Skip to content

Pulmonary Fibrosis Prognosis Prediction utilising Quantum Machine Learning - Variational Circuits

License

Notifications You must be signed in to change notification settings

nuvinga/fibro-quannet

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Fibro-QuanNet: Pulmonary Fibrosis Prognosis Prediction using Quantum Machine Learning 🫁

Research 🔬 project by Nuvin Godakanda Arachchi

This research project has been conducted as partial completion of my BEng (Hons) in Software Engineering degree program for the Final Project dual-credit module.

The codebase of the Minimum Viable Product developed can be found in the Fibro-QuanNet - Codebase repository.

Abstract

Pulmonary fibrosis is a progressive lung condition caused by damaged or scarred lung tissue obstructing the exchange of carbon dioxide and oxygen gasses in the alveoli Thereby, leaving the body deprived of the oxygen required for blood oxygenation and less lung volume. As per state-of-the-art medical practice, the deterioration/ scarring of the lung tissue is not entirely reversible or correctable, merely leaving patients with symptom management using therapy and clinical drug trials. An accurate judgment of the lung function decline is crucial for the management and trial treatment of the patient.

This research project endeavours to automate the process of prognosis prediction of pulmonary fibrosis using a hybrid-classical quantile regression hybrid model built using a variational quantum circuit. Although quantum computing is still in its formative years, research activities done in similar domains have proved to have immaculate success in both the correctness and speed of the results. The project explores the advantages one might gain by utilizing the development of quantum computing over classical computational approaches, which will facilitate and encourage more optimization of machine learning using quantum computing.

The model has shown promising results so far, with a Laplace Log Likelihood matrix of -7.13, and a mean absolute error of just 212.31. For a regression model trained with a small dataset such as the OSIC dataset with just 700+ DICOM images with its metadata, the evaluations are noticeable and promising.

File Structure

  • 'Project-Proposal'
  • The project proposal documentation includes the initial set of documentation for the project kick-off.
  • 'PSPD' - Project Specification
  • The requirement gathering stage by identifying all potential stakeholders of the system had been documented in this report. The initial design and implementation of the core functionality of the Fibro-QuanNet have also been documented.
  • 'Thesis'
  • The final documentation of the project with a thorough critical analysis of the existing literature in the domains. The results of the models developed have also been discussed in detail.
  • Presentation
  • The slide deck used for the demonstration of the project during the viva-voce of the research module.

Data

The dataset utilized in this project can be accessed through the Open-Source Imaging Consortium.

Research Usage Disclaimer

All proceedings of this research project have been submitted to both DOI and the Turn-it-in global referencing platform, where direct utilization of the project or any usage of the content without necessary citations may be identified as plagiarism and an academic offence. Citations can be made to either the code/ documents repositories or the ResearchGate DOI publication.