Expresso-AI: Explainable Video-Based Deep Learning Models for Depression Diagnosis

2026-06-24 · Source: Takara TLDR - Daily AI Papers · Field: Health & Wellbeing — Mental Health & Psychological Support, Medical Devices & Health Technology · Depth: Advanced, medium

Summary

Expresso-AI introduces a novel framework for explainable video-based deep learning models designed for automatic depression severity diagnosis. This system fine-tunes Deep Convolutional Neural Networks (DCNNs), initially pre-trained on Action Recognition datasets, using facial videos from the AVEC depression dataset. The framework interprets the DCNN's decisions by analyzing saliency maps, focusing on specific face regions and temporal expression semantics to generate both visual and quantitative explanations. This approach provides greater insight into the model's reasoning, addressing a critical gap in affect-specificity and interpretability in current automated depression diagnosis methods. Furthermore, Expresso-AI has demonstrated improved predictive performance compared to previous single-face benchmarks for visual depression diagnosis, successfully developing a framework that generates hypotheses from facial model decisions while enhancing predictive capabilities.

Key takeaway

For AI Scientists developing diagnostic tools for mental health, Expresso-AI demonstrates that integrating explainability is crucial for clinical adoption. You should prioritize frameworks that interpret deep model decisions, such as analyzing saliency maps from facial videos, to provide transparent reasoning. This approach not only enhances trust with healthcare professionals but also improves predictive performance over single-face benchmarks, guiding your development towards more effective and clinically relevant diagnostic systems.

Key insights

Expresso-AI provides explainable deep learning for depression diagnosis by interpreting DCNN saliency maps from facial videos, improving predictive performance.

Principles

• Interpretable models enhance clinical utility.
• Temporal video analysis improves diagnosis.
• Saliency maps reveal model reasoning.

Method

Fine-tune DCNNs pre-trained on Action Recognition datasets using AVEC depression facial videos. Interpret model saliency maps by examining face regions and temporal expression semantics to generate visual and quantitative explanations.

In practice

• Enable early depression diagnosis.
• Support mental health treatment planning.
• Provide objective diagnostic measures.

Topics

• Depression Diagnosis
• Explainable AI
• Video Analysis
• Deep Learning
• Facial Expression Recognition
• Saliency Maps

Code references

• Lingxi-mental-health/LingxiDiagBench

Best for: Computer Vision Engineer, AI Scientist, Research Scientist, Domain Expert

Related on AIssential

• SketchXplain: Intuitive Visual Explanations of Image Classifiers with Sketches — SketchXplain generates intuitive, sketch-based visual explanations for image classifiers by combining saliency, concept models, and sketch optimization.
• Explainable AI based ensemble model for the identification of Schizophrenia prodromal phase — An XAI-integrated ensemble ML model accurately predicts Schizophrenia prodromal symptoms for early clinical intervention.
• Translating black-box medical AI models into interpretable global decision logic — Class-Association Manifold Learning (CAML) explains black-box medical AI by visualizing global decision rules on low-dimensional manifolds.
• Causal Explanations for Image Classifiers — `rex` provides principled, causality-based black-box explanations for image classifiers, outperforming existing tools in efficiency and explanation size.
• Explaining AI Without Code: A User Study on Explainable AI — Integrating XAI into no-code ML platforms improves usability and trust, but must balance novice accessibility with expert diagnostic depth.
• Interpretable facial dynamics as behavioral and perceptual traces of deepfakes — Deepfakes exhibit measurable behavioral fingerprints, especially during emotional expression, detectable through interpretable facial dynamics.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.