| Question 1: Why is it desirable to use outcome models in biologically based treatment planning? |
| Reference: | 1. Stewart and Li, “BGRT: Biologically guided radiation therapy–The future is fast approaching!,” Med. Phys. 34, 3739–3751 (2007).
2. AAPM TG-166 Report. |
| Choice A: | Because they are required by optimization algorithm. |
| Choice B: | Because they are simple to use. |
| Choice C: | Because they can improve dose calculation accuracy. |
| Choice D: | Because they can describe dose response. |
| Choice E: | Because they are available. |
| Question 2: A biophysical outcome model most likely includes: |
| Reference: | 1. Ling and Li, “Over the next decade the success of radiation treatment planning will be judged by the immediate biological response of tumor cells rather than by surrogate measures such as dose maximization and uniformity,” Med. Phys. 32, 2189–2192 (2005).
2. AAPM TG-166 Report |
| Choice A: | Maximum dose |
| Choice B: | Minimum dose |
| Choice C: | Mean dose |
| Choice D: | V20 |
| Choice E: | EUD, TCP and/or NTCP |
| Question 3: Why machine learning is useful for radiomics or radiogenomics modeling? |
| Reference: | El Naqa I, A Guide to Outcome Modeling in Radiotherapy and Oncology: Listening to the Data, CRC Press Taylor & Francis Group, Boca Raton, FL, 2018. |
| Choice A: | Radiomics and radiogenomics are algorithms based on machine learning |
| Choice B: | Statistical methods can only be used for testing but not for building such models |
| Choice C: | Machine learning allow for discovering complex patterns in the data such as encountered in radiomics and radiogenomics |
| Choice D: | None of the above |
| Question 4: Radiomic models and signatures embedded or confounded by underlying features are a concern when drawing conclusions about features and disease relationships. Assess the most appropriate statement with respect to safeguarding predictive power of these models: |
| Reference: | M Welch et al. Vulnerabilities of radiomic signature development: The need for safeguards. Radiotherapy and Oncology, 130 (2019): 2-9;
Ke Nie et al (2019) NCTN Assessment on Current Applications of Radiomics in Oncology. Int J Radiation Oncol Biol Phys, Vol. 104, No. 2, pp. 302e315, 2019
A Zwanenburg, S Leger, M Vallières, S Löck. Image biomarker standardisation initiative. arXiv:1612.07003 (2016-2019) |
| Choice A: | open software packages and standardized implementations (e.g., IBSI) for texture features should be used to ensure reproducibility |
| Choice B: | models and features should be tested to determine added prognostic and predictive accuracy compared to accepted clinical factors |
| Choice C: | features should be tested for underlying dependencies using statistical analysis or by perturbing the data in controlled ways |
| Choice D: | image quality (e.g. artifacts) should be assessed in a preprocessing step and contouring information included |
| Choice E: | all of the above are possible safeguards |
| Question 5: Visualization tools can be used to overcome the obstacle of interpretability in feature analysis using machine learning. Select the single true statement: In deep learning using convolutional neural networks (CNN): |
| Reference: | Diamant et al (2019) Deep learning in head & neck cancer outcome prediction. Sci. Rep. (2019) 9:2764 https://doi.org/10.1038/s41598-019-39206-1 |
| Choice A: | hand-crafted radiomic features are used in a regression model to predict outcomes |
| Choice B: | texture features have to be manually extracted and checked for dependencies before they can be used for prediction |
| Choice C: | amongst others, gradient class activation maps (Grad-CAM) are a possible tool to depict what areas of the image the CNN found most relevant for outcome prediction |
| Choice D: | we have to expect that its performance will not rival that of traditional radiomics using hand-crafted features |
| Choice E: | none of the above is true |
| Question 6: Outcome modeling is the process to characterize the behavior of tissue response to a treatment. In radiotherapy, it plays an important role in: |
| Reference: | El Naqa I. “A guide to outcome modeling in radiotherapy and oncology – Listening to the data”. CRC Press Taylor & Francis’s Group, Boca Raton, FL 2018. |
| Choice A: | Understanding response to different therapeutic cancer agents |
| Choice B: | Personalization of treatment |
| Choice C: | Designing of new clinical trials |
| Choice D: | All of above. |
