| 1 |
What is the primary goal of the article according to its introduction?
|
3. To explore advancements, applications, and challenges of generative AI in medical imaging |
|
This viewpoint provides a comprehensive overview of synthetic data in medical imaging and critically analyses the advancements, applications, and challenges of this field. The introduction focuses on explaining how generative AI models, such as Med-PaLM, Med-Gemini, and MedImageInsight, are transforming medical imaging by enabling multimodal data integration and image generation
|
The article states that it aims to analyze the advancements, applications, and challenges of generative AI in medical imaging. This is supported by references 1 - 10 in the introduction, which discuss developments in large language models and generative imaging models such as ChatGPT, Med-PaLM, DALL-E, and Stable Diffusion.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 2 |
How do generative AI models differ from traditional discriminative models in healthcare applications?
|
2. Generative models produce new data rather than only classify or interpret |
|
This proposed definition emphasizes the functional and intentional aspects of synthetic data, focusing on its strategic application in tackling complex scientific challenges rather than simply mimicking the statistical properties of the original data.
|
Generative models learn the underlying distribution of data to produce new samples, while discriminative models focus on classification or prediction. According to Jordon, Szpruch, Houssiau et al. (2022) in “Synthetic data – what, why and how?” (arXiv, 2022), generative models are designed to synthesize realistic data that can capture essential patterns of the original dataset without duplicating it.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 3 |
What is meant by the term “model as a dataset”?
|
2. A dataset created manually by experts |
|
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 4 |
Which statement correctly distinguishes physics-informed and statistical models?
|
3. Physics-informed models incorporate biological or physical principles |
|
Physics-informed models are primarily rule-based approaches that incorporate domain-specific knowledge and physics principles through mathematical equations and explicit constraints to generate realistic and physically plausible data. Rather than learning the patterns directly from data, these models encode expert knowledge and known physics laws (eg, fluid dynamics, tissue biomechanics, or radiation physics) to simulate biological phenomena.
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 5 |
According to the article, what does the “image generation trilemma” describe?
|
2. Trade-offs among image diversity, quality, and speed |
|
The image generation trilemma, which represents the trade-offs between three key aspects of generative models: diversity, quality, and speed
VAEs excel in generating diverse samples quickly but can compromise on image quality. GANs strike a balance, providing good quality and diversity but can suffer from mode collapse, thereby restricting the diversity. DDPMs prioritise high-quality and diverse samples at the cost of a slow generation speed.
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 6 |
What is the Human Turing Test used for in medical image synthesis?
|
2. To assess realism of synthetic medical images by experts |
|
The human Turing test involves domain experts who are asked to discern between real and derived medical images.42 This assessment provides insights into the perceptual quality and realism of generated images, which is crucial for medical imaging, in which accuracy and fidelity are paramount
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 7 |
Which of the following is NOT mentioned as a potential benefit of synthetic data in healthcare?
|
2. Preserving patient privacy |
|
Although synthetic datasets can help to preserve patient privacy by generating anonymized data, concerns regarding potential data copying still exist.If a generative model is trained on a specific dataset and can replicate images that closely resemble the original data, then the model might inadvertently reveal sensitive patient information
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 8 |
What is one major ethical concern associated with generative AI in medical imaging?
|
2. Data copying and patient reidentification |
|
Generative models can inadvertently reveal sensitive patient information when they reproduce images that closely resemble the original data.
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 9 |
What regulatory precedent did the article cite for synthetic data technologies?
|
|
|
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 10 |
What is the main purpose of the article?
|
2. To compare and evaluate ASCVD risk prediction models in East Asia |
|
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 11 |
Which of the following models was originally developed for a Western population?
|
1. Framingham Risk Score |
|
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 12 |
Why might Western-based risk prediction models overestimate ASCVD risk in East Asian populations?
|
2. East Asians have lower baseline incidence of ASCVD |
|
In this review, we highlight the similarities and differences in the epidemiology, diagnosis, and treatment of ASCVD for individuals of East Asian origin who immigrated to the United States and their offspring (“East Asian Americans”) compared with those living in East Asia (“East Asian natives”)
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 13 |
What is the key advantage of the China-PAR model compared to Western-based models?
|
1. It includes both genetic and lifestyle factors |
|
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 14 |
Which of the following variables is not typically included in ASCVD risk prediction models discussed in the article?
|
4. Genetic ancestry markers |
|
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 15 |
What is a major difference between the Suita Score and the Framingham Risk Score?
|
2. Suita Score was designed for a Japanese population using local epidemiological data |
|
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 16 |
According to the article, what is a potential benefit of developing East Asia–specific risk models?
|
3. They improve accuracy and reduce overestimation of risk |
|
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 17 |
Which factor was highlighted as influencing ASCVD risk differences among East Asian countries?
|
2. Cultural and dietary variations, such as salt intake and lifestyle |
|
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 18 |
What future direction does the article suggest for improving ASCVD risk prediction?
|
2. Using multimodal AI-based prediction integrated with regional data |
|
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 19 |
Which statement best explains the key difference in how VAEs, GANs, and DDPMs generate medical images according to the figure?
|
3. DDPMs iteratively remove noise through reverse diffusion rather than using encoder–decoder or discriminator structures. |
|
DDPMs generate data by learning to reverse a noising process. The model starts with a sample from a simple distribution (eg, Gaussian noise) and iteratively denoises the sample using a learned Markov chain. At each step, the model estimates the gradient of the data distribution and refines the sample accordingly. By repeatedly applying this process, DDPMs can produce high-quality samples that closely resemble the training data. The figure depicts the forward diffusion process that gradually adds noise to the data and the reverse diffusion process that progressively denoises the sample to generate a clean output.
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 20 |
Which of the following best explains the trend shown in Figure comparing age-standardized and crude CVD mortality rates among East Asian countries?
|
1. Japan and South Korea show low age-standardized CVD mortality rates because of smaller populations. |
|
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|