Friday, October 16, 2026
Molloy University faculty and students are making strides in cancer-related research, using artificial intelligence to explore how computers could predict the spread of breast cancer and identify brain tumors in MRI scans.
Helen Dang, a professor of mathematics and computer science, leads research examining whether AI can analyze genetic information to predict where breast cancer may spread. She also mentors senior Liam Fleck, who is developing separate software to distinguish normal brain scans from scans showing tumors.
The projects use different data and remain research-stage efforts, not approved clinical diagnostic tools.
“Teaching and research are in my genes,” Dang wrote in an email. “I am particularly passionate about healthcare.”
Gina Florio, dean of Molloy’s School of Arts and Sciences, said Dang joined the university in September 2023 after working as a research scientist at the University of Massachusetts-Boston.
Dang applies AI to bioinformatics, health care and disease prediction, with the goal of improving disease detection, diagnostics and clinical decision-making, Florio said. She described Dang’s role in student research as providing direction, teaching advanced methods and overseeing projects while allowing students to work independently.
“She provides vision and oversight,” Florio said.
Florio said the students contribute to projects that are part of Dang’s research agenda.
One project examines whether a hybrid model combining classical and quantum computing can predict where breast cancer has metastasized. The framework, called a Hybrid Quantum-Classical Neural Network for Metastatic Site Prediction, is designed to classify whether cancer has spread to the brain, lungs, bone or liver.
The project uses gene-expression data. Its dataset contains 65 breast cancer patient samples from a publicly available database. The original dataset contained more than 22,000 probes, so researchers used processing and feature-selection methods to narrow the information relevant to the model for prediction.
The model combines quantum circuits with classical neural-network layers to identify genetic patterns associated with metastatic sites.
The model achieved 94.4 percent classification accuracy on the research dataset. The study also reported 95.6 percent precision and a 94 percent recall and F1 score. An F1 score is a classification metric that evaluates whether a model correctly identifies a category.
Those figures measure the model’s ability to classify metastatic sites within the dataset. They do not mean it diagnoses breast cancer with 94.4 percent accuracy or establish that it can predict cancer’s spread in patients outside the study. Testing with larger, independent patient groups is needed before the approach could be considered for clinical use.
“If this sort of technology existed when I was younger, there may have been ways of getting yearly checkups for all kinds of cancer,” Fleck said.
Over three years, Dang has mentored five projects involving nine students. Two additional cancer-related AI projects emerged from a capstone course in spring 2026, and two papers were presented at Molloy’s undergraduate research conference.
Fleck’s separate project focuses on brain tumors and MRI scans. The 22-year-old senior said it began as a computer science capstone project last January. He chose the subject partly because cancer affected his family. His father, Christopher, was diagnosed with stage 4 lymphoma when Fleck was 12 and died when he was 16.
The assignment was intended as a one-semester proof of concept, but Fleck continued developing it. He now works independently on the software with Dang’s mentorship and oversight.
The program uses MRI scans and training data from publicly available databases. It is designed to distinguish normal brain scans from scans showing three types of brain tumors. Fleck said it compares scans with patterns learned from examples.
He reported that the model’s accuracy was 68 percent in April or May, rose to 89 percent by the end of May and reached 92 percent about a week before his interview. Those are development-stage results, not clinical-trial findings.
Fleck said further gains have become harder as accuracy has increased. The software is not commercially viable and needs more data, he said. A major challenge is obtaining enough accurately labeled MRI images. Fleck needs help from people qualified to interpret scans and identify tumor types.
He has begun contacting universities, including Stony Brook, and hopes to approach Mercy Hospital in Rockville Centre to explore whether more data might be available. Neither contact is an established partnership. The project has not been used in real-world medical applications, and further development is needed to address data limitations and medical approval requirements.
Fleck hopes to improve the model and eventually broaden its capabilities to help identify more cancers. Routine whole-body cancer screening would require advances beyond his current software. For now, he is focused on improving the model.
Florio said undergraduate research lets students apply classroom learning to practical problems while working with faculty experts. “Undergraduate research is a high impact educational practice,” she said, describing hands-on work on real-world problems.
Florio said she wants students in every discipline to build meaningful relationships with faculty mentors while exploring their interests.
“I want all of my students, regardless of discipline, to be able to engage both with their faculty ... in meaningful mentoring relationships, but also with their content, with their area of interest,” Florio said.
For Fleck, the project is personal. Losing his father informs his hope that better technology could help families.
“Losing a parent as a kid is not something easy to do,” he said.
