Read this article in Thai

Thailand has about 2,000 radiologists, who between them screen over 30 million X-rays produced by the country’s healthcare system annually. That means on average, a radiologist assesses 15,000 images a year, a heavy workload these medical professionals say can affect their performance.

“A single radiologist often reviews 60 to 200 chest X-rays daily, spending between 5 and 15 minutes on each. That number can even exceed 300 in busier hospitals,” said Dr. Narissara Chobarun, radiologist in the thoracic radiology unit at the Faculty of Medicine Ramathibodi Hospital, Mahidol University. “This can lead to fatigue, which could affect the speed and accuracy of their diagnosis, particularly for subtle or early-stage lesions that are often difficult to detect.”

To address this issue, Ramathibodi Hospital developed the RAMAAI CXR Solution, an AI assistant running on Microsoft Azure’s cloud platform that screens chest X-rays and helps doctors prioritize severe or infectious conditions.

RAMAAI can detect up to 16 types of diseases and abnormalities, including lung nodules, pneumonia and COPD, or chronic obstructive pulmonary disease.

It also has a dedicated module for diagnosing tuberculosis, a lung infection commonly known as TB. RAMAAI not only detects TB, it can predict from the X-ray image whether the patient is at the infectious stage – a critical capability for controlling disease spread in high-incidence countries like Thailand. “It plays a vital role in initial screening. If an image shows a high probability of abnormality, the system flags it for immediate review, ensuring patients receive timely diagnosis and care,” said Dr. Narissara. “Additionally, the system employs a heatmap to highlight suspicious areas, enhancing the thoroughness of chest X-ray interpretations.”

The AI was developed in-house and trained on data from chest X-rays of patients at Ramathibodi Hospital. This gives it a deep understanding of Thai people’s physiology and disease patterns, for more accurate and reliable diagnoses of the local population. 

Dr. Chayanin Nitivattanakul, Assistant Dean for Innovation and Partnership and Attending Radiologist in the Thoracic Radiology Unit at the Faculty of Medicine Ramathibodi Hospital, Mahidol University said, “When it was tested during development on historical X-rays where lung cancer diagnoses were initially missed, RAMAAI demonstrated a 72% increase in lesion identification.”

Improving healthcare outcomes 

RAMAAI was rolled out in 8 public hospitals nationwide including Bangkok, Samut Prakarn, Chonburi, Lampang and Chaing Rai serving about 2,000 medical professionals.

So far, it has processed over 1,500-2,000 medical images daily, with more than 500,000 images analyzed to date. According to Dr. Chayanin, using RAMAAI has boosted doctors’ overall lesion detection accuracy by over 20%

Scaling impact for equitable healthcare 

Ramathibodi Hospital now hopes to expand the use of RAMAAI to more hospitals and further improve its technology.

Another 10 public hospitals are expected to adopt it by June this year. Meanwhile, Thailand’s Department of Medical Services plans to pilot the AI assistant across its affiliated hospitals.

This expansion prioritizes remote areas, ensuring equitable access to quality diagnostics. RAMAAI is also engineered for integration with mobile X-ray units, which are used in rural regions with fewer medical facilities.

Having Microsoft Azure as its cloud platform supports these future plans. “We chose Microsoft Azure because Microsoft is a global leader, offering internationally standardized technology and innovation, paramount data security, and flexible scalability to support future expansion to numerous hospitals and users,” said Dr. Chayanin. 

The hospital is also experimenting with AI models available in Microsoft Foundry, including BiomedCLIP, a medical imaging model, and the Phi‑3‑Mini compact language model. These are being integrated into its CXRReportGen, a pilot multimodal AI tool that can help draft written reports after the X-rays have been analyzed. “RAMAAI helps identify abnormalities more effectively,” said Dr. Chayanin. “Earlier intervention for patients leads to timely care and reduced complications, infections, and disease spread. Ultimately, this lowers overall public healthcare costs.”