Luận văn unified deep neural networks for anotomical site classification and lesion segmentation for upper gastrointestinal endoscopy

ee eee 44 4.7 EndoUnet - Confusion matrix on anatomical site classification task GA NGÌu sea H Hướng a ew aes we SB RS eR oe 49 4.8 SFMNet - Confusion matrix on anatomical site clas

Hanoi University of Science and Technology School of Information and Communication Technology

7

D

Master Thesis in Data Science

Unified Deep Neural Networks for Anatomical Site

Classification and Lesion Segmentation for Upper

Author’s Declaration

Thereby declare that I am the sole author of this Uhesis The results in this work

are not complete copies of any other works

STUDENT

Nguyen Duy Manh

14 Qutlie of the thesis

2 Artificial Intelligence aud Machine Learning

2.1 Basia concepts

2.21 Supervised learning 2.2.2 Unsupervised uming 0.0 02 0000000

22.8 Reinforcement learning 2.3 Techniques

Convolnrional Neural Network

References ñ2

3.9 Overview comparison between FPN and FaPN [15] 38 3.10 Feature alignment module [15] 2.0 2.00000 ee eee eee 39 3.11 Feature selection module [I5] eee 39 4.1 Demostration of upper GI

4.2 Some samples in anatomical dataset

4.3 Some samples in lesion dataset SE

4.4 Some samples in HP dataset 0.0 ee eee 44

4.7 EndoUnet - Confusion matrix on anatomical site classification task

GA NGÌu sea H Hướng a ew aes we SB RS eR oe 49

4.8 SFMNet - Confusion matrix on anatomical site classification task on

2.3.2.3 Motivation 23.24 Activation function

23.5.1 The Transformer 23.5.2 Transformers for Vision 2

Squeeze and excitation module .-

3.2.5 Feature-atigned pyramid network 3.2.6 Classifiers " An R

Metrics and loss functions

3.9 Overview comparison between FPN and FaPN [15] 38 3.10 Feature alignment module [15] 2.0 2.00000 ee eee eee 39 3.11 Feature selection module [I5] eee 39 4.1 Demostration of upper GI

4.2 Some samples in anatomical dataset

4.3 Some samples in lesion dataset SE

4.4 Some samples in HP dataset 0.0 ee eee 44

4.7 EndoUnet - Confusion matrix on anatomical site classification task

GA NGÌu sea H Hướng a ew aes we SB RS eR oe 49

4.8 SFMNet - Confusion matrix on anatomical site classification task on

Abstract

Image Processing is a subfield of computer vision concerned with comprehending and extracting data from digital images ‘There are several applications for image processing in various fields, including face recognition, optical character recognition,

main orientations: (1) as a computer-aided diagnosis to help the physicians for an

efficient and early diagnosis, with a better harmonization and less contradictory diagnosis; (2) to enhance the medical care of patients with better-personalized ther- apies; and (3) to improve the human wellbeing, for example by analyzing the spread

multiple simultaneous tasks pertaining to the upper gastrointestinal (G1) tract On

a dataset of 11469 endoscopic images, the models were evaluated and produced

relatively positive results.

3.9 Overview comparison between FPN and FaPN [15] 38 3.10 Feature alignment module [15] 2.0 2.00000 ee eee eee 39 3.11 Feature selection module [I5] eee 39 4.1 Demostration of upper GI

4.2 Some samples in anatomical dataset

4.3 Some samples in lesion dataset SE

4.4 Some samples in HP dataset 0.0 ee eee 44

4.7 EndoUnet - Confusion matrix on anatomical site classification task

GA NGÌu sea H Hướng a ew aes we SB RS eR oe 49

4.8 SFMNet - Confusion matrix on anatomical site classification task on

Reinforcement Learning

Abstract

Image Processing is a subfield of computer vision concerned with comprehending and extracting data from digital images ‘There are several applications for image processing in various fields, including face recognition, optical character recognition,

main orientations: (1) as a computer-aided diagnosis to help the physicians for an

efficient and early diagnosis, with a better harmonization and less contradictory diagnosis; (2) to enhance the medical care of patients with better-personalized ther- apies; and (3) to improve the human wellbeing, for example by analyzing the spread

multiple simultaneous tasks pertaining to the upper gastrointestinal (G1) tract On

a dataset of 11469 endoscopic images, the models were evaluated and produced

relatively positive results.

