At the Global AI Impact Summit being held at Bharat Mandapam from February 16 to 20, one of the busiest corners was the health pavilion, where a striking installation stopped visitors in their tracks.
A life-sized mannequin lay flat on a hospital bed while a robotic arm moved over it, mimicking an ultrasound scan. Nearby, four screens displayed simulated vital parameters, offering a glimpse of how artificial intelligence could reshape diagnostics in the near future.
The demonstration showcased an AI model designed to assist radiologists by analysing ultrasound images in real time and flagging potential abnormalities. According to the developers, the system can significantly shorten the time taken to prepare diagnostic reports.
“This model scans the ultrasound and highlights areas of concern for the radiologist in real time,” said Varun Dube, who heads innovation at Wipro, which is developing the technology. “Once the radiologist verifies the scan, the report can be generated almost immediately, accelerating the entire diagnostic cycle,” he added.
Currently, patients often wait several hours, or sometimes a day, for ultrasound reports. By narrowing the radiologist’s focus to areas that may need attention and enabling automated measurements, the AI system aims to speed up both analysis and reporting.
Dube emphasised that the technology is meant to assist doctors, not replace them. “AI will not remove radiologists,” he said. “It helps them make decisions faster and handle more cases efficiently.”
The project has been under development for about a year and involves collaboration with academic and research institutions, including the Indian Institute of Technology (IIT) Delhi, Indian Institute of Science (IISc), Bangalore. Training has been carried out using a combination of open-source, synthetic, and real-world datasets.
However, Dube noted that regulatory approval and clinical validation remain critical steps before such systems can be widely deployed. “Because this is healthcare, the technology must be foolproof and go through trials and government approvals. Patient safety is the priority,” he said.
Data availability is another major challenge. AI systems require large, unbiased datasets to perform reliably, but privacy concerns often limit access to patient data. “Healthcare data is highly sensitive. Without adequate data, developing robust AI models becomes difficult,” Dubey said.
The company plans to work with manufacturers of ultrasound and other imaging equipment to embed the AI model directly into their machines. The software-based approach means the system could potentially be integrated into existing devices without major hardware changes, making adoption easier for hospitals and diagnostic centres.
While the demonstration focused on healthcare, similar AI-driven visual analysis models are being applied in other sectors as well. Dube described applications ranging from inspecting corrosion in industrial pipelines to surveying hazardous environments where sending humans would be risky, such as boilers, sewers, or disaster sites.
In such cases, AI models can be deployed on robots, drones, or other devices to detect cracks, leaks, or structural damage and relay information in real time. According to Dube, these inspection technologies are already seeing use in sectors such as oil and gas and civic infrastructure in various parts of the world.
Back at the summit pavilion, the robotic ultrasound demonstration continued to draw crowds, offering a tangible example of how AI could soon become an invisible but integral layer in medical devices, helping doctors work faster, patients receive results sooner, and healthcare systems manage growing demand more efficiently.