| 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 |
|
The study is aimed at providing a detailed overview of image synthesis in the medical imaging field, exploring its gradual development, uses, and problems it faces in the current world.
|
The main purpose of the study is directly stated to be as such in the abstract. Synthesized images in the context of the study are clearly defined within the article as artificially generated datasets that closely mirror and take reference from actual non-generated data. This means that image synthesis within the study is basically a form of Generative AI involvement in medical imaging.
|
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 |
|
Generative AI models have the capability to produce new synthetic data, diverging from traditional discriminative models, which do not show the same function and are solely able to interpret data and make decisions based on the data they do.
|
In the article, generative artificial intelligence is referred to as being set apart from traditional discriminative models by its potential in generating synthetic data, as opposed to discriminative models merely being able to classify and analyze directly given data.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 3 |
What is meant by the term “model as a dataset”?
|
3. Sharing trained model weights instead of raw data |
|
The "model as a dataset" concept allows data stored in internal parameters, or weights, to enable others to generate synthetic images based on the characteristics of the initial data.
|
The patterns and notable traits in the training data are stored as CONDENSED versions of themselves within weights when the "model as a dataset" concept is used.
|
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 mainly apply previously established laws and expert-provided knowledge within physics to create accurate representations of phenomena in biology, whereas statistical models rely on input data and distribution types to create samples.
|
Physics-informed models involve the use of specific physics principles and field-related knowledge (in this case, biology) for information in data generation. In contrast, statistical models approach data generation through the usage of input data and distributions in different ways, such as VAEs, GANs, and DDPMs.
|
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 |
|
To acquire two of the three aspects (stated as diversity, quality, and speed) in the generated samples, one aspect must be underprioritized in favor of the other two.
|
AS STATED IN THE ARTICLE:
If a model can generate samples across a diverse range at high speeds (e.g., VAEs), the quality of the samples generated will likely be compromised.
If a model can generate samples with good quality and speed (e.g., GANs), it may be susceptible to mode collapse, meaning limited variation coverage, leading to reduced diversity.
If a model can generate diverse samples with high quality (e.g., DDPMs), the samples will take a longer time to be generated.
|
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 Turing Test is primarily used to allow for crucial insights that allow for the adjustment and refinement of quality.
|
In the Turing Test, experts distinguish real images from synthetic ones, allowing for feedback on improving the quality of image synthesis processes.
|
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?
|
4. Eliminating all medical biases permanently |
|
Synthetic data may be susceptible to medical bias concerns, leading to low representation in specific groups and diminished accuracy.
|
Training data used for model training may be biased towards specific demographics and minoritized groups, leading to the finished product being influenced by said biases in generating images.
|
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 |
|
When the model can accurately replicate the original data, sensitive information concerning patients may be at risk of being exposed.
|
Medical image pixels have information that allows patients to be reidentified, leading to difficulty in preserving patient anonymization.
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|
| 9 |
What regulatory precedent did the article cite for synthetic data technologies?
|
2. FDA clearance of synthetic MRI as image-processing software |
|
FDA's clearance of synthetic MRI is stated as an example of a framework for synthetic medical imaging evaluation.
|
The FDA needs good evidence and clinical validation to reliably confirm that the results in the use of synthetic images and normal traditional images by radiologists remain consistent in MRI 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 |
|
The study is aimed at comparing
|
|
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 |
|
The Framingham Risk Score was intended to be a model developed for general Western populations.
|
The Framingham Risk Score was developed and used data from a limited population within Framingham, a city in Massachusetts, and mainly reflects Western groups due to being based on Caucasians.
|
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 |
|
|
|
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?
|
4. It was calibrated using national data representing diverse regions in China |
|
The China-PAR model was specifically made for China, having higher calibration as a result, and directly representing groups in China.
|
The China-PAR model uses data directly from China cohort studies in development.
|
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 |
|
Genetic ancestry markers are mostly not part of the core equation for ASCVD risk prediction models.
|
Genetic ancestry markers are typically not core components of ASCVD risk prediction, and are generally viewed as supplementary information.
|
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 |
|
While the Suita Score was designed specifically for the Japanese populace, the Framingham Risk Score was made for general Western populations.
|
The Suita Score uses data and was validated directly by Japanese people, whereas the Framingham Risk Score uses data based on Americans.
|
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 |
|
East Asia-specific risk models are more specific in their purposes, which means that there will be more accuracy in predicting risk.
|
East Asia-specific risk models take data and are validated directly based on data from East Asian populations, meaning that they are more suitable for the purpose of estimating ASCVD risks in East Asia.
|
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 |
|
East Asian lifestyles led to different predicted rates of ASCVD risk than opposed to predicted rates of ASCVD risk by Westerners.
|
Certain different aspects of East Asian lifestyles as opposed to Western lifestyles were shown to affect the final predicted risks.
|
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. |
|
The model for DDPMs shows the use of a reverse diffusion process to gradually denoise the sample in order for a cleaner output result.
|
In generating data, DDPMs use a learned Markov chain to iteratively denoise a sample, refining the sample according to data distribution gradient estimations.
|
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?
|
2. Mongolia and North Korea demonstrate higher CVD mortality due to older population structures alone. |
|
|
|
7 |
-.50
-.25
+.25
เต็ม
0
-35%
+30%
+35%
|