Projects | Dionne M. Aleman

Prediction of COVID-19 severity

Fri, 05 Nov 2021 13:59:33 +0000

Early and effective detection of severe infection cases during a pandemic can significantly help patient prognosis and resource allocation. However, data may be slow to become available, and questions about the applicability of global datasets to local communities abound. Will global trends apply to a particular community with a particular population demographic profile, e.g., age, comorbidity prevalence, socio-economic status, etc.?

To tackle this issue, we develop machine learning that tools that (1) can accurately predict COVID-19 severity at the time of RT-PCR testing, before any blood tests, imaging, or presentation at hospitals; and (2) can be applied to small datasets reflective of a local population to best train predictions for a specific healthcare unit’s population. We frame the prediction problem as both an imbalanced classification, as severe cases are far less likely than mild cases, and as an anomaly detection problem. With a small dataset with few features, our tools can accurately predict hospitalization, ICU, and death, and can be easily re-trained for other emerging pandemics.

Radiotherapy treatment planning

Fri, 05 Nov 2021 13:57:21 +0000

Elekta’s Leksell Gamma Knife® (stereotactic radiosurgery)

The introduction of Elekta’s Leksell Gamma Knife® PERFEXION™ has taken the field of stereotactic radiosurgery to a new level. Unlimited access to cranial volume and full automation approach open a new research area for the researchers who have been already working on the previous Lekshell Gamma Knife unit for years. Although this method is primarily designed to be performed on brain tumors, today, it is considered an effective treatment for several other conditions, including arteriovenous malformations and pituitary tumors.

The new capabilities of the PERFEXION™ unit require new methods of treatment plan design. In previous Gamma Knife® designs, the collimator (the component that adjusts the shape/location of the radiation delivery) had to be manually adjusted, rendering complex treatments too labor intensive for clinical viability. Because the movement of the collimator in the new PERFEXION™ unit is now automated, complex treatment plans that can very tightly conform to the targeted treatment area can be delivered in clinical settings.

In order to deliver a high quality treatment, we select collimator positions based on optimization methods. The optimization algorithms are designed to deliver an appropriate amount of dose to the target area while simultaneously avoiding sensitive healthy tissues, thereby leaving the patient with a high quality of life and fewer side effects after treatment.

Total body irradiation (TBI)

Diseases requiring bone marrow transplants, also called stem cell transplants, include leukemia, lymphoma, sickle cell disease and aplastic anemia, as well as some immunodeficiencies. As part of the conditioning process to prepare the patient for the bone marrow transplant, the patient may be treated with total body
irradiation (TBI). The purpose of TBI is to eliminate the underlying disease and to suppress the recipient’s immune systems, thus preventing rejection of new donor stem cells. Once the conditioning treatment is complete, the patient receives the bone marrow transplant to restore healthy bone marrow function.

The goal for this research program is to design treatments that will not irradiate the whole body, but instead will only focus on the bone marrow. Such a treatments are no longer total body irradiation, but total marrow irradiation (TMI).

TMI treatment plans will be developed using a mathematical model that will provide for treatments resulting in the desired eradication of existing bone marrow cells while simultaneously avoiding organs and healthy tissues that do not require irradiation. This will be achieved by using intensity modulated radiation therapy (IMRT), a type of radiation therapy capable of delivering any distribution of radiation intensity from each beam. The resulting treatments will be better able to eradicate the patient’s existing bone marrow with potentially fewer side effects from radiation overdose, thereby improving the patient’s quality of life and preparation for a bone marrow transplant.

Automated quality assurance

The process of validating a radiotherapy treatment plan requires a quality assurance review by an expert. If the plan is deemed acceptable, it proceeds to treatment. Otherwise, it is returned to a dosimetrist to revise the treatment to meet standards for the particular clinic. This review is time-consuming, taking as much as tens of thousands of man-hours annually at large treatment centers, and may be subject to human errors.

The goal of this research is to automate the QA process using machine learning to learn clinic treatment standards and flag treatments that do not meet those standards. There can be little tolerance for incorrectly identifying an erroneous plan, as there may significant health implications for the patient. Notably, a plan that does not conform to standards may not be erroneous; it may be part of trial of new treatments or the patient may require unusual treatment. Thus, any flagged plan must be manually reviewed.

The challenge in correctly classifying plans as acceptable or not is twofold. First, radiotherapy QA datasets are extremely high-dimensional due to the number of features required to represent dose distributions. Second, the datasets are highly imbalanced in that most stored plans are acceptable, as unacceptable plans are generally overwritten with improved plans. We develop classification and anomaly detection tools to identify unacceptable plans with a low false negative rate, and interpretation methods to communicate to dosimetrists how to improve unacceptable plans.

