🦠 ForamDeepSlice: AI-Powered Foraminifera Classification

Automated Fossil Species Identification from Micro-CT Scans Using Deep Learning

KVL Staff on Project

Abdelghafour Halimi
abdelghafour.halimi@kaust.edu.sa

Didier Barradas-Bautista
didier.barradas@kaust.edu.sa

Ronell Sicat
ronell.sicat@kaust.edu.sa

Building 1, Level 0, Office 0125

Collaborators

Ali Alibrahim
ali.alibrahim@kaust.edu.sa

Abdulkader M. Afifi
abdulkader.alafifi@kaust.edu.sa

Research Breakthrough

Foraminifera are microscopic organisms whose fossilized shells provide insight into the history of Earth's climate. Our ForamDeepSlice framework uses deep learning to automatically classify 12 foraminifera species from micro-CT scans with unprecedented accuracy.

  • 📊 Tested on 97 specimens representing 27 species, ForamDeepSlice achieved 95.6% accuracy and 99.6% top-3 accuracy across 109,617 2D slices.
  • 🎯 The method includes an interactive dashboard for real-time classification and 3D slice matching.
  • 🔬 This work establishes new benchmarks for AI-assisted micropaleontological identification.

🔬 Research Visualization The workflow diagram below illustrates our comprehensive deep learning pipeline for automated foraminifera classification. The paper entitled “ForamDeepSlice: A High-Accuracy Deep Learning Framework for Foraminifera Species Classification from 2D Micro-CT Slices” is available for download and review here.

foramdeepslice_workflow_v1.jpg

🎯 KVL's Contribution KVL's visualization scientists Abdelghafour Halimi, Didier Barradas-Bautista, and Ronell Sicat contributed significantly to the development of the ForamDeepSlice framework through:

  • Design and implementation of PatchEnsemble strategy for improved classification accuracy
  • Curation of a comprehensive micro-CT foram dataset with rigorous specimen-level splitting
  • Evaluation and benchmarking of 7 state-of-the-art CNN architectures
  • Development of an interactive dashboard with real-time classification and 3D slice matching
  • Optimization of preprocessing pipeline and data augmentation strategies
  • Public release of the ForamDeepSlice framework for scientific community use
,