drooping, facial asymmetry, or facial weakness could be detected before family members or caregivers notice it?

This idea may be especially meaningful for older adults, high-risk patients, and people living alone.

From a preventive medicine perspective, the greatest value of AI in this context is not that it replaces clinicians. Its real value is that it may help shorten the silent interval between the first appearance of symptoms and the first medical response. This is crucial because stroke warning signs such as sudden weakness of the face, arm, or leg require urgent action, and early hospital arrival can affect treatment opportunities.

In simple terms, stroke prevention is not only about controlling blood pressure, diabetes, smoking, dyslipidemia, or physical inactivity. It is also about detecting symptoms earlier, sending alerts faster, intervening more promptly, and reducing the time lost before treatment.

In stroke care, every minute matters.

From my perspective, future prevention strategies may increasingly explore how AI can support early detection and rapid response, helping reduce the risk that stroke is recognized too late.

Reference:
https://www.nature.com/articles/s41598-025-28513-5

18/05/2026

Can artificial intelligence help identify people at high risk of diabetes earlier?

Diabetes is a long-term health condition that can lead to serious complications if it is not detected and managed in time. Early screening is very important, especially for women and people with risk factors such as high blood glucose, high BMI, family history, or pregnancy-related metabolic changes.

In our study, we tested AI models that can learn from routine health data. The data included simple health indicators such as: Glucose level, BMI, Age, Blood pressure, Insulin, Pregnancy history, Family-related diabetes risk.

The goal was not to replace doctors. The goal was to see whether AI can support healthcare workers by identifying people who may need further testing, counseling, or follow-up.

The encouraging point is that these models performed well in classifying diabetes risk. More importantly, we used explainable AI methods to understand the prediction. The model showed that glucose level was the most important factor, and pregnancy-related information also played a role.

This makes the result easier to understand:

AI is not just giving a label.

It can help show which health factors may be contributing to diabetes risk.

For community health, this can be useful in several ways:

- Early diabetes screening
- Health check-up programs
- Women’s health monitoring
- Remote or mobile health services
- Identifying people who need follow-up care
- Supporting doctors and nurses in decision-making

However, AI should always be used carefully. A prediction is not a diagnosis. People classified as high risk still need proper clinical testing and medical consultation.

The future direction is to build models using local Vietnamese and Asian health data so that AI tools can become more accurate, practical, and meaningful for our communities.

Read more details at:
- Paper 1: https://link.springer.com/chapter/10.1007/978-3-032-14055-5_6
- Paper 2: https://arxiv.org/pdf/2207.01848

Diabetes can lead to several long-term severe health complications; more than 199 million women are affected by diabetes, with projections indicating a rise to 313 million by 2040. Feature Tokenizer Transformer (FT-Transformer) is proposed to predict diabetes in 1858...

17/03/2026

A cross-sectional study is science's way of taking a photograph of a population — capturing who has a condition, who doesn't, and what factors seem to go with it, all at a single point in time. It is one of the most widely used tools in public health, and one of the easiest to understand once you see it clearly.

A cross-sectional study takes a single snapshot of a group of people at one moment in time. No follow-up. No before and after. Just: right now, who has this condition, and who doesn't?

Limitations of this study:
+ Chicken-and-egg problem: Did heavy screen time cause poor sleep, or did students who sleep badly end up scrolling more? This study cannot answer that.
+ Self-report bias: Students may over- or under-report their screen time and sleep quality. Objective data (wrist sensors, phone usage logs) would strengthen the study.
+ Confounding variables: Anxiety, caffeine use, roommate noise, study stress — all could independently cause poor sleep AND correlate with screen use. This study doesn't control for them.
+ What this study IS good for: It establishes that the association exists and justifies a longitudinal cohort study to test causation. Cross-sectional research is where most hypotheses begin.

Receive the full document at: https://drive.google.com/drive/folders/1kOtMEkb1R46Ij-wa-frjM5rk10rS0cpX?usp=sharing

02/10/2022

Do you think like me? 😇🥲🥲

Sometimes, my question took one week to get the answer 🥹

Source: Internet

24/09/2022

There are the top 10 Data science use cases that were used by retail industries to grow their industries in today’s world.

1. Price optimization
2. Personalized Marketing
3. Fraud detection in Retail
4. Utilizing Social Media
5. Implementing Augmented Reality
6. Merchandising
7. Location of New Store
8. Inventory Management
9. Customer Sentiment Analysis
10. Recommendation System

Read more information at: https://www.analyticsvidhya.com/blog/2021/05/data-science-use-cases-in-retail-industry/

In this article let's see the top 10 Data science use cases that were used by retail industries to grow their industries in today's world.

17/09/2022

How is your model? 👀👀

Source: internet

12/09/2022

Do you agree with us?
Source: Twitter