Advanced

Convolutional Neural Networks for Classification of Prostate Cancer Metastases Using Bone Scan Images

Belcher, Lewis LU (2017) FYTM04 20161
Department of Astronomy and Theoretical Physics
Abstract
Convolutional neural networks (CNNs) are used to classify directly on bone scan images in two medical tasks: classifying anterior / posterior pose, and classifying bone scan hotspots as metastatic / non-metastatic in patients with prostate cancer and suspected metastatic disease. The networks trained produce highly accurate results in both tasks and current methods are outperformed for all tested body regions when classifying metastatic / non-metastatic hotspots. For one such dataset current methods obtain an area under receiver operating characteristic (ROC) score of 0.9352. By utilising CNNs and other developments in the field an area under ROC of 0.9739 is obtained for the same test set. We consider this a remarkable result given the... (More)
Convolutional neural networks (CNNs) are used to classify directly on bone scan images in two medical tasks: classifying anterior / posterior pose, and classifying bone scan hotspots as metastatic / non-metastatic in patients with prostate cancer and suspected metastatic disease. The networks trained produce highly accurate results in both tasks and current methods are outperformed for all tested body regions when classifying metastatic / non-metastatic hotspots. For one such dataset current methods obtain an area under receiver operating characteristic (ROC) score of 0.9352. By utilising CNNs and other developments in the field an area under ROC of 0.9739 is obtained for the same test set. We consider this a remarkable result given the exclusion of hand-designed heuristics used in current methods. (Less)
Popular Abstract
Recent developments in machine learning have inspired a surge of interest in artificial neural networks (ANNs). Efficient algorithms, novel designs and increased hardware capability have led to a dominance of machine learning over humans in various paradigms: games such as chess and GO, and now even in areas such as image recognition. The number of fields in which machine learning is becoming not only applicable, but increasingly beneficial, is growing rapidly.

Convolutional neural networks (CNNs) are a particular class of neural networks designed to work with locally correlated data, such as image data, and the development has led to a dramatic increase in the performance of image classification. In light of this, CNNs and other... (More)
Recent developments in machine learning have inspired a surge of interest in artificial neural networks (ANNs). Efficient algorithms, novel designs and increased hardware capability have led to a dominance of machine learning over humans in various paradigms: games such as chess and GO, and now even in areas such as image recognition. The number of fields in which machine learning is becoming not only applicable, but increasingly beneficial, is growing rapidly.

Convolutional neural networks (CNNs) are a particular class of neural networks designed to work with locally correlated data, such as image data, and the development has led to a dramatic increase in the performance of image classification. In light of this, CNNs and other recently developed enhancements are implemented in this project to improve methods for identifying prostate cancer metastases (the spread of cancer away from the primary site) using bone scan images.

By using CNNs, a substantial increase in the performance of identifying prostate cancer metastases was realised. If implemented, this can lead to more informed diagnoses, and ultimately, better survival rates for those affected by one of the most common forms of cancer worldwide. (Less)
Please use this url to cite or link to this publication:
author
Belcher, Lewis LU
supervisor
organization
course
FYTM04 20161
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Theoretical Physics, Machine Learning, Supervised Learning, Artificial Neural Networks, Convolutional Neural Networks, Image Classification, Deep Learning, Prostate Cancer, Metastatic Disease
language
English
id
8900840
date added to LUP
2017-01-27 10:59:00
date last changed
2017-01-27 10:59:00
@misc{8900840,
  abstract     = {Convolutional neural networks (CNNs) are used to classify directly on bone scan images in two medical tasks: classifying anterior / posterior pose, and classifying bone scan hotspots as metastatic / non-metastatic in patients with prostate cancer and suspected metastatic disease. The networks trained produce highly accurate results in both tasks and current methods are outperformed for all tested body regions when classifying metastatic / non-metastatic hotspots. For one such dataset current methods obtain an area under receiver operating characteristic (ROC) score of 0.9352. By utilising CNNs and other developments in the field an area under ROC of 0.9739 is obtained for the same test set. We consider this a remarkable result given the exclusion of hand-designed heuristics used in current methods.},
  author       = {Belcher, Lewis},
  keyword      = {Theoretical Physics,Machine Learning,Supervised Learning,Artificial Neural Networks,Convolutional Neural Networks,Image Classification,Deep Learning,Prostate Cancer,Metastatic Disease},
  language     = {eng},
  note         = {Student Paper},
  title        = {Convolutional Neural Networks for Classification of Prostate Cancer Metastases Using Bone Scan Images},
  year         = {2017},
}