| 1 |
How does the concept of “model as a dataset” reshape traditional data-sharing practices in medical imaging?
|
It promotes open patient data repositories. |
|
The model stores compressed representations of key features and relationships learned from the training dataset, making data sharing more efficient and potentially more privacy-preserving. |
Section “Synthetic datasets – Generative models”, paragraph beginning “The advancement of generative artificial intelligence introduces a new concept in data sharing, which we refer to as a model as a dataset.” |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 2 |
Which analytical conclusion can be drawn about the trade-offs between physics-informed and statistical models?
|
Physics-informed models are more interpretable but computationally intensive. |
|
The article shows that physics-informed models explicitly incorporate scientific knowledge, while statistical models learn patterns directly from data. Therefore, the choice depends on whether interpretability or scalability is the priority. |
Section “Synthetic datasets – Generative models”, paragraphs beginning “Physics-informed models are primarily rule-based approaches…” and “In contrast to physics-informed models, statistical models learn from data patterns…” |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 3 |
Why is “mode collapse” considered a critical problem in GAN-based medical image synthesis?
|
It improves the uniformity of generated samples. |
|
Medical imaging datasets require broad coverage of disease presentations. If diversity is lost, synthetic data may not adequately represent real-world clinical populations. |
Section “Synthetic datasets – Generative models”, paragraph discussing the AI generation trilemma, particularly the sentence “GANs excel at generating high-quality samples but might not always capture all data variations, leading to low mode coverage, known as mode collapse.” Also referenced in “Use cases in medical imaging”, where repeated training on generated outputs may worsen mode collapse. |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 4 |
Why are healthcare-specific metrics preferred over general-purpose metrics such as FID or SSIM?
|
They depend on ImageNet-trained features. |
|
Metrics like FID or SSIM mainly measure visual similarity, whereas medical applications require evaluation of diagnostic accuracy and clinical relevance. |
Section “Health-care-specific metrics”, paragraphs beginning “Evaluating synthetic medical images requires metrics tailored to health-care needs…” and “Similarly, anatomical accuracy is being prioritised…” |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 5 |
What does the article identify as the key tension between privacy preservation and image fidelity?
|
Fidelity metrics can guarantee anonymization. |
|
As synthetic images become closer to real patient images, the likelihood of data copying, re-identification, or privacy leakage may increase. |
Section “Challenges and considerations – Patient privacy and data copying”, paragraphs beginning “Although synthetic datasets can help to preserve patient privacy…” and “Unlike tabular data, medical images contain patient-identifying information embedded within the pixel values…” |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 6 |
Why is the FDA’s approval of synthetic MRI technology significant for future AI-generated data?
|
It establishes a framework for validating synthetic data equivalence in clinical use. |
|
FDA approval is significant because it demonstrates regulatory acceptance of AI-generated medical imaging technologies and validates their potential clinical utility, paving the way for broader adoption of synthetic data applications.
The article highlights synthetic MRI and image transformation techniques as practical clinical uses of generative AI. Regulatory approval helps establish trust and supports future deployment. |
Section “Use cases in medical imaging”, paragraph discussing image transformations and accelerated MRI, together with the “Future directions” section stating that regulatory bodies such as the FDA will play a crucial role in establishing standards for responsible use. |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 7 |
Which strategy would best mitigate demographic bias in generative models according to the article?
|
Applying diversity-aware training and fairness constraints |
|
The article emphasizes that bias often originates from imbalanced source datasets. Improving subgroup representation and evaluating fairness can reduce disparities in model performance. |
Section “Potential biases”, paragraph beginning “The use of synthetic datasets and generative models raises important bias considerations…” especially the discussion of “diversity-aware sampling,” “adversarial debiasing techniques,” and the recommendation that researchers “regularly audit the generated data for fairness and representativeness.” |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 8 |
How do DDPMs exemplify versatility in healthcare image synthesis?
|
They can perform multiple tasks such as denoising, inpainting, and anomaly detection without retraining. |
|
The article highlights that DDPMs are not limited to creating new images; they can also modify, reconstruct, and enhance existing medical images, making them useful across a wide range of clinical and research applications. |
Section “Use cases in medical imaging”, paragraphs beginning “Generative models also excel at image transformations…” and “DDPMs have enabled inpainting…” |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 9 |
What analytical insight does the article provide about integrating AI-generated medical images into education and research?
|
It eliminates the need for patient participation in studies. |
|
Synthetic images allow learners and researchers to study conditions that may be difficult to collect in sufficient numbers from real-world datasets, thereby enhancing training and scientific investigation. |
Section “Potentials and promises” and “Increased dataset size and diversity”, especially the discussion on improving dataset diversity, rare disease representation, and medical education. |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 10 |
Why is regional calibration essential when applying risk prediction models across countries?
|
To ensure identical risk cutoffs |
|
The article notes that even among East Asian countries, cardiovascular disease patterns vary considerably, making direct application of foreign models less accurate. |
Article “Atherosclerotic Cardiovascular Disease Risk Prediction Models in China, Japan, and Korea”, section discussing current risk calculators and the paragraph beginning “Even among countries classified with similar risk levels…” |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 11 |
What analytical conclusion can be drawn when comparing the China-PAR and Framingham models?
