Kanta Saito, Wilson Huynh, and Elijah Saloma in front of their title slide at the 2026 UH Mānoa Spring Undergraduate Showcase

Three undergraduate researchers from the Information and Computer Sciences (ICS) department at UH Mānoa — Kanta Saito, Wilson Huynh, and Elijah Saloma — presented their joint project at the 2026 Spring Undergraduate Showcase, the end-of-semester research conference co-organized by the Undergraduate Research Opportunities Program (UROP) ↗ and the Honors Program. The hybrid event was held on May 8, 2026 at UH Mānoa, drawing undergraduate presenters from across the campus for oral and poster sessions.

Their presentation — “Developing a Data Standardization Pipeline for Evaluation of Mammography AI in Hawaiʻi” — addressed a recurring obstacle for any group trying to evaluate commercial or research-grade mammography AI on Hawaiʻi data: imaging archives are heterogeneous in format, metadata, and labeling conventions, and most published AI models assume DICOM inputs and population characteristics that differ from what is actually collected in the islands. The students designed a pipeline that ingests, harmonizes, and quality-checks mammography images and associated metadata so downstream AI models can be benchmarked consistently on Hawaiʻi cohorts — a prerequisite for evaluating whether models trained elsewhere perform adequately for the Native Hawaiian, Pacific Islander, and Asian women who make up most of the HIPIMR registry.

The team

The project was carried out by the three ICS undergraduates with mentorship from:

  • Arianna Bunnell — graduate-student mentor; PhD candidate in ICS at UH Mānoa and a member of the Shepherd Research Lab. Read Arianna’s profile.
  • Peter Sadowski — senior mentor; Associate Professor in ICS at UH Mānoa, where he directs the Sadowski Lab ↗ on machine learning for the physical and life sciences.
  • John Shepherd — senior mentor; Director of the Shepherd Research Lab at the University of Hawaiʻi Cancer Center. Read John’s profile.

The pairing reflects an ongoing collaboration between the Sadowski Lab in ICS and the Shepherd Research Lab at the UH Cancer Center on AI methods for breast-cancer imaging — and on building the trainee pipeline that supports it.

Kanta Saito, Wilson Huynh, and Elijah Saloma after their presentation

Why this matters

Standardization sounds dull next to model architecture, but in practice the data-wrangling layer determines whether any AI evaluation in Hawaiʻi is comparable, reproducible, or even meaningful. By building this pipeline as undergraduates, Kanta, Wilson, and Elijah have taken on a piece of infrastructure the lab will rely on directly as new mammography AI tools are evaluated against HIPIMR data over the coming year.

Congratulations to Kanta, Wilson, and Elijah on a well-presented project, to Arianna for her mentorship, and to UROP and the UH Mānoa Honors Program for organizing the showcase. More information about the event is available at the UH Mānoa Undergraduate Showcase ↗.