MAD-UV: The 1st INTERSPEECH Mice Autism Detection via Ultrasound Vocalization Challenge

2025-01-08

AI Summary

Current mouse models of autism spectrum disorder (ASD) lack objective, quantitative tools for vocal phenotype identification. Method: We introduce the first benchmark dataset for automated ASD classification based on ultrasonic vocalizations (USVs) in mice and pioneer the application of speech signal processing techniques to neurodevelopmental disease animal models. We propose a multi-band analytical framework integrating both ultrasonic and audible-frequency components, and design a convolutional neural network (CNN) classifier that fuses three types of time-frequency spectrogram features—including those from the audible band—to support both segment-level and individual-level classification. Results: Audible-band features yield the highest discriminative performance, achieving unweighted average recall (UAR) of 0.600 (segment-level) and 0.625 (individual-level). These results validate the feasibility of automated ASD detection via vocal phenotypes and establish a novel interdisciplinary methodology and benchmark platform for biomarker discovery in translational neuroscience.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Natural Language Processing: Speech
• Machine Learning: Neuro-Symbolic Learning

Application Category

• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

The Mice Autism Detection via Ultrasound Vocalization (MAD-UV) Challenge introduces the first INTERSPEECH challenge focused on detecting autism spectrum disorder (ASD) in mice through their vocalizations. Participants are tasked with developing models to automatically classify mice as either wild-type or ASD models based on recordings with a high sampling rate. Our baseline system employs a simple CNN-based classification using three different spectrogram features. Results demonstrate the feasibility of automated ASD detection, with the considered audible-range features achieving the best performance (UAR of 0.600 for segment-level and 0.625 for subject-level classification). This challenge bridges speech technology and biomedical research, offering opportunities to advance our understanding of ASD models through machine learning approaches. The findings suggest promising directions for vocalization analysis and highlight the potential value of audible and ultrasound vocalizations in ASD detection.

Problem

• Autism Detection
• Ultrasound Vocalizations
• Mouse Model

Innovation

• Ultrasound Vocalization Recognition
• Autism Detection
• Convolutional Neural Network (CNN)