- Title Information
- Title
- Using Convolutional Neural Network for Early Breast Cancer Detection
- Type of Resource (primo)
- dissertations
- Name:
Personal
- Name Part
- Rusnak, Anna
- Role
- Role Term:
Text
- creator
- Name:
Personal
- Name Part
- Srivastava, Vikas
- Role
- Role Term:
Text
- Advisor
- Name:
Personal
- Name Part
- Dawson, Michelle
- Role
- Role Term:
Text
- Reader
- Name:
Personal
- Name Part
- Tripathi, Anubhav
- Role
- Role Term:
Text
- Reader
- Name:
Corporate
- Name Part
- Brown University. Biology and Medicine: Biomedical Engineering
- Role
- Role Term:
Text
- sponsor
- Origin Information
- Copyright Date
- 2022
- Physical Description
- Extent
- iii, 38 p.
- digitalOrigin
- born digital
- Note:
thesis
- Thesis (Sc. M.)--Brown University, 2022
- Genre (aat)
- theses
- Abstract
- Objective: The objective of this study was to create a convolutional neural network
(CNN) capable of predicting the size, elastic modulus, and the location of flaws embedded
in soft materials. One potential application for this methodology is early breast cancer detection.
This would make breast cancer screening and diagnostic safe for those who cannot
undergo ionizing radiation and more accessible to those who do not have access to mammography.
Methods: Simulations of a breast containing a tumor of various elastic moduli and
sizes were created. A pressure load simulating an ultrasound signal emitted from a probe
was applied to the outside surface of the breast. The resulting signal, along with known
elastic modulus and size of the tumor, was used to train a convolutional neural network.
Convolutional neural networks consist of convolution layers where the input data is being
transform by element-wise multiplication with convolution kernel, pooling layers that down
sample the resulting data, and fully connected (FC) layers that connect the input and output.
To experimentally validate the model, tissue mimicking phantoms will be created and imaged
using an Olympus-1 ultrasound machine.
Results: MAPE for the training data set was 0.733 and 5.66 for size and elastic modulus,
respectively. MAPE for the testing data set was 1.11 and 7.83 for size and elastic modulus,
respectively.
Conclusions: The convolutional neural network trained on simulation data was evaluated
to be highly effective and accurate. Future work is required to experimentally validate
and expand the model to be able to characterize smaller tumors and tumor density.
- Subject (fast)
(authorityURI="http://id.worldcat.org/fast", valueURI="http://id.worldcat.org/fast/00817247")
- Topic
- Artificial intelligence
- Subject (fast)
(authorityURI="http://id.worldcat.org/fast", valueURI="http://id.worldcat.org/fast/01715717")
- Topic
- Diagnostic ultrasonic imaging
- Subject (fast)
(authorityURI="http://id.worldcat.org/fast", valueURI="http://id.worldcat.org/fast/01775421")
- Topic
- Abaqus (Electronic resource)
- Language
- Language Term (ISO639-2B)
- English
- Record Information
- Record Content Source (marcorg)
- RPB
- Record Creation Date
(encoding="iso8601")
- 20220706