Multimodal AI model can guide personalized TB treatments

A team of researchers from the University of Michigan developed a multimodal AI model to predict treatment outcomes for tuberculosis (TB) patients. Their analysis of data from patients around the world could lead to personalization of tuberculosis treatment.

“Tuberculosis is the deadliest infectious disease in the world, affecting millions of people each year. My lab’s goal is to develop innovative new solutions to stop the spread of drug-resistant pathogens like those that cause tuberculosis. tuberculosis,” said Sriram Chandrasekaran, corresponding and associate author. professor of biomedical engineering.

In the study, published in iScience, researchers analyzed multimodal data, including various biomedical data from clinical tests, genomics, medical imaging, and drug prescriptions from TB patients. By analyzing data from patients with varying levels of drug resistance, they discovered biomedical characteristics predictive of treatment failure. They also discovered effective treatment regimens against specific groups of patients with drug-resistant tuberculosis.

“Our multimodal AI model accurately predicted treatment prognosis and outperformed existing models that focus on a narrow set of clinical data,” Chandrasekaran said.

“We identified effective treatment regimens against certain types of drug-resistant TB in all countries, which is very important due to the spread of drug-resistant TB,” added Awanti Sambarey, first author and postdoctoral researcher.

Using AI, the team examined more than 5,000 patients. “We are talking about real data, so patients in different countries have different admission protocols,” she said.

“We worked with more than 200 biomedical characteristics in our analysis; we looked at demographic information such as age and gender as well as treatment history. We also noted whether patients had other comorbidities, such as HIV, and then we worked with multiple image features such as their x-rays, CT scans, pathogen data, drug resistance data, as well as the genomic features and mutations of the pathogen.

“It’s really difficult, clinically, to look at all the data as a whole,” Sambarey said. “Typically, you look at them separately. I think that’s where AI comes in handy. When clinicians look at all that data, it can be overwhelming. Here, our research is able to identify the characteristics most significant clinics.”

The team also studied the impact of the type of drug resistance present. “You can look at a specific snapshot of data, such as genomic features, discover mutations in the infectious pathogen, and ask what some of the long-term treatment implications are,” she added.

Surprisingly, they found that certain drug combinations worked better in patients with certain types of resistance but not others, leading to treatment failure.

Researchers also found that drugs with antagonistic pharmaceutical interactions could lead to worse outcomes. “Using AI to eliminate antagonistic drugs early in the drug discovery process can prevent treatment failure down the line,” Chandrasekaran noted. “The AI ​​model can also potentially be adapted to identify treatment regimens suitable for people with certain comorbidities.”

“Instead of a one-size-fits-all treatment approach, we hope that studying multimodal data will help doctors treat patients with more personalized treatments to achieve the best results,” Sambarey said.