AI evidence extraction

AI-Augmented Systematic Review of Remote Sensing and Predictive Modelling for Mycotoxin Risk Monitoring in Cereal Crops Across Central and Balkan Europe

László Radócz, Attila Nagy, Nikolett Szöllősi, Nikolett Éva Kiss, Andrea Szabó, János Tamás et al. · Remote Sensing · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10

Relevance

0/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.3390/rs18132063

Methodology & findings

Study design

AI-augmented systematic review using a four-stage automated pipeline comprising PICO domain scoring, SBERT semantic deduplication, Thompson-sampling reinforcement learning applied to 36,038 corpus records (2010–2025), yielding 156 included studies with inter-rater κ = 0.81 (95% CI: 0.74–0.88).

Main result

The study found that "satellite multispectral imaging dominated the literature (91.7% of studies); random forest and gradient boosting models achieved R2 = 0.74–0.80 for aflatoxin B1 and deoxynivalenol prediction in CBE maize and wheat when integrating vegetation indices, land surface temperature, and precipitation covariates." Additionally, "deep learning surpassed classical ML in annual study count from 2021, reaching ~60% relative share by 2025, though the performance advantage narrows at field scale relative to laboratory hyperspectral benchmarks (98–99% accuracy)."

Research paradigm

Empirical/Positivist (quantitative synthesis of computational and remote sensing studies)

Author conclusions

The authors conclude that "Climate-driven satellite mycotoxin prediction emerges as the field's active research frontier." The systematic review demonstrates that satellite multispectral imaging with machine learning models shows strong predictive performance for mycotoxin risk assessment in Central and Balkan European cereals, with deep learning methods increasingly dominant, though substantial geographic and toxin-specific research gaps remain.

Risk of bias

Publication bias (systematic reviews typically favor published, positive results); Language bias (search limited to 2010-2025 corpus); Geographic bias (limited coverage for four CBE nations); Selection bias related to database indexing and search strategy; Variability in study quality across 156 included studies; Publication bias (preference for successful machine learning applications); Geographic bias (zero studies for four CBE nations suggests regional representation gaps); Toxin-specific bias (T-2/HT-2 toxins underrepresented in Balkan states); Technology bias (multispectral imaging overrepresented at 91.7%); Sample size heterogeneity affecting performance comparisons; Potential automation pipeline bias from SBERT semantic deduplication and Thompson-sampling reinforcement learning; Selection bias: Zero eligible studies for four CBE nations and for T-2/HT-2 toxins across Balkan states; Geographic representation bias: Five-percentage-point CBE–global performance gap; Study design scope: Multi-toxin design differences affecting performance comparisons; Sample size heterogeneity across included studies

Limitations

The authors note that "A five-percentage-point CBE–global performance gap is largely consistent with differences in sample size and multi-toxin design scope rather than algorithmic access." Additionally, "The country × mycotoxin gap matrix identifies zero eligible studies for four CBE nations and for T-2/HT-2 toxins across the Balkan states," indicating significant geographic and toxin-specific knowledge gaps in the literature.

Open questions raised

Zero eligible studies for four CBE nations and for T-2/HT-2 toxins across the Balkan states. Limited field-scale validation relative to laboratory benchmarks. Need for climate-driven satellite mycotoxin prediction development in the CBE region.
Zero eligible studies for four CBE nations
Complete absence of T-2/HT-2 toxin studies across Balkan states
Field-scale validation needed for deep learning models versus laboratory hyperspectral benchmarks
Geographic coverage gaps in Central and Balkan European region
Need for climate change integration in satellite-based prediction models

Data: not_statedCode: not_statedExtracted from: pdfAgreement 54%

Explore related topics

Systematic Review Automation Literature Discovery