Mecha Health raises $4.1M seed round to build next-generation foundation models for radiology

Radiology has never stood still.

From film to PACS, from early CT to whole-body MRI, radiologists have adapted to each wave of technological change. But the pace hasn’t slowed, it’s only accelerated.

Today, imaging volume is soaring, exam complexity keeps multiplying and comparison workflows stretch across years of studies and dozens of modalities. Every radiologist knows the feeling: more images, more priors, more clicks, same number of hours in the day.

Despite the established history of AI vendors in radiology, most companies do not build systems that truly fit the way radiologists actually work.

Mecha Health, a San Francisco-based startup founded out of UCL and Y Combinator, has raised a $4.1M seed round led by Valia Ventures, with participation from Y Combinator, to build systems that change this.

Instead of building single-task algorithms that classify isolated findings, Mecha Health is developing full-context foundation models trained to read scans the way radiologists do:

• Multi-anatomy
• Multi-modality
• Multi-view
• Prior-aware
• Structured reporting

These models don’t just detect abnormalities, they generate full radiology reports, reading pixels and voxels across entire studies, comparing priors and incorporating clinical indication and technique.

“Our focus is simple,” said co-founder Ahmed Abdulaal. “Radiologists need tools that don’t ask them to change how they work. They need tools that understand how they work.”

The funding will accelerate development of Mecha Net XR, a generalist radiograph model capable of reading chest, abdomen, spine, hand, foot, knee, shoulder, and hip X-rays, and Mecha Net CT, a full-context foundation model that reads entire CT chest volumes and generates draft reports in seconds.

“A radiologist doesn’t open a scan to answer one question,” said co-founder Deji Ijishakin. “They’re reasoning through anatomy, modality, history, priors, and context all at once. Foundation models are the first technology that can operate at that level of complexity.”

Mecha Health will exhibit for the first time at the RSNA Annual Meeting, Booth #5311 in the AI Showcase, where the team will unveil new product capabilities and demonstrations.

“Radiology is reaching an inflection point,” said Abdulaal “The workload keeps rising. The complexity keeps rising. Foundation models represent the first real opportunity since PACS to reshape the entire reporting experience.”