AI can identify drug-resistant infections from microscopy images
University of Cambridge study finds
Groundbreaking research led by scientists at the University of Cambridge has demonstrated the potential of artificial intelligence (AI) to identify drug-resistant infections from microscopy images quickly.
This advancement promises to revolutionise how healthcare professionals diagnose and treat diseases, potentially reducing the diagnostic timeline from days to mere hours.
Antimicrobial resistance, a rapidly escalating global health threat, is increasingly limiting effective treatment options and raising grave concerns about the emergence of untreatable infections.
Traditional testing methods often require days for bacteria to be cultured, tested against antibiotics, and analysed, delaying crucial treatment decisions and risking patient outcomes.
In a study published in Nature Communications, researchers from Professor Stephen Baker’s Lab utilised AI to discern drug-resistant Salmonella typhimurium bacteria from standard microscopy images. This bacterium is known to cause severe gastrointestinal and typhoid-like illnesses, with symptoms ranging from fever to potentially life-threatening conditions.
Dr Tuan-Anh Tran, formerly of the University of Oxford and now at the University of Cambridge, highlighted the AI’s ability to detect subtle features indicative of resistance invisible to the human eye.
“The beauty of the machine learning model is that it can identify resistant bacteria based on a few subtle features on microscopy images that human eyes cannot detect,” said Dr Tran.
The AI model, a potential game-changer in healthcare diagnostics, was trained on high-resolution microscopy images of Salmonella typhimurium isolates exposed to varying levels of the antibiotic ciprofloxacin.
Remarkably, the algorithm accurately predicted the bacteria’s susceptibility or resistance in just six hours, a significant reduction compared to the conventional 24-hour culture and testing period.
Dr Sushmita Sridhar, formerly at the University of Cambridge and now at the University of New Mexico and Harvard School of Public Health, stressed the potential impact of this technology on clinical practice.
“While the current approach relies on single-cell resolution imaging and may not be universally deployable yet, it demonstrates the feasibility of rapidly predicting drug resistance with minimal parameters about bacterial structure,” said Dr. Sridhar.
The following research phase aims to expand the scope of the bacteria analysed, potentially extending the AI’s capabilities to identify resistance across multiple species and antibiotics.
Dr Sridhar added: “Ultimately, our goal is to develop methods that can directly assess susceptibility and resistance from complex clinical samples such as blood or urine, which would be transformative for clinical diagnostics.”
The study underscores the transformative potential of AI in combating antimicrobial resistance, offering a glimpse into a future where advanced technology, when harnessed effectively, can play a pivotal role in safeguarding public health.
Featured image: AI accurately predicted the Salmonella typhimurium bacteria’s susceptibility or resistance in just six hours. Credit: University of Cambridge
