AI breakthrough could end deadly guesswork in pneumonia diagnosis

 

What you need to know:

1. Researchers from UC San Francisco have developed an artificial intelligence-based system that diagnoses pneumonia and other lung infections with over 90 per cent accuracy, potentially transforming care for critically ill patients.
2. The system correctly identified infections 96 per cent of the time by combining analysis of medical records using GPT-4 with a blood biomarker known as FABP4, dramatically reducing unnecessary antibiotic use.
3. In countries like Kenya, pneumonia remains a leading cause of death among young children and laboratory resources are limited.

In the shadowy corridors of intensive care units, doctors face a deadly guessing game. A patient lies struggling to breathe, their lungs compromised, their body failing. Is it pneumonia? Another infection? Or something else entirely?

The wrong answer does not just mean wasted antibiotics, it can mean the difference between life and death. Yet, despite decades of medical advancement, accurately diagnosing lung infections in critically ill patients remains one of medicine's most worrying challenges.

Now, a groundbreaking discovery from UC San Francisco researchers promises to change that, offering hope for nations where respiratory infections claim thousands of lives annually. The breakthrough combines artificial intelligence (AI) with a novel biomarker to diagnose pneumonia and other lower respiratory infections with over 90 per cent accuracy.

In a study published in Nature Communications, the system correctly identified infections 96 per cent of the time, potentially reducing inappropriate antibiotic use by over 80 per cent. “We've devised a method that gives results much faster than a culture, and it could be easy to implement in the clinic,” explained Dr Chaz Langelier, the study's senior author. “We're confident that it could lead to faster diagnosis and curtail the unnecessary use of antibiotics.”

The system pairs GPT-4 artificial intelligence analysis of medical records with a biomarker called FABP4—a gene that regulates inflammation and shows reduced expression in immune cells compared to normal lung tissue. The study looked at data from two sets of critically ill patients: 98 were recruited before the Covid-19 pandemic, and most had bacterial infections, while 59 were recruited during the pandemic, and most had viral infections, including Covid-19.

For Kenya, where pneumonia kills around 15,000 under-fives annually and ranks among the leading causes of death, this technology could be transformative. The country's healthcare system faces chronic shortages of trained specialists and laboratory resources, with many rural facilities lacking access to culture testing that can take days to produce results.

An AI-powered diagnostic tool requiring only standard medical records and a simple blood test could democratise access to accurate pneumonia diagnosis across the 47 counties. Additionally, Kenya's growing antibiotic resistance crisis, driven partly by inappropriate antimicrobial prescribing makes the system's ability to reduce unnecessary antibiotic use particularly valuable.

The World Health Organization estimates that antibiotic resistance could push 24 million people into extreme poverty by 2030, with low-income countries bearing the heaviest burden.