Documentation Index

Comprehensive documentation for the Computational Pathology Research Framework.

Getting Started

Installation and Setup

Getting Started Guide - Complete installation and setup guide
System Requirements - Hardware and software prerequisites
Installation - Step-by-step installation instructions

Tutorials

Your First Model - Step-by-step PCam tutorial
Working with Real Data - PatchCamelyon and CAMELYON16 workflows
API Reference - Complete API documentation

Training and Evaluation

PatchCamelyon (PCam)

PCAM_BENCHMARK_RESULTS.md - Benchmark results and analysis
PCAM_COMPARISON_GUIDE.md - Comparing model architectures
Topics: Patch-level classification, data augmentation, baseline comparisons

CAMELYON16 Slide-Level

CAMELYON_TRAINING_STATUS.md - Training guide and best practices
CAMELYON_SLIDE_LEVEL_IMPLEMENTATION.md - Implementation details
Topics: Slide-level aggregation, attention mechanisms, feature extraction

Evaluation Metrics

Model performance analysis
ROC curves and confusion matrices
CSV export for downstream analysis

Architecture and Design

System Overview

ARCHITECTURE.md - Complete system architecture
Components: Data loaders, model architectures, training loops, evaluation pipelines

Model Architectures

Baseline Models: ResNet, DenseNet, EfficientNet
Slide Classifiers: Attention-based aggregation, pooling strategies
Pretrained Integration: torchvision and timm model loading

Data Pipeline

PCam Dataset: Patch extraction and preprocessing
CAMELYON Dataset: HDF5 feature storage, slide-level batching
Augmentation: Standard transforms, normalization

Deployment

Docker Deployment

DOCKER.md - Complete Docker guide
Topics: Container setup, GPU support, production deployment

REST API

deploy/README.md - API deployment instructions
Endpoints: Model inference, batch processing, health checks

Model Export

ONNX Export: Cross-platform deployment
TorchScript: Production optimization
Quantization: Model compression

Development

Contributing

CONTRIBUTING.md - Contribution guidelines
Topics: Code style, pull requests, issue reporting

Testing

TESTING_SUMMARY.md - Test suite documentation
Coverage: 62% code coverage, 500+ unit tests
Topics: Unit tests, integration tests, property-based testing

Build System

MAKEFILE.md - Makefile usage guide
Commands: Build, test, lint, format, clean

Performance and Optimization

Performance Analysis

PERFORMANCE.md - Optimization guide
Topics: Profiling, bottleneck analysis, GPU utilization

Model Profiling

Inference latency measurement
Memory usage analysis
Throughput optimization

Results and Analysis

Benchmark Results

DEMO_RESULTS.md - Demo training results
Metrics: Accuracy, AUC, training time, convergence analysis

Project Portfolio

PORTFOLIO_SUMMARY.md - Complete project overview
Status: Current capabilities, limitations, future work

Roadmap

ROADMAP_TO_REAL_DATASETS.md - Real dataset integration plan
Timeline: Short-term goals, long-term vision

API Reference

Core Modules

Data Loading

from src.data import PatchCamelyonDataset, CAMELYONSlideDataset

PatchCamelyonDataset: Patch-level image loading
CAMELYONSlideDataset: Slide-level feature loading
collate_slide_bags: Variable-length batch collation

Models

from src.models import SimpleClassifier, SimpleSlideClassifier
from src.models.pretrained import load_pretrained_encoder

SimpleClassifier: Patch-level classifier
SimpleSlideClassifier: Slide-level aggregation
load_pretrained_encoder: Pretrained model loading

Training

from src.training import train_epoch, evaluate

train_epoch: Single epoch training loop
evaluate: Model evaluation with metrics

Utilities

from src.utils import set_seed, save_checkpoint, load_checkpoint

set_seed: Reproducibility utilities
save_checkpoint: Model checkpointing
load_checkpoint: Checkpoint loading

Quick Reference

Common Commands

Training

# PCam training
python experiments/train_pcam.py --config experiments/configs/pcam.yaml

# CAMELYON training
python experiments/train_camelyon.py --config experiments/configs/camelyon.yaml

Evaluation

# PCam evaluation
python experiments/evaluate_pcam.py \
  --checkpoint checkpoints/pcam/best_model.pth \
  --data-root data/pcam

# CAMELYON evaluation with CSV export
python experiments/evaluate_camelyon.py \
  --checkpoint checkpoints/camelyon/best_model.pth \
  --save-predictions-csv

Testing

# Run all tests
pytest tests/ -v

# Run with coverage
pytest tests/ --cov=src --cov-report=html

Model Profiling

# Profile inference time
python scripts/model_profiler.py \
  --checkpoint models/best_model.pth \
  --profile-type time

# Export to ONNX
python scripts/export_onnx.py \
  --checkpoint models/best_model.pth \
  --output models/model.onnx

Troubleshooting

Common Issues

CUDA Out of Memory

Reduce batch size in config file
Use gradient accumulation
Enable mixed precision training

Slow Training

Check GPU utilization with nvidia-smi
Increase number of data loader workers
Use faster data augmentation

Poor Model Performance

Verify data preprocessing
Check learning rate schedule
Increase training epochs
Try different architectures

Additional Resources

External Documentation

Research Papers

PatchCamelyon: Veeling et al. (2018) - “Rotation Equivariant CNNs for Digital Pathology”
CAMELYON16: Bejnordi et al. (2017) - “Diagnostic Assessment of Deep Learning Algorithms”

Community

Archived Documentation

Historical documentation and implementation notes are available in archive/.