A Hybrid Classical-Quantum Neural Network to Predict Future Lung Cancer Risk From a Single Low-Dose CT
Principal Investigator
Name
Weixin Lu
Degrees
M.sc
Institution
New York University
Position Title
PhD Student
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-1208
Initial CDAS Request Approval
Mar 25, 2024
Title
A Hybrid Classical-Quantum Neural Network to Predict Future Lung Cancer Risk From a Single Low-Dose CT
Summary
Low-dose Computed Tomography contains necessary information predictive of future lung cancer risk. Machine Learning models such as Sybil, PLCOm2012, LLP use classical neural networks to efficiently predict lung cancer risk in 6 years without additional demographic or clinical data. However, deep 3-D convolutional network contains large numbers of parameters and requires large computation in back propagation when training the models. Comparing with classical computing, quantum computing gives us a more powerful tool to solve complex problems from a quantum perspective. We propose a hybrid classical-quantum neural network based on Sybil and compare their accuracy and computational cost. We explore an essential and low-dimensional embedding of images for the input of quantum neural network layers.
Aims
1. A hybrid classical-quantum neural network with high accuracy and low computational cost comparing to classical neural networks.
2. Comparison between classical and quantum implementation at different medium layers.
3. A good image embedding network based on 3D ResNet.
4. Effective quantum error mitigation.
Collaborators
Javad Shabani, New York University