Swathi Kiran's team at Boston University's Center for Brain Recovery is using artificial intelligence to answer a question that has long stumped neurologists: when a bilingual person suffers a stroke, which language should they focus on during recovery? The answer, it turns out, may not be the one they want to prioritize—but a personalized "digital twin" can now predict the path that works best for their brain.
About 2 million Americans live with aphasia, a communication disorder that typically strikes after a stroke or brain injury, disrupting the ability to speak, read, or understand language. For those who speak multiple languages, recovery becomes exponentially harder. Unlike monolingual patients, bilingual people with aphasia face a unique neuroscience problem: their languages share resources in the brain while maintaining separate control mechanisms, meaning a stroke's damage reverberates across both linguistic systems simultaneously rather than affecting just one.
Traditionally, clinicians made recovery decisions based on guesswork or convenience. Some asked patients which language they preferred and provided therapy in that language regardless of whether it was optimal. Others simply defaulted to English, whether or not the patient actually used it most frequently. "Before this study, clinicians usually decided which language a bilingual aphasia patient should focus on by asking the patient what language they wanted to focus on and providing therapy in that language whether or not that was an optimal language, or if they did not speak the languages the patient spoke, simply providing therapy in English, whether or not that was an optimal language," Kiran explained in a recent Q&A.
That approach missed a crucial opportunity. Kiran and her colleagues published research in npj Digital Medicine describing an AI model called BiLex that changes the equation entirely. Rather than guessing, BiLex creates a computer-based simulation of each patient's individual language system—a "digital twin" trained on their lifetime language history and tailored to their specific brain damage after stroke. Researchers can then safely experiment on this digital twin, simulating different therapies to see which language pathway offers the best recovery prospects.
What makes BiLex fundamentally different from the AI tools most people encounter online is that it doesn't simply recognize patterns or generate plausible-sounding answers. Instead, it models the neurobiology of how language actually breaks down and recovers in the brain, with separate systems for each language and shared meaning representations that mirror how a bilingual brain actually works. "It can be 'experimented on' safely—researchers can simulate brain damage and try different therapies to see what works best for that specific patient," Kiran noted.
The implications ripple beyond the laboratory. For patients facing months or years of speech therapy, the difference between focusing on the right language and the wrong one could mean the difference between meaningful recovery and months of effort yielding limited results. And as June is recognized as National Aphasia Awareness Month, the work underscores how much remains unknown about a condition affecting millions—and how technology, grounded in neuroscience, can illuminate paths forward.
Kiran's team publishes more than 30 papers annually on brain plasticity, language recovery, and bilingualism. With BiLex, they've moved beyond asking patients what they want and begun asking their brains what they actually need.