| 1 |
What is the primary goal of the article according to its introduction?
|
To explore advancements, applications, and challenges of generative AI in medical imaging |
|
The introduction clearly stae the purpose of the article, it also explains that the article is focused on giving a full overview of synthetic data and explaining how generative AI would chngae the medical imaging. The article also explains what it can be used for and what challenges come along with it. Since the beginning of the article, they clearly explain that they want to look at the progress in this field and connect it to real-world applications. They also talk about the ethical and technical concerns that shouldn't be ignored. The other options such as hospital management, economic impacts, policy comparisons, or designing new models aren't really presented at the main purpose of the article in the introduction.
|
Based on the article, the foundation for creating synthetic medical images is key generative models like VAEs, GANs, and DDPMs. The deifition of synthetic data was purposed by The Royal Society and The Alan Turing Institute in 2022, which explain what syntehtic data truly means.. Throughout the paper, the authors supoort their claim by using specific resources such as Chambon et al. (2022), Pinaya et al. (2022), and Khosravi et al. (2024), which show how synthetic data can improve model performance.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 2 |
How do generative AI models differ from traditional discriminative models in healthcare applications?
|
Generative models produce new data rather than only classify or interpret |
|
Generative models are designed to create new data samples rather than use to classsify or interpret inputs that is already existed. In medical imaging, they can generate synthetic MRI, CT, or X-ray images that resemble the real data of the patient. This distinguishes them from discriminative models since it only predict labels the outcome from given inputs.
|
In machine learning, discriminative models learn probabilities such as
P(y/x) while genertaive models learn data distributions like P(x) or joint distributions P(x,y)- these learning enables sampling and data synthesis.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 3 |
What is meant by the term “model as a dataset”?
|
Sharing trained model weights instead of raw data |
|
The term “Model as a dataset” means that insread of sharing raw medical data it share trained model weights. The weights encode statistic patterns from the data which can be used to generate synthetic samples without exposing patient infomation
|
This idea is based on representation learning. Neural network's parameters will act like compressed summaries of the data it has learned from which allows people to work together without sharing their raw data. This help protect privacy and federated and distributed learning which is training a model on multiple devices or places instead of one location only.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 4 |
Which statement correctly distinguishes physics-informed and statistical models?
|
Physics-informed models incorporate biological or physical principles |
|
Physicsinformed models use a biological or physical laws in their modeling process. These models are able to integrate equations to predicts something, not like others model that only learn patterns from the data.
|
They combine scientific equations with data, they rely on patterns learned from data rather than rely fully on statistic like those statistical models.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 5 |
According to the article, what does the “image generation trilemma” describe?
|
Trade-offs among image diversity, quality, and speed |
|
The image generation trilemma describe the trade-off between image quality, diversity, and sampling speed. This mean improving one aspect will often compromises another depend on the model architecture.
|
Improving anyone of these factors usually reduces the others due to the model design and computational limits.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 6 |
What is the Human Turing Test used for in medical image synthesis?
|
To assess realism of synthetic medical images by experts |
|
The Human Turing Test is use to evaluates whether a clinical experts can actually distinguish the different between synthetic medical images and the real images.
|
In medicine, human judgement is crucial as realism will affects the diagnostic trust.
|
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?
|
Eliminating all medical biases permanently |
|
Synthetic data does not automatically remove bias and may even repeat existing biases.
|
Synthetic data does not automatically remove bias and may even repeat existing biases.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 8 |
What is one major ethical concern associated with generative AI in medical imaging?
|
Data copying and patient reidentification |
|
A key ethical concern is that models might copy real patient data and risk reidentification.
|
Neural networks can memorize data, which may lead to privacy risks.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 9 |
What regulatory precedent did the article cite for synthetic data technologies?
|
FDA clearance of synthetic MRI as image-processing software |
|
FDA clearance of synthetic MRI shows that synthetic imaging tools can be officially approved.
|
Medical AI must meet safety and performance standards before clinical use.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 10 |
What is the main purpose of the article?
|
To introduce new diagnostic imaging technologies |
|
The article mainly discusses how generative AI is used in medical imaging.
|
It reviews current technology and explains its benefits and challenges in healthcare.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 11 |
Which of the following models was originally developed for a Western population?
|
Framingham Risk Score |
|
The Framingham Risk Score was developed using data from a Western population.
|
Risk models work best when applied to populations similar to the ones they were built from.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 12 |
Why might Western-based risk prediction models overestimate ASCVD risk in East Asian populations?
|
East Asians have lower baseline incidence of ASCVD |
|
Western models may overestimate risk in East Asians because their baseline disease rates are lower.
|
Risk prediction depends on population event rates, so differences affect accuracy.
|
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?
|
It was calibrated using national data representing diverse regions in China |
|
China-PAR was developed by using national Chinese data, which make it more accurate for China.
|
Using local data improves model calibration and reliability.
|
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?
|
Genetic ancestry markers |
|
Genetic ancestry markers are not usually included in traditional risk models.
|
Genetic ancestry markers are not usually included in traditional risk models.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 15 |
What is a major difference between the Suita Score and the Framingham Risk Score?
|
Suita Score was designed for a Japanese population using local epidemiological data |
|
The Suita Score was created specifically for the Japanese population.
|
Population-specific models not only reflect local health patterns but also improve prediction accuracy.
|
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?
|
They improve accuracy and reduce overestimation of risk |
|
East Asia–specific models reduce risk overestimation and also improve the accuracy.
|
Matching the model to local disease rates improves calibration.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 17 |
Which factor was highlighted as influencing ASCVD risk differences among East Asian countries?
|
Cultural and dietary variations, such as salt intake and lifestyle |
|
Diet and lifestyle differences, such as salt intake, affect ASCVD risk across East Asia.
|
Environmental and behavioral factors influence cardiovascular disease rates.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 18 |
What future direction does the article suggest for improving ASCVD risk prediction?
|
Using multimodal AI-based prediction integrated with regional data |
|
in the future, risk prediction may use AI that combines different types of health data.
|
Precision medicine aims to improve prediction by integrating multiple data sources.
|
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?
|
DDPMs iteratively remove noise through reverse diffusion rather than using encoder–decoder or discriminator structures. |
|
DDPMs create images by slowly removing noise, unlike VAEs or GANs.
|
Diffusion models use a reverse noise process, which is different from encoder–decoder or adversarial methods.
|
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?
|
Despite differences in age structures, Japan maintains low mortality rates in both measures, suggesting effective prevention and healthcare systems. |
|
Japan’s low mortality rates suggest a strong prevention and healthcare systems.
|
Age-standardized rates shows a comparisons between countries by adjusting for age differences.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|