Advanced

Recognizing Microscopic Structures: Dense Semantic Segmentation of Multiple Histopathological Classes using Fully Convolutional Neural Networks

Isaksson, Johan LU (2016) In Master’s Theses in Mathematical Sciences FMA820 20161
Mathematics (Faculty of Engineering)
Abstract
In order to alleviate the financial burden on the healthcare sector as well as relax its employees’ workload, there is a need to introduce novel tools that automate some of the tasks that today are performed manually. Especially pathology poses a problem with few pathologists, demanding manual labour and unnecessary work on benign tissue. As a response, the DOGS project aims to develop a tool to automate or assist in Gleason grading of histopathological images from prostate biopsies. It is probable that such a tool would benefit from having access to individually segmented, pathologically relevant objects from the images. Moreover, considering recent advances in deep learning and its frequently impressive performance on various image... (More)
In order to alleviate the financial burden on the healthcare sector as well as relax its employees’ workload, there is a need to introduce novel tools that automate some of the tasks that today are performed manually. Especially pathology poses a problem with few pathologists, demanding manual labour and unnecessary work on benign tissue. As a response, the DOGS project aims to develop a tool to automate or assist in Gleason grading of histopathological images from prostate biopsies. It is probable that such a tool would benefit from having access to individually segmented, pathologically relevant objects from the images. Moreover, considering recent advances in deep learning and its frequently impressive performance on various image analysis tasks, it is natural to approach this challenge from a deep learning perspective.

This thesis proposes several fully convolutional neural networks to be used for dense semantic segmentation on histopathological images. The networks’ architectures are all initially based on already proven networks but are modified in various ways to achieve better performance. Being a supervised machine learning task, the ground truth required to train the network has been developed as a part of the thesis. The best-performing network obtained an accuracy of 79.71 % mean intersection over union and the networks presented plausibly equaled or outperformed state-of-the-art methods in nuclei segmentation. However, further work is deemed necessary for reaching adequate segmentation performance. Several suggestions for possible future directions of work are presented, as well as obstacles that have to be considered moving onwards. (Less)
Popular Abstract
To make a significant dent in the issues the healthcare sector faces today in terms of costs and overextended employees, a great increase of viable automated tools will sooner or later be needed. For pathologists this is no different. However, the complexity and size of microscopy images makes automated analysis of them difficult. This thesis achieved promising and first-of-its-kind results in tackling the very underexplored challenge of recognizing several structures in microscopy images at once. By using a deep learning approach heavily inspired from a pair of popular artificial neural networks a score of 80 % mean IU was reached. The resulting networks are prospects for use in several different preprocessing steps in medical image... (More)
To make a significant dent in the issues the healthcare sector faces today in terms of costs and overextended employees, a great increase of viable automated tools will sooner or later be needed. For pathologists this is no different. However, the complexity and size of microscopy images makes automated analysis of them difficult. This thesis achieved promising and first-of-its-kind results in tackling the very underexplored challenge of recognizing several structures in microscopy images at once. By using a deep learning approach heavily inspired from a pair of popular artificial neural networks a score of 80 % mean IU was reached. The resulting networks are prospects for use in several different preprocessing steps in medical image analysis applications – possibly enabling or improving automated tools in the pathological field in the future. (Less)
Please use this url to cite or link to this publication:
author
Isaksson, Johan LU
supervisor
organization
course
FMA820 20161
year
type
H2 - Master's Degree (Two Years)
subject
keywords
segmentation, semantic segmentation, dense semantic segmentation, multiple classes, pathology, digital pathology, histopathology, histopathological, prostate cancer, gleason, gleason grading, gleason scoring, image analysis, machine learning, deep learning, artificial neural networks, ANN, convolutional neural networks, CNN, fully convolutional neural networks, Digital Pathology for Optimized Gleason Score in Prostate Cancer, DOGS
publication/series
Master’s Theses in Mathematical Sciences
report number
LUTFMA-3306-2016
ISSN
1404-6342
other publication id
2016:E47
language
English
id
8895081
date added to LUP
2017-04-18 17:22:24
date last changed
2017-04-18 17:22:24
@misc{8895081,
  abstract     = {In order to alleviate the financial burden on the healthcare sector as well as relax its employees’ workload, there is a need to introduce novel tools that automate some of the tasks that today are performed manually. Especially pathology poses a problem with few pathologists, demanding manual labour and unnecessary work on benign tissue. As a response, the DOGS project aims to develop a tool to automate or assist in Gleason grading of histopathological images from prostate biopsies. It is probable that such a tool would benefit from having access to individually segmented, pathologically relevant objects from the images. Moreover, considering recent advances in deep learning and its frequently impressive performance on various image analysis tasks, it is natural to approach this challenge from a deep learning perspective.

This thesis proposes several fully convolutional neural networks to be used for dense semantic segmentation on histopathological images. The networks’ architectures are all initially based on already proven networks but are modified in various ways to achieve better performance. Being a supervised machine learning task, the ground truth required to train the network has been developed as a part of the thesis. The best-performing network obtained an accuracy of 79.71 % mean intersection over union and the networks presented plausibly equaled or outperformed state-of-the-art methods in nuclei segmentation. However, further work is deemed necessary for reaching adequate segmentation performance. Several suggestions for possible future directions of work are presented, as well as obstacles that have to be considered moving onwards.},
  author       = {Isaksson, Johan},
  issn         = {1404-6342},
  keyword      = {segmentation,semantic segmentation,dense semantic segmentation,multiple classes,pathology,digital pathology,histopathology,histopathological,prostate cancer,gleason,gleason grading,gleason scoring,image analysis,machine learning,deep learning,artificial neural networks,ANN,convolutional neural networks,CNN,fully convolutional neural networks,Digital Pathology for Optimized Gleason Score in Prostate Cancer,DOGS},
  language     = {eng},
  note         = {Student Paper},
  series       = {Master’s Theses in Mathematical Sciences},
  title        = {Recognizing Microscopic Structures: Dense Semantic Segmentation of Multiple Histopathological Classes using Fully Convolutional Neural Networks},
  year         = {2016},
}