Welcome to this repository showcasing my course project on artificial intelligence (AI). This project extensively explores the impact of various preprocessing techniques and configurations of artificial neural networks (specifically Convolutional Neural Networks, CNNs) on model performance in a medical setting.
Using a dataset comprising over 170,000 expert-annotated, de-identified cells derived from the bone marrow smears of 945 patients, this project aims to develop a proficient model for classifying bone marrow cells. The cells in the dataset were stained using the May-Grünwald-Giemsa/Pappenheim stain. Various preprocessing methods, such as normalization, standardization, and feature selection, were utilized to prepare the data for model training. 📊🔬💡
The goal of this project is to provide an exhaustive analysis of the effects of preprocessing techniques and CNN configurations on model performance. This project also aims to offer insights into best practices for these tasks in real-world applications. The results from this research should prove to be a valuable resource for both students and practitioners seeking to understand the impact of these factors on model performance, particularly in the context of bone marrow cell classification. 🏥💼🎓
This project has been implemented using Python, incorporating popular libraries such as Keras and Numpy. The code is thoroughly documented and organized into clear, modular functions for easy understanding and reuse. 📚🐍👩💻
I welcome any feedback or comments on this project. Thank you for your time and interest!
- Detailed Report with Work and Conclusions (PDF)
- CNN Model - Colab Notebook
- SVM Model - Colab Notebook
Matek, C., Krappe, S., Münzenmayer, C., Haferlach, T., & Marr, C. (2021). An Expert-Annotated Dataset of Bone Marrow Cytology in Hematologic Malignancies [Data set]. The Cancer Imaging Archive. https://doi.org/10.7937/TCIA.AXH3-T579
Matek, C., Krappe, S., Münzenmayer, C., Haferlach, T., and Marr, C. (2021). Highly accurate differentiation of bone marrow cell morphologies using deep neural networks on a large image dataset. https://doi.org/10.1182/blood.2020010568
Clark K, Vendt B, Smith K, Freymann J, Kirby J, Koppel P, Moore S, Phillips S, Maffitt D, Pringle M, Tarbox L, Prior F. The Cancer Imaging Archive (TCIA): Maintaining and Operating a Public Information Repository, Journal of Digital Imaging, Volume 26, Number 6, December, 2013, pp 1045-1057. DOI: 10.1007/s10278-013-9622-7