Healthcare imaging requires infrastructure built for scale, interoperability, security, and data complexity.

"One of the primary reasons we chose Google Cloud is the Healthcare API, which gives us a healthcare-forward way to stream and store imaging data," says Sokka. "The other is the AI stack—both the out-of-the-box tools and the research behind medical-grade models."

The Cloud Healthcare API handles DeepHealth's image streaming, delivering more than 20 million medical images per year with security, privacy, and data governance compliance capabilities designed to support healthcare data requirements. Multi-zone replication keeps availability high even across regions. DeepHealth uses Gemini Enterprise Agent Platform to run agentic tooling across radiology workflows, mainly routing the right exams to the right specialists based on qualifications and expertise.

Gemini handles auto-summarization of radiology reports. For example, when a radiologist dictates their findings, Gemini helps draft or summarize report content that, once finalized, referring physicians can act on, cutting reporting time down for an initial treatment decision. But the benefit goes beyond speed. As Sokka puts it, "AI can help improve consistency."

For specialized clinical AI, DeepHealth works closely with Google Research on MedGemma, a medical-grade language model purpose-built for healthcare. "Not all conventional LLMs are medical grade," says Sokka. "MedGemma is one of the strongest for medical awareness and medical use cases, and we've fine-tuned it for our specific needs."

DeepHealth uses BigQuery to curate data cohorts for AI model training and to provide performance analytics to customers, from physicians tracking their own outcomes to technologists monitoring scan quality. Looker sits on top of the stack to deliver organizational dashboards across the network, so even non-technical users can still gain insights.

DeepHealth is best known for its clinical AI applications with flagship tools in cancer screening, spanning breast, lung, prostate, and thyroid cancer.

When the industry moved from 2D to 3D mammography, cancer detection rates improved 10–12%. DeepHealth's AI, layered on top of 3D mammography, can deliver an additional 21% improvement⁴. "Equally important is that we're also finding more aggressive cancers earlier⁵, so the physician can provide treatment options sooner," says Sokka.

In lung cancer, DeepHealth's solutions support radiologists across more than 90% of the NHS England’s Lung Cancer Screening Programme's sites. Before the program, just 29% of lung cancers were found at an earlier, more treatable stages. Now, 76% are.⁶ "Stage shifting disease, particularly cancer, can mean saving lives," says Sokka.