| 1 |
How does the concept of “model as a dataset” reshape traditional data-sharing practices in medical imaging?
|
It enables sharing of learned model weights instead of sensitive raw images. |
|
The model weights allow for similar characteristics as the old data but can help with patient privacy because it does not reveal the exact data of the patients. |
The model weights take in data but then generate new data but still has similar characteristics to the old one, which makes it applicable in applications. |
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 physics models are a lot more applicable because they are simulating things that would happen based on real existing equations. The statistical models give a lot of data, but require more interpretation. |
The physics models use equations to simulate what would happen in each scenario, which makes it a lot more realistic than just data alone. |
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 reduces image realism and variety by producing repetitive outputs. |
|
GANs do not broadly sample so it can reduce the diversity and overall accuracy. |
The GANs have high quality and are quite fast but are not broad, which can lead to the data being very limited. |
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 better capture clinical accuracy and diagnostic relevance. |
|
The general ones are too general and are not specifically made for evaluating health. |
Looking at medical images needs models that are specifically designed for health. The ones that are too general may not be able to interpret the medical images. |
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?
|
Higher realism may risk reproducing identifiable patient data. |
|
If models trained on data sets make images that are too similar to the ones they were trained on, it could risk leaking patient information. |
This is found under "Patient privacy and data copying". |
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. |
|
It creates a framework for evaluating new models to make sure that the models are actually accurate and can be used to help in imaging. |
It is found under the "Future directions". |
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 |
|
This helps with making sure that the populations are distributed and not underrepresented. This would allow for the data to be accurate and applicable to each population by making sure that they are properly included. |
If a population is underrepresented, then the claims made can not be applied to that population because they did not contribute a lot to the data. |
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. |
|
They can decode images by themselves. |
They go through a reverse process to decode the scans. |
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 enhances training by providing diverse, realistic datasets without ethical breaches. |
|
The models can quickly scan through a lot of data to allow doctors to make claims that are accurate and can help. A large data set can make it so the model does not produce an image that is too close to the data. |
Some models can help present raw data and make objective decisions while humans sometimes cannot make objective decisions. |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 10 |
Why is regional calibration essential when applying risk prediction models across countries?
|
To adjust for population-specific incidence and lifestyle differences |
|
Some things are limited to only small population and do not apply to the other part of the population. These can create claims that are not specific enough to be applicable to that small population and is not accurate. |
Underrepresentation in populations will make it not accurate and applicable in that population. |
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 China-PAR model will make it so that the data is collected from the population in that country so it is actually accurate towards that country. |
The Framingham models use factors that are common in Western countries but may not account for things that are important in Asian populations. |
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 mortality reflects poor access to screening. |
|
The difference between the Mortality Rates for Major CVD graphs is high. |
This makes it so that the data might not be accurate and might not have been reported. |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 13 |
What analytical limitation arises when using Western-derived coefficients in East Asian models?
|
It introduces systematic overestimation of ASCVD probability. |
|
The models in the West use factors that are in the West and may not factor in the ones that are important in the Asian population. |
This makes it so that it is not fully including everything that can help predict the CVD for Asian populations. |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 14 |
What policy implication can be derived from country-specific risk models?
|
They allow for targeted national prevention programs. |
|
Each population has different things that affect them, so having a model for each population allows programs to prevent CVD to be made. |
Models specific to an Asian population can allow prevention of CVD in those populations because they factor in the things that actually affect CVD in that population. |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 15 |
If a model excludes socioeconomic variables, what analytical consequence might occur?
|
Improved accuracy |
|
Socioeconomic variables are also important because they can affect other things such as diet, which could affect CVD. |
Models using these factors can help predict and reduce the chances of CVD. |
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 information |
|
The models can help analyze the data and factor in things that are important for CVD. |
This will help with making sure that the factors that are important towards CVD are properly used and can be applied for preventing CVD. |
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. |
|
Mongolia has high rates of CVD but South Korea has low. South Korea puts a lot of funding into doing an analysis on CVD. |
South Korea putting a lot of money into preventing CVD helps reduce the high CVD in the population. |
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 |
|
This helps make sure that data from many countries that have similar populations can help combine and make a prevention program that can be applicable for all of them. |
Populations that have similarities can be combined together and create a program that can be useful for all of them because they are similar. |
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 VAEs and DDPMs both have diversity in their images. |
The diversity is in the Image Generation Trilemma. |
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. |
|
Strokes are highest in the graph. |
It is highest for all of the countries in the graphs. |
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|