Abstract

Image Processing is a subfield of computer vision concerned with comprehending and extracting data from digital images ‘There are several applications for image processing in various fields, including face recognition, optical character recognition,

main orientations: (1) as a computer-aided diagnosis to help the physicians for an

efficient and early diagnosis, with a better harmonization and less contradictory diagnosis; (2) to enhance the medical care of patients with better-personalized ther- apies; and (3) to improve the human wellbeing, for example by analyzing the spread

multiple simultaneous tasks pertaining to the upper gastrointestinal (G1) tract On

a dataset of 11469 endoscopic images, the models were evaluated and produced

relatively positive results.

Abstract

Image Processing is a subfield of computer vision concerned with comprehending and extracting data from digital images ‘There are several applications for image processing in various fields, including face recognition, optical character recognition,

main orientations: (1) as a computer-aided diagnosis to help the physicians for an

efficient and early diagnosis, with a better harmonization and less contradictory diagnosis; (2) to enhance the medical care of patients with better-personalized ther- apies; and (3) to improve the human wellbeing, for example by analyzing the spread

multiple simultaneous tasks pertaining to the upper gastrointestinal (G1) tract On

a dataset of 11469 endoscopic images, the models were evaluated and produced

relatively positive results.

Reinforcement Learning

Abstract

Image Processing is a subfield of computer vision concerned with comprehending and extracting data from digital images ‘There are several applications for image processing in various fields, including face recognition, optical character recognition,

main orientations: (1) as a computer-aided diagnosis to help the physicians for an

efficient and early diagnosis, with a better harmonization and less contradictory diagnosis; (2) to enhance the medical care of patients with better-personalized ther- apies; and (3) to improve the human wellbeing, for example by analyzing the spread

multiple simultaneous tasks pertaining to the upper gastrointestinal (G1) tract On

a dataset of 11469 endoscopic images, the models were evaluated and produced

relatively positive results.

Reinforcement Learning

References ñ2

Reinforcement Learning

2.3.2.3 Motivation 23.24 Activation function

23.5.1 The Transformer 23.5.2 Transformers for Vision 2

Squeeze and excitation module .-

3.2.5 Feature-atigned pyramid network 3.2.6 Classifiers " An R

Metrics and loss functions

Detailed sevlings of MiT-B2 and MiT-B3 0.0.00

Number of images in each anatomical site and lighting mode Accuracy comparison on the three classification taska

Dive Score comparison on the segmentation task 0

Number of parameters and speed of models

43

a7

48 +

Reinforcement learning components 2 00 ee 6

Mlustration of a deep learning model [2] 9

Architecture of a CN 13

Sparse connectivity, viewed from above [Đ] l5

Common activation functions [5] 16

Architecture of an FON [6] 0.0.0 bende eee 19

Architecture of VGGI6 [J] eee 20

Architecture of EndoUNet 31 VGG19-based shared block + 82 ResNet50-based shared bloek c2 262 38 DenseNet121-based shared block 33 EndoUNet decoder configuration 34

SFMNet architecture Grouped compact generalized non-local (CGNL) module [13]

2.3.2.3 Motivation 23.24 Activation function

23.5.1 The Transformer 23.5.2 Transformers for Vision 2

Squeeze and excitation module .-

3.2.5 Feature-atigned pyramid network 3.2.6 Classifiers " An R

Metrics and loss functions

3.9 Overview comparison between FPN and FaPN [15] 38 3.10 Feature alignment module [15] 2.0 2.00000 ee eee eee 39 3.11 Feature selection module [I5] eee 39 4.1 Demostration of upper GI

4.2 Some samples in anatomical dataset

4.3 Some samples in lesion dataset SE

4.4 Some samples in HP dataset 0.0 ee eee 44

4.7 EndoUnet - Confusion matrix on anatomical site classification task

GA NGÌu sea H Hướng a ew aes we SB RS eR oe 49

4.8 SFMNet - Confusion matrix on anatomical site classification task on

References ñ2

2.3.2.3 Motivation 23.24 Activation function

23.5.1 The Transformer 23.5.2 Transformers for Vision 2

Squeeze and excitation module .-

3.2.5 Feature-atigned pyramid network 3.2.6 Classifiers " An R

Metrics and loss functions

Detailed sevlings of MiT-B2 and MiT-B3 0.0.00

Number of images in each anatomical site and lighting mode Accuracy comparison on the three classification taska

