In India, the challenge is stark. "We have just 20,000 radiologists in the country, and about 8,000 focus solely on ultrasounds. That leaves only 12,000 specialists to handle chest X-rays, CTs, and MRIs for 1.4 billion people," explains Ashutosh Pathak, CTO of DeepTek.
DeepTek’s founders, Ajit Patil and Dr. Amit Kharat saw this widening gap as an opportunity for technological intervention to increase imaging penetration in India and across the globe. Drawing from Ajit’s experience working with Japanese tech giants like NTT Data and Dr. Kharat’s deep radiology expertise, they envisioned AI as the key to transforming radiology.
The company boasts a growing global footprint, with over 500 active customers across India, APAC (Singapore, Thailand, Indonesia, Malaysia, Philippines, Japan), and with US FDA approvals and CE marks, fueling their expansion into the US and UK markets. In addition, Augmento is deployed as the national platform for radiology AI by the Singapore government.
The Radiology Crisis: Impact and Inefficiencies
Deeptek’s CTO, Ashutosh emphasizes that the skewed patient-to-radiologist ratio significantly impedes India’s ability to scale clinical diagnostics, resulting in:
- High workload & burnout: “There’s a lot of pressure and burden on the radiologist to do the reading, which could result in errors and delays,” explains Ashutosh. “The fast pace of work and shortage of radiology reading experts make the problem even worse.”
- Limited access beyond metros: “Many times, you can’t even provide imaging services beyond the Metro regions,” he adds, highlighting the disparity in healthcare access in rural India.
- Delays in diagnosis & treatment: The increasing volume of imaging studies, combined with a shortage of experts, leads to longer turnaround times for critical diagnoses.
DeepTek's AI-Powered Solutions: Augmento and Genki
DeepTek’s AI-assisted radiology solutions—Augmento and Genki—are designed to alleviate radiologists’ workloads, enhance diagnostic accuracy, expand access beyond urban centers, and accelerate turnaround times, leading to faster treatments.
At the core of both Augmento and Genki is DeepTek’s proprietary AI model for chest X-rays, which adopts an organ-based approach rather than the conventional single-finding-based model.
Ashutosh explains, “Unlike AI solutions that detect only specific conditions (pneumonia, pleural effusion, etc.), DeepTek’s model evaluates the entire chest for abnormalities, enhancing its diagnostic capability. It is one of the only globally available commercial AI models with US FDA certification, underscoring its clinical reliability and effectiveness.”
Decoding Augmento: AI-Powered PACS Integration
Augmento is a next-gen PACS + AI deployment platform. Picture Archiving and Communication System (PACS) is the standard software for managing medical images with a worklist for easy access and enabling them to view and interpret the images to generate reports.
Unlike traditional on-premise PACS from providers like GE, Philips, and Siemens, Augmento leverages cloud technology to deploy AI models, reducing infrastructure needs while enhancing medical image management and interpretation.
Key Features of Augmento:
- AI-Integrated PACS: Augmento enables AI-driven decision support, allowing radiologists to leverage advanced analytics without disrupting their time-tested workflow.
- Multi-Modality Integration: Supports AI model integration across multiple imaging modalities, including CT, MRI, X-ray, and ultrasound.
- Seamless Third-Party AI Deployment: Radiologists and healthcare providers can integrate various AI models from different vendors into Augmento’s ecosystem.
- Optimized Workflow Efficiency: By embedding AI insights directly into radiologists’ worklists, Augmento reduces turnaround time, enhances diagnostic accuracy, and minimizes the risk of human oversight.
Decoding Genki: AI-Driven Public Health Screening
Key Features of Genki:
- AI-Enabled Chest X-Ray Analysis: Genki’s can analyze chest X-rays in under 30 seconds to detect abnormalities, significantly speeding up the diagnostic process.
- Portable and Offline Functionality: Works on a basic laptop without requiring internet connectivity or high-end GPUs, making it ideal for field deployment.
- Cost-Effective Public Health Screening: Reduces the need for expensive and time-consuming laboratory-based sputum tests, enabling large-scale screening.
- Immediate Triage and Decision Support: Helps healthcare workers identify TB or other chest-related abnormalities on-site, allowing faster patient management and treatment.
