Biological uncertainties regarding specific radiosensitivity of each patient constitutes a large part of the unknown patient-specific parameters. Therefore, the treatment plan for each patient should be carefully devised to incorporate such uncertainties. In these projects, I used robust and convex optimization methods to directly incorporate the tissue-specific uncertainties in the so-called “alpha-beta ratio” into the treatment plan optimization problem and obtained robustly-optimized plans with respect to the uncertainty intervals.
Ajdari, A, Ghate, A. (2016). Robust spatiotemporally integrated fractionation in radiotherapy. Operations Research Letter. 44(4): 544-549.
Ajdari, A, Ghate, A. (2016). A model predictive control approach for discovering nonstationary fluence-maps in radiotherapy, Winter Simulation Conference, Washington D.C. 2065-2075.
Adaptive optimization of treatment plans
In order to move towards a more personalized treatment, one needs to incorporate patient-specific biological information. In these works, I put forth the theoretical framework for incorporating such information into the planning problem and to obtain biologically-adapted personalized treatment plans. These formulations take as inputs the original plans, and updates these plans based on the “observations” of the patients’ biological parameter. In this way, patients’ treatment plans are truly tailored to each patient’s specific needs, and the way the tumor dynamically and stochastically responds to the treatment.
Ajdari, A, Saberian, F, Ghate, A. (2019). A theoretical framework for learning tumor dose-response uncertainty in individualized spatiobiologically integrated radiotherapy, INFORMS Journal on Computing (in press).
Ajdari, A, Ghate, A, Kim, M. (2018). Adaptive treatment-length optimization in spatiobiologically integrated radiotherapy, Physics in Medicine & Biology 63(7):075009.
Optimal stopping in radiation therapy
Whereas the previous works developed mostly theoretical foundations for addressing biological uncertainties and biomarker-based treatment adaptation, during my postdoctoral studies at MGH, I have been focusing on real-world implementations of these theoretical methods to devise a practical clinically-viable framework for personalized radiotherapy. Towards this end, three main areas were identified as needing further investigations: (1) lack of viable mid-treatment biomarker of treatment response, (2) lack of transparent predictive model of treatment outcome, (3) lack of a systemic framework for treatment adaptation that seamlessly combines mathematics, biology, and oncology. To address these gaps, several projects have been undertaken. The search for validated mid-treatment biomarkers are the focus of the first paper, the second paper proposes a novel Bayesian framework for transparent modeling of the treatment response, and the last paper proposes a theoretical foundation for a systemic approach to treatment monitoring and adaptation.
Ajdari, A, Shusharina, N, Liao, Z, Mohan, R, Bortfeld, T (2019). Mid-Treatment [18]F-FDG PET Uptakes Can Predict Symptomatic Radiation Pneumonitis in Non-Small Cell Lung Cancer Patients. Internatoinal Journal of Radiation Oncology.Biology.Phyisics. 105 (1): S224.
Ajdari, A, Shusharina N, Liao, Z, Mohan, R, Bortfeld, T (2019). A novel machine learning-Bayesian network model for prediction of radiation pneumonitis: Importance of mid-treatment information. International Conference on the Use of Computers in Radiation Therapy. Montreal, Canada, June 17-21, 2019.
Ajdari, A, Niyazi, M., Nicolay, N et al (2019). Towards optimal stopping in radiation therapy. Radiation Therapy and Oncology, May 2019, vol. 134, 96–100.
Simulation & analysis of emergency department
Due to their chaotic nature, the care process of patients in the emergency departments are subject to a host of uncertainties, which could substantially delay the care process and affect treatment outcome. In these works, I used data analytics to first analyze the effect of “wasted” times on the outcome of treatment, showing that every 10 minutes increase in non-value added times reduce the likelihood of favorable outcome by 9%. Subsequently, I develop data-driven simulation-based solutions for optimizing the care process of TBI patients.
Ajdari, A, Boyle, L.N., Kannan, N et al. (2017). Simulation of the Emergency Department Care Process for Pediatric Traumatic Brain Injury. Journal for Healthcare Quality 40(2):110-118.
Ajdari, A, Boyle, L.N., Kannan, N et al. (2017). Examining Emergency Department Treatment Processes in Severe Pediatric Traumatic Brain Injury. Journal for Healthcare Quality 39(6):334-344.
ONGOING PROJECTS
My current research focuses on analyzing various imaging and blood-based biomarkers of radiotherapy treatment outcome and developing novel data-driven approaches to biologically adapt the treatment plan for each individual patient by integrating Optimization, Machine learning and Bayesian statistics.
A hybrid Random Forest-Bayesian Network predictive model of chemo-radiation toxicity in non-small cell lung cancer (with T. Bortfeld and D. Craft).
Adjustable robust optimization (ARO) for treatment length optimization of radiotherapy plans (with S. Eikelder, D. Hertog, and T. Bortfeld).
A distributed Bayesian Network model for multi-outcome modeling of liver metastasis radiotherapy (with Y. Xie, C. Richter, T. Hong, D. Duda, and T. Bortfeld).
Adaptive treatment planning using mid-treatment [18]-FLT PET imaging in head-and-neck cancer (with S. Eikelder, D. Hertog, R. Jeraj, P. Ferjancic, and T. Bortfeld).
Expert-assisted natural language programming (NLP) for building a biomarker network for liver cancer treatment (with M. Walczak, C. Collicott, K-H. Kuefer, and T. Bortfeld).