I completed my PhD in operations research at the University of Toronto and now work as a Senior Applied Research Scientist in the Advanced Technology Group at ServiceNow.
My research interests broadly cover computational models that provide actionable insights for new contexts. These interests have led to me to combine predictive and prescriptive analytics to address real world problems. Specifically, I have focused my attention on computer vision and combinatorial optimization, which has helped me develop tools for personalized medicine in radiotherapy to treat complex cancers.
Objective.To establish an open framework for developing plan optimization models for knowledge-based planning (KBP).Approach.Our framework includes radiotherapy treatment data (i.e. reference plans) for 100 patients with head-and-neck cancer who were treated with intensity-modulated radiotherapy. That data also includes high-quality dose predictions from 19 KBP models that were developed by different research groups using out-of-sample data during the OpenKBP Grand Challenge. The dose predictions were input to four fluence-based dose mimicking models to form 76 unique KBP pipelines that generated 7600 plans (76 pipelines × 100 patients). The predictions and KBP-generated plans were compared to the reference plans via: the dose score, which is the average mean absolute voxel-by-voxel difference in dose; the deviation in dose-volume histogram (DVH) points; and the frequency of clinical planning criteria satisfaction. We also performed a theoretical investigation to justify our dose mimicking models.Main results.The range in rank order correlation of the dose score between predictions and their KBP pipelines was 0.50-0.62, which indicates that the quality of the predictions was generally positively correlated with the quality of the plans. Additionally, compared to the input predictions, the KBP-generated plans performed significantly better (P< 0.05; one-sided Wilcoxon test) on 18 of 23 DVH points. Similarly, each optimization model generated plans that satisfied a higher percentage of criteria than the reference plans, which satisfied 3.5% more criteria than the set of all dose predictions. Lastly, our theoretical investigation demonstrated that the dose mimicking models generated plans that are also optimal for an inverse planning model.Significance.This was the largest international effort to date for evaluating the combination of KBP prediction and optimization models. We found that the best performing models significantly outperformed the reference dose and dose predictions. In the interest of reproducibility, our data and code is freely available.
@article{Babier:2022ab,author={Babier, Aaron and Mahmood, Rafid and Zhang, Binghao and Alves, Victor G.L. and Barrag{\'a}n-Montero, Ana Maria and Beaudry, Joel and Cardenas, Carlos E. and Chang, Yankui and Chen, Zijie and Chun, Jaehee and Diaz, Kelly and David Eraso, Harold and Faustmann, Erik and Gaj, Sibaji and Gay, Skylar and Gronberg, Mary and Guo, Bingqi and He, Junjun and Heilemann, Gerd and Hira, Sanchit and Huang, Yuliang and Ji, Fuxin and Jiang, Dashan and Carlo Jimenez Giraldo, Jean and Lee, Hoyeon and Lian, Jun and Liu, Shuolin and Liu, Keng-Chi and Marrugo, Jos{\'e} and Miki, Kentaro and Nakamura, Kunio and Netherton, Tucker and Nguyen, Dan and Nourzadeh, Hamidreza and Osman, Alexander F.I. and Peng, Zhao and Dar{\'\i}o Quinto Mu{\~n}oz, Jos{\'e} and Ramsl, Christian and Joo Rhee, Dong and David Rodriguez, Juan and Shan, Hongming and Siebers, Jeffrey V and Soomro, Mumtaz H and Sun, Kay and Usuga Hoyos, Andr{\'e}s and Valderrama, Carlos and Verbeek, Rob and Wang, Enpei and Willems, Siri and Wu, Qi and Xu, Xuanang and Yang, Sen and Yuan, Lulin and Zhu, Simeng and Zimmermann, Lukas and Moore, Kevin L. and Purdie, Thomas G. and McNiven, Andrea L. and Chan, Timothy C.Y.},doi={10.1088/1361-6560/ac8044},journal={Physics in Medicine and Biology},keywords={automated planning; inverse optimization; inverse problem; knowledge-based planning; open data; optimization; radiotherapy},month=sep,number={18},pmid={36093921},pst={epublish},title={OpenKBP-Opt: an international and reproducible evaluation of 76 knowledge-based planning pipelines},url={https://iopscience.iop.org/article/10.1088/1361-6560/ac8044},volume={67},year={2022},bdsk-url-1={https://doi.org/10.1088/1361-6560/ac8044}}
Advising student-driven analytics projects: a summary of experiences and lessons learned
In this paper, we describe a course project in which teams of undergraduate students propose and execute an end-to-end analytics project to solve a real-world problem. The project challenges students to implement machine learning, optimization, simulation, or a combination of these three techniques on real-world data that they collect. A designated project advisor helps each team refine its project and assesses the quality of the resulting work. In our analysis of 58 past projects, we show that students developed solutions for a wide range of topics by employing various methodologies. However, most teams encountered similar challenges that project advisors helped them overcome with tailored feedback. Based on feedback from 106 previous students, the project experience was largely positive and helped them prepare for their future careers. We believe that this type of hands-on project is conducive to the development of important data analytics skills.
@article{Babier:2022aa,author={Babier, Aaron and Fernandes, Craig and Zhu, Ian Y.},date-modified={2023-02-06 23:24:36 -0500},doi={10.1287/ited.2022.0275},eprint={https://doi.org/10.1287/ited.2022.0275},journal={INFORMS Transactions on Education},month=aug,number={2},pages={121-135},title={Advising student-driven analytics projects: a summary of experiences and lessons learned},url={https://doi.org/10.1287/ited.2022.0275},volume={23},year={2022},bdsk-url-1={https://doi.org/10.1287/ited.2022.0275}}
