| Question 1: Which of the following is a generative model? |
| Reference: | Foster, David. Generative deep learning: teaching machines to paint, write, compose, and play. O'Reilly Media, 2019.
Tomczak, Jakub M. Deep Generative Modeling. Springer, Cham, 2022. 1-12. |
| Choice A: | A model trained to map a set of noisy imaging measurements of an object to a useful estimate of the object. |
| Choice B: | A model trained to map measurement noise to an estimate of an object. |
| Choice C: | A model trained to map independent and identically distributed (i.i.d.) samples from a tractable probability distribution to i.i.d. samples approximately from an unknown, high-dimensional data distribution. |
| Choice D: | None of the above |
| Question 2: A medical image GAN trained using conventional methods is guaranteed to learn the statistics of the medical images that are relevant to a downstream clinical application. |
| Reference: | Kelkar, Varun A., et al. "Assessing the ability of generative adversarial networks to learn canonical medical image statistics." arXiv preprint arXiv:2204.12007 (2022). |
| Choice A: | True |
| Choice B: | False |
| Question 3: Which of the following statements is accurate? |
| Reference: | Kelkar, Varun A., et al. "Assessing the ability of generative adversarial networks to learn canonical medical image statistics." arXiv preprint arXiv:2204.12007 (2022). |
| Choice A: | State-of-the-art GANs produce visually realistic images. Therefore, one can use them in medical imaging applications. |
| Choice B: | If a GAN can reproduce the first- and second-order statistics of the data it is trained on, it can be used for medical imaging applications. |
| Choice C: | GANs may suffer from mode dropping and mode collapse. |
| Choice D: | None of the above |
| Question 4: Which of the following statement is correct regarding spatial resolution of deep-learning-based noise reduction (DLR) in CT? |
| Reference: | Solomon J, et al. Noise and spatial resolution properties of a commercially available deep learning-based CT reconstruction algorithm. Med Phys. 2020;47(9):3961-71. Epub 2020/06/09. doi: 10.1002/mp.14319. PubMed PMID: 32506661. |
| Choice A: | It can reduce noise while maintaining the spatial resolution for any structures |
| Choice B: | It improves the in-plane spatial resolution |
| Choice C: | It may degrade the spatial resolution |
| Choice D: | It improves the cross-plane spatial resolution |
| Question 5: Which one of the following parameters might affect the generalizability of a DLR method in CT? |
| Reference: | Huber NR, et al. Random Search as a Neural Network Optimization Strategy for Convolutional-Neural-Network (CNN)-based Noise Reduction in CT. Proc SPIE Int Soc Opt Eng. 2021;11596. Epub 2021/02/01. doi: 10.1117/12.2582143. PubMed PMID: 35386837; PMCID: PMC8982987. |
| Choice A: | Reconstruction kernel |
| Choice B: | Radiation dose |
| Choice C: | All of the above |
| Question 6: Deep-learning has been applied to improve quantitative MRI in assessing human tissue microstructures? |
| Reference: | Li Feng, Dan Ma, Fang Liu, “Rapid MR relaxometry using deep learning: An overview of current techniques and emerging trends”. NMR in Biomedicine.2020;e4416. https://doi.org/10.1002/nbm.4416 |
| Choice A: | True |
| Choice B: | False |
