Skip to content

DominikBeese/SSM4SPM

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

SSM4SPM

Multimodal Surgical Process Modeling With State Space Models
(Dominik Beese, 2025)

Computer-assisted intervention (CAI) uses AI to improve the accuracy, efficiency, and safety of surgical procedures, including pre-operative planning, intra-operative guidance, and post-operative analysis. A modern approach is to use a combination of convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to analyze surgical videos, providing a detailed understanding of phases, instruments and anatomical structures. Recent advances in state-space models (SSMs), such as the Mamba architecture, improve the efficiency of sequence modeling compared to Transformers by reducing computational complexity while maintaining performance. This work applies AI to analyze videos of cataract surgeries and compares CNN-Mamba architectures with other models such as LSTMs, GRUs and Transformers. The results highlight the potential of SSMs for the advancement of multimodal surgical video analysis.

Real-Time Phase Recognition

Content

The repository contains the following elements:

  • 📂 Data
    • 🗃 Phase, Instrument, Anatomy, and Irregularity annotations
    • 📂 Preparation code for downloading and preprocessing the dataset
    • 📂 Visualization code for generating plots
  • 📂 Code
    • 📂 Models
      • 📜 Mamba and Transformer implementation
      • 📜 CNN-RNN architecture implementation
    • 📂 Training
      • 📜 Callback, Loss, and Metric implementations
      • 📜 Grid search implementation
  • 📂 Experiments
    • 📜 Code for running the experiments
    • 📂 All experiments
      • 📂 Data splits
      • 🖼 Evaluation plots
      • 📜 Evaluation results
      • 🖼 Analysis plots

About

Data and code for my Master's thesis "Multimodal Surgical Process Modeling With State Space Models".

Resources

License

Stars

Watchers

Forks

Contributors

Languages