Researchers at the University of Cambridge are utilizing artificial intelligence (AI) tools in the objective to combat antibiotic resistance. Led by Professor Stephen Baker, the team developed a machine learning tool that can distinguish resistant bacteria from susceptible ones. The tool uses microscopy images to identify bacteria that are resistant to common antibiotics such as ciprofloxacin. The technique promises to significantly reduce diagnosis time for antibiotic resistance and consequently, transform treatment plans for deadly infections like typhoid fever.
The study, detailed in Nature Communications, was primarily focused on Salmonella Typhimurium, which leads to severe gastrointestinal illnesses and potentially fatal invasive diseases. Salmonella, most commonly contracted by humans through contaminated food, possesses certain strains resistant to antibiotics. Dr. Tuan-Anh Tran explains that the machine learning model was able to identify resistant bacteria through certain unique features in the microscopy images, a detail unnoticeable to the human eye.
The research process encompassed numerous steps, such as sample preparation which included growing S. Typhimurium samples in a liquid nutrient media, with some exposed to varying concentrations of ciprofloxacin and others not. The team used a sophisticated microscope to take detailed pictures of the bacteria at multiple time intervals, after which the analysis was performed with specialized software to gather and assess 65 different features from each bacterial cell.
The data was then utilized in the development of the machine learning model to recognize patterns related to antibiotic resistance. The Cambridge team observed that antibiotic-resistant bacteria could be correctly identified about 87% of the time. Furthermore, the study discovered that even when not exposed to antibiotics, resistant bacteria showed specific visual patterns completely distinct from susceptible bacteria- suggesting that antibiotic resistance changes the bacteria’s appearance subtly (though notably to AI detection).
Currently, common and prevailing methods demand several days for bacterial culture and testing against specific antimicrobials. This new AI-based solution would likely provide results within hours, enabling faster prescription of effective antibiotics, thereby bettering patient outcomes and minimizing the spread of resistant bacteria.
The future objectives of the research team involve applying their AI methodology to more complex clinical samples such as blood or urine, and also on different types of bacteria and antibiotics. Their goal includes making this technology more accessible to hospitals and clinics worldwide. The effort is indeed pivotal given the increasing threat of antibiotic resistance. Recognizing this, similar AI initiatives in antibiotic research are being developed in institutions like MIT, heralding a broader trend of AI-driven innovation in the field. The consistent and effective application of the new AI tool in real-world clinical scenarios will be the defining success of this breakthrough.
