AI tools aim to improve diagnosis of children's speech disorders

When Marisha Speights began her career as a speech-language pathologist in Nashville, Tennessee, she used standard screening and assessment methods. However, upon moving to Jackson, Mississippi, she found these methods less effective for preschools serving poorer families. "It was, 'I don't think this child has a speech or language issue, but the test says they're at risk.' And also the other way, of it not identifying children I thought were at risk," Speights said.

Speights took her concerns to Northwestern University and is now developing an artificial intelligence (AI) system aimed at addressing these issues. Her work is part of the Pediatric Speech Technologies and Acoustics Research Lab (PedzSTAR Lab), which uses acoustic biomarkers to analyze children's speech patterns. The goal is to use AI and machine learning to predict speech disorders.

So far, Speights has collected samples from 400 children of diverse backgrounds and aims to gather over 2,000 samples. "In the current dataset, we have many children that are not represented," she noted.

The PedzSTAR Lab's efforts are part of a broader trend in using AI within speech pathology. Jordan Green from Harvard University commented on the growing excitement around AI in healthcare for its potential applications such as virtual therapists and AI-driven diagnostics.

Nina Benway from the University of Maryland attributes increased AI usage to more data availability, accessible computing power, and mainstream large language models like ChatGPT. "It's been used in the field broadly most by clinicians...but the idea of using AI to assist with treatment has been relatively new," Benway said.

Speights emphasizes that collecting quality speech data from young children presents challenges due to their age-specific needs during assessments. Her approach involves engaging activities with toy farm animals followed by structured tasks.

The University at Buffalo is also exploring AI for diagnosing speech issues with a $20 million grant from the National Science Foundation (NSF). Venu Govindaraju from NSF's National AI Institute for Exceptional Education highlighted public interest in AI's potential: "Everyone knows someone who has children who are struggling or have struggled with some of their language."

Both Govindaraju and Speights clarify that AI will not replace human pathologists but rather assist them under professional supervision. In areas where professionals are scarce, Lauren Arner from the American Speech-Language-Hearing Association believes AI can help manage workloads effectively.

According to ASHA's 2024 survey, increasing diagnoses overwhelm available pathologists. Around 27% consider leaving due to burnout linked partly to insufficient pay and funding.

Speights sees automation as a means to reduce workload while focusing on personalized care for those needing it most. She stresses safeguarding children's data during collection processes at PedzSTAR Lab.

Arner mentions upcoming ASHA guidance on using AI safely in practice settings while Benway underscores validity, reliability, and representation as key considerations when implementing such technologies into clinical workflows: "It's likely AI will be most useful...when it's automating things clinicians already do."