Projects offered by our partners this semester. Each stakeholder brings a real problem and data, and advises the team.
Dr. Hannah Baranes, Coastal Scientist, Coastal Dynamics Lab, Gulf of Maine Research Institute
Maine is installing low-cost tide gauges so the National Weather Service can forecast coastal floods in more places. The slowest, costliest step is surveying each gauge’s exact elevation – without it, the gauge’s readings can’t be turned into water levels. The SWOT satellite, launched in 2022, is the first to measure sea level accurately near the coast. This project compares SWOT’s water heights with Maine’s ~20 existing gauges to see whether the satellite is accurate enough to replace the survey. The payoff: removing one of the biggest bottlenecks to expanding Maine’s flood forecasting network, with results well-positioned for publication.
Dr. Hannah Baranes, Coastal Scientist, Coastal Dynamics Lab, Gulf of Maine Research Institute
Communities judge coastal flood risk from FEMA maps, which ignore sea level rise and are built inconsistently from place to place. Maine is developing a new property-scale model that includes sea level rise, but FEMA maps will stay the legal standard, so the two will exist side by side. Using Massachusetts’ already-released twin of the model, this project finds where and why the two disagree (waves vs. tides and storm surge, and in which kinds of places), then builds a viewer to compare them side by side. The payoff: a viewer that the State adopts and that coastal communities across Maine use directly.
Rory Dunn, Senior Platform Engineering Manager (primary contact), and Dr. Zachary Kabelac, Founder & CEO, Aperture Space, Inc.
Aperture Space, a Roux climate-tech start-up, is building satellite sensors that measure soil moisture and other water properties, and will need a way to calibrate them once they’re in orbit. As a first step, this project asks what free European Sentinel satellites already reveal: it builds a machine learning pipeline that detects open water and saturated soil from Sentinel-1 radar and Sentinel-2 infrared imagery, checked against ground measurements from Oklahoma’s weather-station network. The payoff: a data-science pipeline for sensor calibration that could inform a system used when Aperture’s sensors start operating in space.
Dr. Bill Lenart, Research Scientist, The Roux Institute – working with the CRIPT polymer database
Designing better polymer materials depends on knowing how a material’s composition, manufacturing process and properties connect, but that data exists as small, disparate sets that need expert work to harmonize. CRIPT is a shared database built to fix that. This project queries CRIPT to map which polymer systems have enough complete data to support machine learning, and connects that experimental data with a simulation workflow that builds phase diagrams automatically. The payoff: validating CRIPT’s capacity to accelerate polymer materials discovery, and a map of the critical data gaps the community needs to address.
Dr. Brianne Olivieri-Mui, Assistant Professor, Bouvé College of Health Sciences; Affiliate Faculty, Roux Institute
Hospice care is critically underutilized, and people with HIV disproportionately lack the social and financial support to live at home at the end of life, and may fear discrimination in inpatient settings. Comparing people in hospice with and without HIV, this project asks who is and isn’t using hospice. The payoff: knowing to whom knowledge about hospice care should be targeted, and whom to focus on in qualitative work on why they are not using hospice services – in a study intended for a peer-reviewed journal.
Dr. Brianne Olivieri-Mui, Assistant Professor, Bouvé College of Health Sciences; Affiliate Faculty, Roux Institute
Care quality is usually measured as “of the people who should get a treatment, how many did?” – but those measures are rarely designed for patients with several chronic conditions at once. This project measures care quality for chronic and age-related conditions among older adults with and without HIV who have multiple conditions. The payoff: care quality measures that work for people with multiple comorbid conditions, with a possible extension to the role of insurance providers – publishable as a peer-reviewed paper.
Dr. Brianne Olivieri-Mui, Assistant Professor, Bouvé College of Health Sciences; Affiliate Faculty, Roux Institute
Some hospitals are paid one lump sum to cover a surgery and the 30 days after it. This project compares outcomes – readmissions, complications, deaths – for one surgery type (lower-limb or bowel) at hospitals paid that way versus other ways, including for patients with HIV, who are treated as “high risk”. The payoff: understanding how payment models impact surgical outcomes, including for high-risk patients – intended to be a publishable paper.
Dr. Christine Lary, Research Associate Professor, The Roux Institute
Alzheimer’s disease and related dementias and osteoporosis are diseases of aging that frequently occur together, share common risk factors, and may share molecular mechanisms – suggesting osteoporosis drugs might help prevent or delay dementia. Building on a cohort earlier students defined, this project follows people newly diagnosed with osteoporosis, compares those who start an oral bisphosphonate with those who don’t, and asks whether the treated group develops dementia less often within five years – adding propensity scores to the existing survival model to better account for who gets treated. The payoff: preliminary data to support the hypothesis that osteoporosis drugs may be neuroprotective.
Dr. Christine Lary, Research Associate Professor, The Roux Institute – in collaboration with investigators at The Jackson Laboratory
In collaboration with investigators at the Jackson Laboratory, the stakeholder is interested in vascular contributions to cognitive impairment and dementia. This project looks for early signals – blood markers of metabolism, inflammation and brain damage, genetics, eye scans, and diagnoses such as diabetes – that predict who will develop dementia or vascular dementia. Many of these may not be in the data, so it also records how often each one actually appears in medical records. The payoff: preliminary data for a prediction model of incident dementia or vascular dementia.
Team projects in Data Visualization with Web Technologies, each built with a stakeholder.
Christian Thomas, MD, Medical Director, Hematologist & Medical Oncologist, New England Cancer Specialists
DNA testing of tumors (next-generation sequencing) has become common over the last decade, and the results can point to targeted drugs – sometimes a pill that makes the cancer go away. But even when a test finds a mutation that suggests a life-saving treatment, not every patient gets it: another treatment may be working, insurance may be an issue, or the patient may be afraid. Using a 10-year look back at genomic testing at NECS, this project visualizes who does and doesn’t receive test-informed treatment, and how that relates to factors such as where patients live, education, insurance and performance status. The payoff: new ideas for the treatment approach, or for an intervention or clinical study, that could improve patient outcomes. The work starts out private, under the data-use terms.
Nikki Rutman, Managing Partner & Chief Intelligence Officer, Sentinel Security
Sentinel Security assesses the risks its clients face at specific locations – an organization with several offices, a sports team on the road, an executive traveling abroad. A summer intern built a prototype dashboard from public data, for the US only. This project builds on that prototype: risks at point locations rather than counties (floodplains, hurricanes, earthquakes and tsunamis, protests, crime), toward the global view Sentinel wants, with case studies such as sports facilities and universities. It serves two audiences: Sentinel’s analysts, who dig into the details, and clients, who need a simpler view. The payoff: a dashboard Sentinel can integrate into its website.
Anna Levitsky, Executive Director, Woodie Wheaton Land Trust
The Woodie Wheaton Land Trust conserves about 4,600 acres in Washington County, Maine, where it practices sustainable forestry and wants its land to provide opportunity for local residents. Its executive director is its only full-time staff, and the data it needs is scattered: state databases, word of mouth, and parcel maps from Amity, Maine, that exist only on paper. This project builds an interactive map that applies WWLT’s property-ranking rubric to identify and compare high-priority areas for conservation, with a dashboard that summarizes land already conserved and land yet to be conserved. The payoff: a tool for choosing future projects, and for fundraising, campaigns and communicating the trust’s conservation impact.