Doctors use brain scans called MRIs to look for problems like tumors, but reading these scans by hand takes a lot of time. A specialist has to carefully trace the outline of each part of the brain, one section at a time, and different doctors can draw those lines slightly differently.

We're building a computer program that can look at a brain scan and learn what makes each part of the brain look different from the others — almost like each part has its own unique fingerprint. Our goal is for the program to eventually map out a brain scan automatically, just as accurately as a doctor, but much faster.

This project is run by a team of local high school students in partnership with Penn State Harrisburg, giving us real hands-on experience with a problem that could genuinely help doctors and patients.

Manual segmentation of MRI brain scans is time-consuming, costly, and prone to inter-practitioner variability. Clinicians who need precise anatomical boundaries currently rely on specialists to manually delineate structures slice by slice.

Our team is developing a Python-based program that derives a statistical fingerprint for each brain structure from MRI pixel data. The pipeline samples pixel-level data from specified structures, computes statistical descriptors (mean, variance, skewness, kurtosis), and compares confidence intervals across structures to distinguish them from one another and from tumor tissue. The underlying hypothesis: each brain structure exhibits a unique, reproducible statistical profile based on Hounsfield unit intensity distributions.

This research is conducted by a student research team in partnership with Penn State Harrisburg, working toward an AI classification layer capable of automated brain structure segmentation.