Phuong H. N. Dang is a PhD student from the AUT Institute of Biomedical Technologies (IBTec). He attended the 22nd International Conference on Multimedia Information Technology and Applications (MITA 2026) at the Auckland University of Technology (AUT) in Auckland, New Zealand, held from 21-24 July 2026. As a premier global conference in multimedia information technology, MITA unites academics and industry professionals to exchange the latest innovations in AI-driven media processing, multimedia architectures, smart IoT, and emerging media applications. (http://www.icmita.org/)
Supported by AUT Computer and Information Sciences Research Centre (CISRC), Phuong presented the paper “The Coexisting Confound in Speech-Based Clinical Detection: A Bootstrap Stability and Subgroup Analysis” in the MITACCIS-Springer track. This study proves that speech-based detection studies often wrongly treat depression and PTSD as separate conditions, despite their high co-occurrence with the E-DAIC dataset leads to flawed conclusions. First, the two conditions are so highly correlated (r = 0.850) that nearly all depressed patients also have PTSD, eliminating the possibility of a clean "depression-only" test group. Second, acoustic features are largely unreliable; out of 135, only two consistently signal depression, and none signal PTSD. Finally, while classifiers boast high overall accuracy, they are mostly just identifying healthy subjects (84.8%) while severely failing to detect actual clinical groups (accuracies ranging from 4.2% to 25.8%). The paper concludes that previous E-DAIC studies actually measure "healthy" detection rather than depression detection, and proposes four new practical standards for future research.