Single-entry waitlist surgical scheduling

Thu, 04 Nov 2021 22:45:23 +0000

The research is performed in collaboration the University Health Network (UHN). To improve utilization of resources required in operating room scheduling, e.g., operating room time, surgeon availability, downstream beds (ICU, ward, etc.), hospitals can collaborate in a strategic network to pool resources. This network results in a single-entry waitlist for surgery patients, where patients enter the waitlist and get assigned to a surgeon at a participating hospital.

The resulting scheduling problem is a large and complex optimization, which we solve through novel decomposition techniques. Our findings demonstrate that resources across all participating hospitals are better utilized with a single-entry waitlist, and we use game theory concepts to show that each hospital better achieves its own goals of maximizing surgeries and minimizing wait times by participating in the network. We additionally find that a year’s worth of surgeries can be completed in just 9-10 months, freeing up operating rooms for targeted waitlist reductions.

We additionally examine single-entry waitlists at a provincial level, using simulation and simple routing rules, to understand the efficiency that can be gained by widespread adoption of single-entry waitlists, and to understand what percent of surgical patients should opt-in to a centralized waitlist to realize those efficiencies.

Kidney paired donations

Thu, 04 Nov 2021 22:43:26 +0000

Kidney paired donation programs allow patients registered with an incompatible donor to receive a suitable kidney from another donor, as long as the latter’s co-registered patient, if any, also receives a kidney from a different donor. The kidney exchange problem (KEP) aims to find an optimal collection of kidney exchanges taking the form of cycles and chains. KEP is complex optimization problem, which we solve using graph theory and decomposition methods.

Bone marrow transplants

Thu, 04 Nov 2021 22:39:27 +0000

Diseases requiring bone marrow transplants (BMTs), also called stem cell transplants, include leukemia, lymphoma, sickle cell disease and aplastic anemia, as well as some immunodeficiencies. As part of the conditioning process to prepare the patient for the bone marrow transplant, the patient may be treated with total body irradiation (TBI). The purpose of TBI is to eliminate the underlying disease and to suppress the recipient’s immune systems, thus preventing rejection of new donor stem cells. Once the conditioning treatment is complete, the patient receives the bone marrow transplant to restore healthy bone marrow function.

The success of a bone marrow transplant is uncertain, and depends on many factors, including underlying diagnosis, health status, donor relation, etc. Other important factors may exist that are not yet known or well-understood by clinicians. By examining historical records of BMTs, we develop machine learning tools to predict the success of a BMT with a particular patient and donor, using only data regularly collected during the course of treatment. We transform these predictions into conventional Kaplan-Meier survival functions to help clinicians and patients understand individualized survival probabilities and select the best course of treatment.

We particularly focus on single-center datasets to ensure that predictions are appropriate for the patient mix actually seen at the treating hospital, rather than use very large datasets covering many hospitals and regions, which may bias predictions with respect to any one hospital. Single-center datasets are small, requiring novel approaches to obtain satisfactorily accurate predictions.

Pandemic modeling and planning

Thu, 04 Nov 2021 22:05:30 +0000

During a pandemic disease outbreak, it is important to have an accurate estimate of the number and location of individuals affected. The efforts to do so require a multi-disciplinary collaboration since most of the factors involved are interrelated. Much of the research in this area is dedicated to obtaining the same objectives using homogenous mixing models, where the affected population is classified into three groups; susceptible, infectious and removed. However, to better understand the properties of an outbreak, there are more factors to be explained than the estimated number alone. It is also important to obtain information regarding the area where the disease may spread, the risk levels in the different parts of that area and the possible direction it may continue to spread. Other factors to be considered are the transmission mode of the disease and the role of public facilities such as public transport and buildings, e.g., offices and grocery stores, in the disease transmission. Additional concerns include addressing pedestrian foot traffic as well as children and transient populations.

This research uses agent-based simulation models and contact network models to obtain the estimate the spread of disease through a population generated using data from census, public transportation, school system, etc. sources. We build algorithms to optimize public health mitigation strategies that account for both resource costs and societal costs based on simulation outcomes. We also apply graph theory concepts, specifically focusing on the critical node detection problem (CNDP), to identify optimal individuals in the population to vaccinate. We then use rule-mining techniques to extract actionable vaccination prioritization policies.