|
China-PAR uses local epidemiological data, leading to improved predictive validity. |
|
The Framingham model was developed using Western cohorts and tended to overestimate coronary heart disease risk in Chinese populations. Recalibration improved performance, but locally derived models better reflect regional epidemiology and risk profiles. |
Section “ASCVD Risk Prediction in China”, paragraphs discussing the CMCS cohort and the overestimation of risk by the Framingham Risk Score. |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 12 |
Based on CVD mortality data, what analytical inference can be made about Japan’s position compared to neighboring countries?
|
Japan’s low CVD mortality suggests effective prevention and healthcare systems. |
|
The article reports that Japan had one of the lowest crude and age-standardized cardiovascular mortality rates among the East Asian countries analyzed. |
Figure 1 (Age-Standardized and Crude Mortality Rates for Major CVD) and the section “Epidemiology of ASCVD in East Asian Populations Living in Asia and in the United States.” |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 13 |
What analytical limitation arises when using Western-derived coefficients in East Asian models?
|
It eliminates the need for validation. |
|
Differences in genetics, lifestyle, disease prevalence, and baseline event rates mean that coefficients developed from Western cohorts may not accurately represent East Asian populations, often leading to risk overestimation or underestimation. |
Section “Current State of ASCVD Risk Calculators for East Asian Populations” and “ASCVD Risk Prediction in China,” particularly the discussion of Framingham-based equations overestimating risk and the need for recalibration using local data. |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 14 |
What policy implication can be derived from country-specific risk models?
|
They increase healthcare inequality. |
|
Country-specific models better reflect local epidemiology, risk factor distributions, and disease patterns, leading to more accurate risk assessment and treatment decisions. |
Section “Current State of ASCVD Risk Calculators for East Asian Populations” and “ASCVD Risk Prediction in China,” where the authors emphasize developing and validating models using local population data. |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 15 |
If a model excludes socioeconomic variables, what analytical consequence might occur?
|
Ignored non-biological determinants of disease |
|
The article notes that factors such as acculturation, immigration history, and environmental influences affect ASCVD risk profiles. Ignoring these factors can reduce predictive accuracy and overlook population differences. |
Section “The Impact of Acculturation and Environmental Effects of ASCVD Risk Profiles,” particularly the discussion of environmental, cultural, and immigration-related influences on cardiovascular risk. |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 16 |
How might AI improve next-generation ASCVD risk prediction in East Asia?
|
By integrating multimodal data, including imaging and lifestyle informa |
|
The article suggests future models should incorporate more personalized and population-specific information. AI methods can process multidimensional data and support precision medicine approaches. |
Section “Future Directions,” especially the discussion of novel architectures, advanced analytical methods, and personalized datasets for precision medicine. |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 17 |
What conclusion can be drawn from comparing Mongolia’s and South Korea’s CVD mortality rates?
|
Mortality differences reflect varying effectiveness of national prevention programs.Mortality differences reflect varying effectiveness of national prevention programs. |
|
The mortality figures show Mongolia has the highest CVD mortality rate among the countries compared, while South Korea has one of the lowest rates. |
Figure 1: Age-Standardized and Crude Mortality Rates for Major CVD and the section “Epidemiology of ASCVD in East Asian Populations Living in Asia and in the United States.” |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 18 |
What is the most logical future direction for improving ASCVD models across East Asia?
|
Establishing multinational data-sharing platforms to harmonize regional models |
|
The article repeatedly highlights the limitations of applying generalized models across diverse populations and emphasizes the need for more representative, disaggregated, and locally validated data. |
Section “Future Directions” and the article’s conclusion, which call for greater population-specific research, validation studies, and refinement of risk prediction tools for East Asian populations. |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 19 |
According to the “image generation trilemma” shown in the figure, what analytical conclusion can be drawn about the relative strengths of VAEs, GANs, and DDPMs in medical image synthesis?
|
VAEs and DDPMs perform identically in generating high-fidelity images. |
|
The “image generation trilemma” illustrates that each model occupies a different position within the quality–diversity–speed trade-off. Therefore, model selection should depend on the intended medical application rather than assuming one model is universally superior. |
Figure 2: Image Generation Trilemma (caption below the triangle). The caption states that “VAEs excel in generating diverse samples quickly but can compromise on image quality; GANs strike a balance between good quality and diversity but can suffer from mode collapse; DDPMs prioritise high-quality and diverse samples at the cost of a slow generation speed.” |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 20 |
Based on Figure, what analytical conclusion can be drawn regarding the distribution of cardiovascular disease (CVD) subtypes across East Asian countries?
|
Stroke dominates as the primary cause of CVD death in all East Asian countries equally. |
|
China shows approximately 48% of CVD deaths from stroke versus 41% from IHD, and South Korea shows 47% from stroke versus 36% from IHD. In contrast, Japan has a more balanced distribution, with 39% from stroke and 38% from IHD. The chart also shows a higher proportion of ischemic stroke among total stroke deaths in Japan (63%) compared with China (50%) and South Korea (61%). |
Figure 2: Proportion of Subtypes of CVD in Total CVD Death, pie charts for China, Japan, South Korea, and East Asia. The percentages shown in each pie chart support the comparison of stroke, ischemic heart disease, and other CVD causes across countries. |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|