Student-built app slashes weeks off brain mapping to speed up neuroscience research

Alumnus Alec Soronow shares AI-powered program ‘Bell Jar’ with peers for free on GitHub

Cross-sections of mouse brain
The top horizontal panel displays mouse brain section examples from the Allen Institute for Brain Science's reference atlas. The bottom panel shows the results of Bell Jar image registration.
Group photo of project team members
The Bell Jar project team at UC Santa Cruz: from left to right are postdoc Richard Dickson, graduate student Matthew Jacobs, Alec Soronow, and Euiseok Kim. (Photo by Carolyn Lagatutta)

UC Santa Cruz neuroscientists aiming to better understand how specific brain connectivity contributes to perception, thoughts, and behavior are leveraging artificial intelligence to enhance their study of brain function. By integrating AI, they are streamlining the process of aligning thin slices of mouse brain tissue with a reference atlas, helping to identify key details such as the brain region of origin more efficiently.

The cutting-edge technology was developed by Alec Soronow while he was a student at UC Santa Cruz. He began the project as an undergraduate in the lab of Euiseok Kim, an assistant professor of molecular, cell, and developmental biology, and he continued to work with Kim until the digital tool was built and ready for use.

Soronow named the desktop application Bell Jar, after the semi-autobiographical novel by Sylvia Plath about a young woman's mental breakdown and recovery, which Soronow was reading while working on the initial version of the software. He and Kim have shared Bell Jar with their peers through an article published in the open-access journal eNeuro last month.

In academia, undergraduates occasionally see research projects through from inception to a notable publication that contributes to the broader scientific community. But it is nonetheless inspiring when that does happen, Kim said. “I am a biologist. I don’t know how to code, but I understand the problems in the field,” Kim explained. “Alec, on the other hand, was fearless in addressing these challenges, communicating effectively with both myself and other senior lab members. His work has made a significant impact.”

Addressing a long-standing challenge

Bell Jar offers a valuable tool for neuroscientists, making neuroanatomy analysis more accessible and efficient. For years, researchers have struggled with the limitations of existing tools designed to analyze neural connectivity—many of them developed in MATLAB or other specialized software environments that eventually became outdated or inaccessible due to software updates.

Moreover, previous methods faced challenges in effectively integrating machine learning (ML), restricting their flexibility and accuracy. Bell Jar differentiates itself by leveraging ML techniques to improve accuracy and efficiency. Unlike many predecessors, it is also designed to be highly user-friendly and openly accessible. By sharing the code on Soronow’s GitHub, the team has ensured that researchers worldwide can contribute to its development, customize it for their own projects, and continuously refine its capabilities.

Tedium of traditional brain mapping

Bell Jar was born of necessity. In the Kim Neuroscience Lab’s ongoing research into brain connectivity, analyzing an entire mouse brain manually was at times tedious. To prepare brain samples for analysis, researchers perform a process known as histology. This involves slicing the brain into incredibly thin sections, much like a deli slicer cuts thin slices of meat. However, human error is inevitable: Sections may be slightly too thick, too thin, or damaged in some way. When analyzing these slices, researchers must then compare them to a reference atlas, often making subjective decisions about how to align and interpret the data.

Bell Jar helps solve these challenges by using ML to detect and match neurons across brain sections. “In the past, we had to rely heavily on human intervention, which introduced subjectivity,” Kim said. “Now, Bell Jar can make these determinations more efficiently and with greater accuracy.”

To fully appreciate the importance of Bell Jar, Kim said it is essential to understand the painstaking nature of traditional brain mapping. For a single research project leading to a publication, researchers must process over 100 thin sections from just one brain. Within each of these sections, multiple brain regions need to be identified and analyzed. This process must be repeated across numerous samples, making it an incredibly time-intensive task.

Training new students to perform these analyses can also be difficult, as it requires a deep understanding of neuroanatomy. But with Bell Jar, many of these burdens are lightened. “It enables far more efficient analysis of large experimental datasets that would otherwise have to be looked over extensively by hand,” Soronow said. “Also, reducing the time to results lets a lab more quickly assess if the experiments being conducted are working and if adjustments are needed—enabling higher throughput science.”

To drive this point home, Soronow said, “For a standard experiment in our lab, it saves us about three weeks of manual alignment and counting time per brain.”

Personal connection to brain circuitry

Soronow’s grandmother suffered from dementia and often took care of him when he was young. So his curiosity about the malfunction of the brain’s circuits started at an early age. It was when he took his first developmental biology class in college that it clicked. The quarter after he took that class, Soronow attended a seminar by Euiseok that spoke to his burning desire to better understand the brain.

Soronow won a Dean’s Award from the Baskin School of Engineering in 2022 for the Bell Jar project and graduated that year with a B.S. in biomolecular engineering and bioinformatics, then earned an M.S. in molecular, cell, and developmental biology in 2024. He has since started his own company, PlusTen Intelligence, which is trying to bring the power of neurons to industrial applications by combining AI and engineered neuron-like cells to be powerful sensors for specific molecular targets that fit in a handheld device.

“I’m able to use the unique skill set cultivated in the Kim lab,” Soronow said, “combining wet lab, genetic engineering, innovative software design, and hardware design.”

Kim said the creation of Bell Jar aligns perfectly with the broader research objectives of his lab. By improving efficiency and accuracy, researchers can focus on the bigger picture: understanding brain connectivity and neural function. “This tool allows us to conduct research more quickly and effectively,” Kim said. “That is the main point here.”