Skip to content

aidang3019/LungcancerAI

Repository files navigation

LungcancerAI

LungcancerAI is a personal research project exploring deep learning techniques for lung cancer staging and subtyping. The work focuses on experimenting with convolutional neural network (CNN) architectures using the NSCLC Radiomics dataset, with the goal of helping clinicians interpret imaging data more effectively. This work was presented at the North American Conference for Lung Cancer.

Highlights

  • Achieved 98.06% accuracy on the lung cancer staging task.
  • Achieved 99.6% accuracy on the lung cancer subtyping task.
  • Presented at the North American Conference on Lung Cancer.

Repository Structure

The project is organized as a collection of Jupyter notebooks that document the analysis, experiments, and findings:

Notebook Description
AI_lung_cancer_research.ipynb Baseline exploration of CNN-based approaches for lung cancer staging.
Stage_2_AI_lung_cancer_research.ipynb Follow-up experiments with refined training strategies and model tuning.
AI_stage_and_type_lung_cancer.ipynb Consolidated notebook covering both staging and subtyping tasks.

Dataset

The experiments use the NSCLC Radiomics dataset, which provides curated imaging features for non-small cell lung cancer cases. Ensure you have appropriate access to the dataset and update the notebook paths so they point to your local copy. Refer to the dataset's license and usage guidelines before incorporating it into your work.

Reproducing Results

Each notebook contains the preprocessing, training, and evaluation steps required to reproduce the reported accuracies. To replicate the findings:

  1. Prepare the dataset according to the instructions in the notebooks.
  2. Execute all cells sequentially, paying close attention to the configuration blocks for data paths and training hyperparameters.
  3. Review the final metrics printed or plotted in each notebook.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors