Enterprise AI Analysis
FORAMDEEPSLICE: A HIGH-ACCURACY DEEP LEARNING FRAMEWORK FOR FORAMINIFERA SPECIES CLASSIFICATION FROM 2D MICRO-CT SLICES
Abdelghafour Halimi, Ali Alibrahim, Didier Barradas-Bautista, Ronell Sicat, Abdulkader M. Afifi
This study presents a comprehensive deep learning pipeline for the automated classification of foraminifera species using 2D micro-CT slices derived from 3D scans. We curated a scientifically rigorous dataset of 97 micro-CT scanned specimens spanning 27 species, from which we selected 12 representative species with sufficient specimen counts (at least four 3D models each) for robust classification. To ensure methodological integrity and prevent data leakage, we employed specimen-level data splitting, resulting in 109,617 high-quality 2D slices (44,103 for training, 14,046 for validation, and 51,468 for testing). We evaluated seven state-of-the-art 2D convolutional neural network (CNN) architectures using transfer learning. Our final ensemble model, ForamDeepSlice (FDS), combining ConvNeXt-Large and EfficientNetV2-Small, achieved a test accuracy of 95.64%, with a top-3 accuracy of 99.6% and an area under the ROC curve (AUC) of 0.998 across all species. To facilitate practical deployment, we developed an interactive advanced dashboard that supports real-time slice classification and 3D slice matching using advanced similarity metrics, including SSIM, NCC, and the Dice coefficient. This work establishes new benchmarks for AI-assisted micropaleontological identification and provides a fully reproducible framework for foraminifera classification research, bridging the gap between deep learning and applied geosciences.
Executive Impact Summary
Our analysis of 'FORAMDEEPSLICE' reveals groundbreaking advancements in automated micropaleontological identification, setting new industry benchmarks for accuracy and operational efficiency. The integration of advanced deep learning with a practical, interactive dashboard promises significant returns on investment for geosciences and research.
0
Test Accuracy (Top-1)
0
Top-3 Accuracy
0
Area Under ROC Curve (AUC)
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
ForamDeepSlice: Automated Foraminifera Classification Pipeline
Our AI-driven workflow for foraminifera classification from micro-CT slices follows a robust, multi-stage process, ensuring data integrity, model performance, and practical deployment.
Enterprise Process Flow
Dataset Scoping & Prep
→
Rigorous Dataset Creation
→
Transfer Learning Pipeline
→
Interactive Dashboard
→
Scalable AI Deployment
ForamDeepSlice Achieves Benchmark Accuracy
The ForamDeepSlice (FDS) model demonstrated exceptional performance, achieving a 95.64% test accuracy, setting a new benchmark for AI-assisted micropaleontological identification from 2D micro-CT slices.
95.64%
Test Accuracy (Top-1) with ForamDeepSlice (FDS)
ForamDeepSlice Outperforms Individual Models
Comparing ForamDeepSlice (FDS) against leading CNN architectures highlights its superior overall performance due to the confidence-gated ensemble strategy.
| Model | Accuracy | F1-Score | Top-3 Acc | AUC |
|---|---|---|---|---|
| FDS (Ours) | 0.956 | 0.950 | 0.996 | 0.998 |
| ConvNeXt-Large | 0.951 | 0.941 | 0.996 | 0.998 |
| EfficientNetV2-S | 0.918 | 0.906 | 0.995 | 0.996 |
| NASNet | 0.937 | 0.925 | 0.992 | 0.996 |
| ResNet101V2 | 0.843 | 0.808 | 0.974 | 0.984 |
Targeted Ensemble for Difficult Species: Baculogypsina & Orbitoides
ForamDeepSlice's PatchEnsemble strategy specifically addressed persistent misclassification issues for Baculogypsina and Orbitoides. By conditionally switching to a secondary model (EfficientNetV2-Small) for these 'weak' classes, FDS significantly improved recall for Baculogypsina (0.603 vs. 0.243 in Top-2) and precision for Orbitoides (0.827 vs. 0.684 in Top-2), overcoming morphological ambiguities and extraction-induced damage. This targeted approach prevents dilution of accuracy for well-classified species.
Problem: Morphological Ambiguities
Baculogypsina exhibits highly variable cross-sections due to its asymmetric test, making it difficult for models to consistently classify. Orbitoides often suffer from extraction-induced damage, introducing fractures and missing fragments that degrade classification accuracy.
Solution: PatchEnsemble Strategy
FDS employs a confidence-gated model-switching mechanism. When the primary model is uncertain about 'weak' classes like Baculogypsina or Orbitoides, it defers to a specialized 'patch' model (EfficientNetV2-Small), which has demonstrated higher recall for these challenging taxa.
Impact: Enhanced Accuracy for Difficult Cases
- Baculogypsina: Recall improved from 0.243 (Top-2 ensemble) to 0.603 with FDS.
- Orbitoides: Precision improved from 0.684 (Top-2 ensemble) to 0.827 with FDS.
Advanced ROI Calculator
Estimate the potential return on investment for integrating AI-powered classification into your paleontological workflows. Tailor the inputs to your operational scale and see the projected annual savings and reclaimed expert hours.
Projected Annual Savings
$0
Expert Hours Reclaimed Annually
0
Your AI Implementation Roadmap
Implementing AI solutions like ForamDeepSlice requires a structured approach. Our tailored roadmap ensures a smooth transition, from initial data integration to full-scale deployment and continuous optimization within your enterprise.
Phase 01: Discovery & Strategy Alignment
Comprehensive assessment of existing paleontological workflows, data infrastructure, and identification of key classification challenges. Define success metrics and a clear AI integration strategy.
Phase 02: Data Integration & Custom Model Training
Securely integrate your micro-CT datasets and refine ForamDeepSlice models with transfer learning. Establish robust data governance and leakage prevention protocols.
Phase 03: Interactive Dashboard Deployment & User Training
Deploy the ForamDeepSlice interactive dashboard within your environment. Conduct hands-on training for paleontologists and researchers, ensuring proficiency in real-time classification and 3D matching.
Phase 04: Continuous Optimization & Scalable Integration
Establish feedback loops for model refinement and explore integration with existing geological and biostratigraphic analysis platforms. Scale the solution across multiple projects and data modalities.
Ready to Transform Your Geosciences Research?
Leverage the power of AI to accelerate fossil identification, enhance research accuracy, and free up your experts for higher-value tasks. Book a free consultation with our AI specialists to discuss how ForamDeepSlice can be integrated into your operations.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat