2019 AAPM Annual Meeting
Back to session list

Session Title: The integration of AI and Machine Learning in Medical Physics Applications
Question 1: Historically, what types of architecture excels at modeling sequential information?
Reference:Liu, Pengfei, Xipeng Qiu, and Xuanjing Huang. "Recurrent neural network for text classification with multi-task learning." arXiv preprint arXiv:1605.05101 (2016).
Choice A:Recurrent Neural Networks
Choice B:Convolutional Neural Networks
Choice C:Fully Connected Networks
Choice D:Relational Networks
Question 2: Which type of architecture excels at pixel wise classification?
Reference:Ronneberger, Olaf, Philipp Fischer, and Thomas Brox. "U-net: Convolutional networks for biomedical image segmentation." International Conference on Medical image computing and computer-assisted intervention. Springer, Cham, 2015.
Choice A:ResNet
Choice B:DenseNet
Choice C:VGG
Choice D:Encoder-Decoder Networks
Question 3: Which is not a part in the process of patient-specific QA using machine learning?
Reference:G Valdes, R Scheuermann, CY Hung, A Olszanski, M Bellerive, TD Solberg, “A mathematical framework for virtual IMRT QA using machine learning”, Medical Physics, 2016; 43(7):4323-4334
Choice A:Extracting features using scripts to read IMRT beams from TPS database
Choice B:Building models by calculating all plan complexity metrics affecting passing rate
Choice C:Collecting IMRT QA data for building and re-training virtual QA models
Choice D:Measuring treatment machine outputs
Question 4: Which of the following statement on Virtual IMRT QA is not true?
Reference:G Valdes, MF Chan, SB Lim, R Scheuermann, JO Deasy, and TD Solberg, “IMRT QA using machine learning: A multi-institutional validation”, Journal of Applied Clinical Medical Physics, 2017; 18:5:279-284
Choice A:Different TPS dose calculation model needs a different virtual IMRT QA model
Choice B:Virtual IMRT QA is disease-site dependent
Choice C:Virtual IMRT QA can accurately predict passing rate for any plan
Choice D:Virtual IMRT QA model needs to be refined for different detectors & energies
Question 5: Multi-atlas deformable image registration algorithms have been shown to be very effective (i.e. clinically acceptable for the majority of patients) for delineating:
Reference:McCarroll et al, Retrospective validation and clinical implementation of automated contouring of organs at risk in the head and neck: A step toward automated radiation treatment planning for low- and middle-income countries, Journal of Global Oncology,4, 1-11, 2018
Choice A:Anatomical structures (e.g. parotid)
Choice B:High-risk CTVs
Choice C:Microscopic spread of disease
Choice D:GTVs
Question 6: Deep learning techniques have been shown to be capable of delineating high-risk CTVs in head/neck patients, with the mean agreement between the automatic contours and manual contours of:
Reference:Cardenas et al, Deep learning algorithm for auto-delineation of high-risk oropharyngeal clinical target volumes with built-in dice similarity coefficient parameter optimization function, IJROBP 101(2), 468-478, 2018
Choice A:<1 mm
Choice B:2-4 mm
Choice C:5-15 mm
Choice D:6-20 mm
Back to session list