Real-World Impact
Deployed with the Greater Chennai Corporation for four years, DeepTek’s Genki was used to screen 3,000 people monthly via six mobile X-ray vans. If TB was detected, sputum samples were collected immediately for confirmation and treatment.
Ashutosh pointed out that Genki had increased detection rates 25x—from 25 to 500 cases per 100,000—ensuring early diagnosis and reducing TB spread.
Ashutosh Pathak sums it up, “There is no way this can be done just by adding more radiologists. We simply don’t have enough radiologists to manually screen everyone or analyze every X-ray. AI is making possible what was previously impossible.”
Why DeepTek is commercially successful?
"Building an AI model isn’t enough," says Ashutosh. "If the AI model adds overheads to the radiologists instead of reducing workload, it defeats the purpose. Our models seamlessly integrate into clinical usage, improving efficiency without requiring large infrastructure or retraining for radiologists. It has enabled us to do commercial deployments across the globe. "
DeepTek’s success also lies in its deep understanding of the clinical landscape. Ashutosh shared that Dr. Amit Kharat, DeepTek’s co-founder and Professor of Radiology for over 18 years at Dr. D.Y. Patil University, provides invaluable clinical insights into product development.
Additionally, DeepTek collaborates with over 120 radiologists, ensuring continuous feedback and improvement. For instance, early user testing revealed inefficiencies in workflow design to access AI reports and outputs.
"We initially built a solution expecting radiologists to interact with AI in a certain way, but feedback showed that excessive clicks slowed them down. Instead, they preferred keyboard shortcuts. We have to build workflows for such default behaviors of the user," explains Ashutosh.
Tackling Legacy Systems, Interoperability, and Privacy
"We’ve developed a one-click context switch to integrate with PACS," Ashutosh shares. "Radiologists primarily use PACS for viewing studies. While they are analyzing an image in PACS, they can click a single button, which opens our AI viewer on top of it. This provides a seamless way to access AI insights without switching between multiple systems."
"The goal is that for a radiologist, it's just a click—AI output appears, they visualize it, accept or reject it, get the report, submit, and they’re done. The window closes, and they’re back in PACS.”
Data Privacy: A top priority
“All clinical establishments are extremely wary of sharing or leaking patient PII (personally identifiable information). Now, Augmento is a cloud-based solution. Cloud offers easy scalability, continuous updates, and other benefits that an on-premise system may not provide. But the challenge is transferring scans to the cloud while ensuring data privacy," says Ashutosh.
"Before a scan is sent to Augmento, the modality or local system first sends it to our Gateway—a lightweight software that runs on a simple laptop," Ashutosh explains. “It strips all personal data before sending the scan to the cloud. This process, called anonymization or de-identification, ensures compliance with privacy regulations while still enabling AI-powered analysis and reporting."
Addressing the Fear and Trust in AI
“There can be false positives and false negatives in the AI models. So, does that mean AI can't be used? That’s the most common objection from clinicians. But the key is not to use AI in isolation but to build clinical workflows around it to derive meaningful benefits."
AI can significantly optimize workflows even if it isn’t perfectly accurate. Ashutosh illustrates this with an example from Augmento’s chest X-ray algorithm:
- A radiologist reviewing 150 chest X-rays a day follows a sequential process, meaning a critical case at the 100th scan might be delayed until all previous scans are reviewed.
- However, AI can help triage cases by quickly identifying abnormal scans.
- If 17 out of 150 scans are flagged as abnormal, radiologists can prioritize those cases first, improving turnaround time.
- Even if the AI misclassifies one or two scans, capturing 15 or 16 abnormal cases early leads to better patient outcomes.
“DeepTek's models surpass these benchmarks, currently performing at 90-80 and targeting 90-90,” says Ashutosh.
To further build trust, Ashutosh explained that DeepTek employs Explainable AI and Responsible AI principles in its AI models.
Explainable AI ensures that radiologists can interpret AI outputs effortlessly. Instead of flipping between images, they can overlay AI-generated insights onto the original scan with a simple toggle, allowing quick validation.
Responsible AI provides real-time monitoring and post-deployment analysis. It helps track performance across different demographics, model drift over time, and cases where radiologists agree or disagree with AI suggestions.
