AI-Radiomics PET Diagnostic Efficacy for Alzheimer's Disease: A Systematic Review

A 2025 systematic review and meta-analysis synthesizes data from 9 studies comprising 5,765 participants to compare AI-assisted PET radiomics with conventional PET diagnosis for Alzheimer's disease. The analysis, conducted by Jinglu Duan, includes 25 2×2 tables and evaluates diagnostic performance across different patient strata and tracer types. Results indicate promising potential for distinguishing AD from mild cognitive impairment using proteinopathy PET.

Use Cases

• Benchmarking AI diagnostic models against conventional PET based on synthesized sensitivity and specificity metrics.
• Investigating the heterogeneity of study results in medical AI using the provided HSROC prediction region analysis.
• Comparing the performance of different PET tracers (e.g., proteinopathy vs. 18F-FDG) for early Alzheimer's detection.

Strengths

• Includes data from 9 studies and 5,765 participants, providing a substantial evidence base.
• Follows PRISMA guidelines and uses QUADAS-AI for bias assessment, indicating a structured review process.
• Presents specific diagnostic metrics (e.g., SE 0.89, SP 0.91, AUC 0.96 for proteinopathy PET in AD vs. HC).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The description notes substantial study heterogeneity and limited external validation.

Provenance

Source

Systematic review aggregating studies from PubMed, Embase, and Web of Science.

Collection Method

Literature review and bivariate meta-analysis performed per PRISMA guidelines.

Time Range

Studies searched through April 11, 2025.

Freshness

Last updated 2026-05-15 04:21:44; freshness should be verified.