About Workshop
This 3-day workshop introduces participants to the use of AI and Transformer-based models in spatial biology and digital pathology. Participants will learn how whole-slide images, tissue patches, spatial embeddings, gene expression data, and clinical outcomes can be integrated to study cancer tissue architecture, immune cell patterns, and patient risk prediction. The workshop includes practical Google Colab sessions using open-source tools such as PyTorch, HuggingFace Transformers, Squidpy, Scanpy, OpenSlide, Matplotlib, and Seaborn.
Aim
To provide practical understanding of how Vision Transformers and multi-modal AI models can be used to analyze tissue images, spatial genomics data, and cancer microenvironment patterns for research and clinical translation.
What Participants Will Learn
- Understand the role of AI in spatial biology and digital pathology.
- Learn how Whole-Slide Images are divided into patches for Transformer-based analysis.
- Generate attention maps to identify important tissue regions and immune cell patterns.
- Explore spatial transcriptomics and image-genomics data alignment.
- Apply Transformer models for gene marker prediction, tissue classification, and survival risk scoring.
- Create publication-ready visualizations such as heatmaps, UMAP plots, and Kaplan-Meier curves.
Structure
🗓️ Day 1: Preparing Tissue Topography for Transformers
- Understanding the shift from traditional pathology to AI-driven Spatial Biology.
- How Vision Transformers (ViTs) capture complex cell-to-cell environments where traditional CNNs fail.
- Step-by-step logic of breaking down Whole-Slide Images (WSIs) into patch sequences.
- Interpreting "Attention Maps" to identify hidden tumor-infiltrating immune cells.
💻 Hands-on Lab (Google Colab)
- Task: Convert a standard digital biopsy slide into spatial embeddings and extract feature sequences.
- Deliverable: Generate an interactive AI attention heatmap highlighting critical cellular regions.
🗓️ Day 2: Multi-Modal Fusion (Aligning Images with Genomics)
- Introduction to spatial transcriptomics data structures: mapping cellular morphology against gene expression profiles.
- Building Cross-Attention Transformer models to blend histology images with high-dimensional genetic data.
- Using Contrastive Learning frameworks to predict spatial gene expression directly from raw, low-cost H&E slides.
💻 Hands-on Lab (Google Colab)
- Task: Train a lightweight PyTorch Transformer layer to cross-reference tissue layouts with transcriptomic features.
- Deliverable: Map cross-modal tissue alignment and predict key cancer biomarkers from image inputs alone.
🗓️ Day 3: Clinical Translation & Cancer Microenvironment Profiling
- Mapping cellular "neighborhoods" and identifying immunosuppressive tumor boundaries.
- Utilizing Multiple Instance Learning (MIL) Transformers for whole-slide clinical classification.
- Translating AI-generated spatial biomarkers into publishable, peer-reviewed figures.
💻 Hands-on Lab (Google Colab)
- Task: Run a spatial transformer to process global WSI metrics for patient prognosis forecasting.
- Deliverable: Calculate patient survival risk scores and automatically generate a publication-quality Kaplan-Meier curve using Python.
Important Dates
Registration Ends
4:30 PM
Workshop Dates
2026-08-06
5:30 PM
5:30 PM
What You Will Gain
Outcomes
- Process digital pathology and whole-slide image data
- Generate spatial tissue embeddings from tissue image patches
- Interpret AI attention maps for identifying important tissue regions
- Align histology images with gene expression data
- Build basic multi-modal Transformer workflows
- Analyze cancer microenvironment and cellular neighborhood patterns
- Generate research-ready survival analysis and biomarker visualizations
Who Should Attend
- Students in biotechnology, bioinformatics, biomedical science, and related fields
- PhD scholars and researchers working in cancer biology, pathology, computational biology, or spatial biology
- Academicians and faculty members interested in AI-driven digital pathology and precision medicine
- Professionals from healthcare AI, biomedical research, diagnostics, and precision oncology domains
- Learners interested in Vision Transformers, spatial transcriptomics, and cancer microenvironment analysis