Review of Multimodal AI Systems for Major Depressive Disorder, 40 Studies

40 original studies published after 2015 review the performance of multimodal AI for Major Depressive Disorder. The document, authored by Claudio Crema and last updated in April 2026, analyzes clinical and translational applications using data from MRI, speech, clinical records, and wearables. Reported AI model accuracies are typically between 65% and 85%, but the review notes a systematic lack of external validation.

Use Cases

• Benchmarking AI model performance for MDD detection based on reported accuracy ranges of 65%-85%.
• Comparing the clinical feasibility of different data modalities (e.g., MRI vs. audio-visual features) discussed in the review.
• Identifying research gaps in external validation and stratified analyses for clinical translation as highlighted in the document.

Strengths

• Focuses on 40 original studies published after 2015, providing a recent snapshot of the field.
• Explicitly compares performance and feasibility across multiple data modalities like MRI, speech, and wearables.
• Released under a permissive CC-BY-4.0 license, allowing for broad reuse.

Limitations

• The dataset is a 10.6 KB DOCX review document, not a primary data collection, limiting direct analysis.
• Row count and column-level data are unknown, as the file contains textual analysis rather than structured data.
• The description metadata does not specify the geographic or institutional scope of the reviewed studies.

Provenance

Source

figshare, authored by Claudio Crema.

Collection Method

Literature review from four databases (PubMed, Web of Science, Scopus, Embase).

Time Range

Studies published after 2015.

Freshness

Last updated 2026-04-17 05:21:26.