Dive Score comparison on the segmentation task 0

Number of parameters and speed of models

43

a7

48 +

Reinforcement Learning

Reinforcement learning components 2 00 ee 6

Mlustration of a deep learning model [2] 9

Architecture of a CN 13

Sparse connectivity, viewed from above [Đ] l5

Common activation functions [5] 16

Architecture of an FON [6] 0.0.0 bende eee 19

Architecture of VGGI6 [J] eee 20

Architecture of EndoUNet 31 VGG19-based shared block + 82 ResNet50-based shared bloek c2 262 38 DenseNet121-based shared block 33 EndoUNet decoder configuration 34

SFMNet architecture Grouped compact generalized non-local (CGNL) module [13]

Detailed sevlings of MiT-B2 and MiT-B3 0.0.00

Number of images in each anatomical site and lighting mode Accuracy comparison on the three classification taska

Dive Score comparison on the segmentation task 0

Number of parameters and speed of models

43

a7

48 +

References ñ2

3.9 Overview comparison between FPN and FaPN [15] 38 3.10 Feature alignment module [15] 2.0 2.00000 ee eee eee 39 3.11 Feature selection module [I5] eee 39 4.1 Demostration of upper GI

4.2 Some samples in anatomical dataset

4.3 Some samples in lesion dataset SE

4.4 Some samples in HP dataset 0.0 ee eee 44

4.7 EndoUnet - Confusion matrix on anatomical site classification task

GA NGÌu sea H Hướng a ew aes we SB RS eR oe 49

4.8 SFMNet - Confusion matrix on anatomical site classification task on

Detailed sevlings of MiT-B2 and MiT-B3 0.0.00

Number of images in each anatomical site and lighting mode Accuracy comparison on the three classification taska

Dive Score comparison on the segmentation task 0

Number of parameters and speed of models

43

a7

48 +

Abstract

Image Processing is a subfield of computer vision concerned with comprehending and extracting data from digital images ‘There are several applications for image processing in various fields, including face recognition, optical character recognition,

main orientations: (1) as a computer-aided diagnosis to help the physicians for an

efficient and early diagnosis, with a better harmonization and less contradictory diagnosis; (2) to enhance the medical care of patients with better-personalized ther- apies; and (3) to improve the human wellbeing, for example by analyzing the spread

multiple simultaneous tasks pertaining to the upper gastrointestinal (G1) tract On

a dataset of 11469 endoscopic images, the models were evaluated and produced

relatively positive results.

Abstract

Image Processing is a subfield of computer vision concerned with comprehending and extracting data from digital images ‘There are several applications for image processing in various fields, including face recognition, optical character recognition,

main orientations: (1) as a computer-aided diagnosis to help the physicians for an

efficient and early diagnosis, with a better harmonization and less contradictory diagnosis; (2) to enhance the medical care of patients with better-personalized ther- apies; and (3) to improve the human wellbeing, for example by analyzing the spread

multiple simultaneous tasks pertaining to the upper gastrointestinal (G1) tract On

a dataset of 11469 endoscopic images, the models were evaluated and produced

relatively positive results.

Reinforcement learning components 2 00 ee 6

Mlustration of a deep learning model [2] 9

Architecture of a CN 13

Sparse connectivity, viewed from above [Đ] l5

Common activation functions [5] 16

Architecture of an FON [6] 0.0.0 bende eee 19

Architecture of VGGI6 [J] eee 20

Architecture of EndoUNet 31 VGG19-based shared block + 82 ResNet50-based shared bloek c2 262 38 DenseNet121-based shared block 33 EndoUNet decoder configuration 34

SFMNet architecture Grouped compact generalized non-local (CGNL) module [13]

Reinforcement Learning

2.3.2.3 Motivation 23.24 Activation function

23.5.1 The Transformer 23.5.2 Transformers for Vision 2

Squeeze and excitation module .-

3.2.5 Feature-atigned pyramid network 3.2.6 Classifiers " An R

Metrics and loss functions

References ñ2

Abstract

Image Processing is a subfield of computer vision concerned with comprehending and extracting data from digital images ‘There are several applications for image processing in various fields, including face recognition, optical character recognition,

main orientations: (1) as a computer-aided diagnosis to help the physicians for an

efficient and early diagnosis, with a better harmonization and less contradictory diagnosis; (2) to enhance the medical care of patients with better-personalized ther- apies; and (3) to improve the human wellbeing, for example by analyzing the spread

multiple simultaneous tasks pertaining to the upper gastrointestinal (G1) tract On

a dataset of 11469 endoscopic images, the models were evaluated and produced

relatively positive results.