Fonya explores imaging language model during internship in China

Brandone Fonya, a Carnegie Mellon University Africa student earning his master's degree in engineering artificial intelligence, spent his summer at an internship at Westlake University in Hangzhou, China. One of 199 interns selected from a pool of over 2,100 applicants, Fonya spent his internship working under Professor Kaicheng Yu in the School of Engineering's Autonomous Intelligence Lab, exploring the impact of a language-guided segmentation on prostate Magnetic Resonance Imaging (MRI).

"My experience this summer in Hangzhou was life changing. It is the heart of tech innovation in China and home to Alibaba, Unitree, and DeepSeek, among others," Fonya said. "I collaborated with talented lab members and other interns from top institutions in China and worldwide, gaining hands-on experience in research and innovation."

His internship research project, MedBLIPNet3D: Text Prompt-Guided vision language model for 3D MRI Prostate segmentation, focused on training MedBLIPNet3D, a 3D vision language architecture, to segment prostate MRI scans using natural language prompts.

In his internship proposal, he notes that automated segmentation of medical images is a critical step in numerous clinical workflows, including diagnosis, treatment planning, and disease monitoring. Recent research has demonstrated the feasibility of leveraging vision-language models to enable segmentation guided by prompts in 2D imaging scenarios.

However, the extension of prompt-guided segmentation to 3D volumetric imaging, such as MRI and other radiology modalities remain underexplored due to challenges including computational complexity, variability in acquisition protocols and limited annotated data.

He added that experiments evaluate whole-gland segmentation and segmentation of various prostate zones guided by text prompts, demonstrating the model’s ability to generalize across sites and segmentation objectives.

This semester at CMU-Africa, Fonya is a graduate teaching assistant for 18-751: Applied Stochastic Processes, helping engineering AI students build a strong mathematical foundation for machine learning.

He has also been selected to serve as a reviewer for Applied Machine Learning Days (AMLD) Africa 2026, taking place in Johannesburg, South Africa. AMLD, founded by Ecole Polytechnique Fédérale de Lausanne, is one of Europe's largest machine learning conferences. He is looking forward to contributing to the event by evaluating submissions and ensuring they meet high academic standards.

As for his internship last summer, Fonya said, "Overall, my summer was an unforgettable mix of advanced research and adventure in a distant land, putting my skills to the test. Exploring Shanghai and other cities, going on hikes in the Hangzhou mountains, visiting temples, and encountering a panda for the first time, added rich highlights to the experience."