Research & Talks

Speaking up, so fewer people find out too late.

PinkAlert began with a personal experience of breast cancer as a teenager. That experience became peer-reviewed machine-learning research, an open-source analyzer anyone can use, a patent-pending handheld ultrasound device, and a TEDx talk. Together they form a public effort to make breast health something every young woman understands, and to make early screening affordable enough to reach her.

Arshia Ghatak at TEDxDVHS. Watch on YouTube

Early detection should not depend on who can afford an ultrasound, or who was taught what to look for.

Breast cancer is among the most commonly diagnosed cancers in women worldwide, and early detection is one of the strongest predictors of survival. Yet teenagers and young women are rarely taught how to monitor their own breast health, and the tools that could help, such as ultrasound, sit behind cost and access barriers.

PinkAlert works on three fronts: awareness, so young women know what to look for; access, through a low-cost handheld scanner; and openness, by publishing the research and releasing the code for anyone to study and improve.

Research

Peer-reviewed and published.

More research is on the way: a second paper is in progress.

Machine Learning Models for Breast Cancer Diagnosis Using Ultrasound Images

Arshia Ghatak & Dr. Jeremy HittAmerican Journal of Student ResearchVol. 4, No. 1 · February 2026

The study compares three classical machine-learning models (Support Vector Machine, Random Forest and Multilayer Perceptron) for classifying breast ultrasound images as benign, malignant or normal, using engineered texture, shape and intensity features. Random Forest performed best individually. A weighted ensemble of all three improved classification stability and detected malignant tissue more reliably than any single model, supporting interpretable, low-resource approaches to breast ultrasound research.

Read the paper Download PDF Source code

Ghatak, A., & Hitt, J. (2026). Machine Learning Models for Breast Cancer Diagnosis Using Ultrasound Images. American Journal of Student Research, 4(1). https://doi.org/10.70251/HYJR2348.41777784

Get involved

Three ways to help.

Know your normal

Learn what healthy breast tissue feels like, speak with a doctor about any change, and share that knowledge with friends.

Contribute

Researchers and developers can review the repository, open an issue, or propose improvements to the analyzer.

Share the message

Bring the TEDx talk to your school, club or community. Awareness is where early detection begins.