Aided by AI, new catheter design helps prevent bacterial infections

Bacteria are remarkable swimmers, a trait that can harm human health. One of the most common bacterial infections in a healthcare setting comes from bacteria entering the body through catheters, thin tubes inserted into the urinary tract. Although catheters are designed to suck fluids from a patient, bacteria are able to propel themselves upstream and into the body via catheter tubes using a unique swimming motion, causing $300 million worth of UTIs associated with catheters in the United States each year.

Now, an interdisciplinary Caltech project has designed a new type of catheter tube that hinders the upstream mobility of bacteria, without the need for antibiotics or other antimicrobial chemical methods. Thanks to the new design, powered by new artificial intelligence (AI) technology, the number of bacteria capable of swimming upstream in laboratory experiments has been reduced by 100 times.

The article titled “AI-assisted geometric design of anti-infection catheters” was published in the journal Scientists progress on January 3.

In catheter tubes, the fluid exhibits so-called Poiseuille flow, an effect where the movement of the fluid is faster in the center but slow near the wall, similar to flow in a river current, where the speed of the water varies from rapid in the center to slowing down near the banks. Bacteria, as self-propelled organisms, exhibit a unique “two steps forward along the wall, one step back in the middle” movement that produces their forward progression in tubular structures. Researchers from the Brady laboratory had already modeled this phenomenon.

“I once shared this intriguing phenomenon with Chiara Daraio, simply describing it as a ‘cool thing,’ and her response steered the conversation toward a practical application,” says Tingtao Edmond Zhou, a postdoctoral researcher in chemical engineering and co -first researcher. author of the study. “Chiara’s research often plays with all kinds of interesting geometries, and she suggested approaching this problem with simple geometries.”

Following this suggestion, the team designed tubes with triangular protrusions, like shark fins, along the inside of the tube walls. The simulations yielded promising results: these geometric structures effectively redirected the movement of the bacteria, propelling them toward the center of the tube where the faster flow pushed them downstream. The fin-like curvature of the triangles also generated vortices that further disrupted bacterial progression.

Zhou and his collaborators aimed to verify the design experimentally, but needed additional expertise in biology. For this, Zhou contacted Olivia Xuan Wan, a postdoctoral researcher in the Sternberg laboratory.

“I study nematode navigation and this project resonated deeply with my specialized interest in movement trajectories,” says Wan, who is also co-first author of the new paper. For years, the Sternberg laboratory has been conducting research on the navigation mechanisms of the nematode Caenorhabditis elegans, a soil organism the size of a grain of rice commonly studied in research laboratories and thus has numerous tools to observe and analyze the movements of microscopic organisms.

The team quickly moved from theoretical modeling to practical experimentation, using 3D printed catheter tubes and high-speed cameras to monitor bacterial progression. Tubes with triangular inclusions resulted in a reduction in upstream bacterial movement by two orders of magnitude (a 100-fold decrease).

The team then continued simulations to determine the most effective triangular obstacle shape to prevent bacteria from swimming upstream. They then fabricated microfluidic channels analogous to common catheter tubes with optimized triangular designs to observe the movement of E. coli bacteria under various flow conditions. The observed trajectories of E. coli in these microfluidic environments aligned almost perfectly with the simulated predictions.

The collaboration grew as researchers sought to continue to improve the geometric design of the tubes. Artificial intelligence experts from the Anandkumar Lab provided the project with cutting-edge AI methods called neural operators.

This technology was able to speed up catheter design optimization calculations so that they did not require days but minutes. The resulting model proposed changes to the geometric design, further optimizing the triangular shapes to prevent even more bacteria from swimming upstream. The final design improved the effectiveness of the initial triangular shapes by an additional 5% in simulations.

“Our journey from theory to simulation, experimentation and, finally, real-time monitoring within these microfluidic landscapes is a compelling demonstration of how theoretical concepts can come to life, offering tangible solutions to real-world challenges,” says Zhou.