Protocol for a systematic review and meta-analysis: Beyond Accuracy — Generalizability, Data Leakage, and Validation Reliability in AI-Based Seed Cultivar Identification (version 0.4)
Year: 2026
Type: article
Abstract: This record contains version 0.4 of the protocol for a systematic review and meta-analysis of validation practices in machine-learning-based identification of seed species, cultivars, genotypes and accessions from images. The review assesses validation independence (Seed-AI hierarchy V0–V4), data leakage and risk of bias in seven domains (D1–D7), reporting and reproducibility, within-study comparisons of split units, and internal-to-external performance change. Timing (stated for transparency): this protocol is deposited after the searches, screening and first-reviewer data extraction were completed, and before the independent second reviewer and the arbiter returned their assessments. Version 0.4 fixed the decision rules for the internal-to-external analysis (K14–K20) and added a proportion meta-analysis (RQ1) and a meta-regression (RQ2). These are dated protocol amendments, not a prospective pre-registration. The first reviewer (R1) is an AI system (Claude, Anthropic) supervised by the authors; the second reviewer and arbiter are human. Note: the date 2026-11-11 printed in the files is a clerical error; the files were deposited on 2026-11-10 (see record timestamp). Files: (1) English registration summary; (2) full protocol in Turkish.
Institution: Çanakkale Onsekiz Mart Üniversitesi
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7225730680
FWCI:
Cites:
Cited by:
Related to: