Implementation of Real-Time Computer-Aided Diagnosis System with Quantitative Staging and Navigation on Customizable Embedded Digital Signal Processor

[127] Endoscopic

artefact

detection

https://ead2019.grand-challenge.org/

challenge

(EAD2019),

[128] S. Ali, F. Zhou, C. Daul, B. Braden, A. Bailey, S. Realdon, J. East, G.

Wagnires, V. Loschenov, E. Grisan, W. Blondel, and J. Rittscher,

“Endoscopy artifact detection (EAD 2019) challenge dataset,” Computing

Research Repository, CoRR, vol. abs/1905.03209, 2019.

[129] S. Yang and G. Cheng, “Endoscopic artefact detection and segmentation

with deep convolutional neural network.,” Proceedings of the 2019

Challenge on Endoscopy Artefacts Detection (EAD2019), Venice, Italy,

vol.2366 of CEUR Workshop Proceedings. CEUR-WS.org, 2019.

[130] I. Oksuz, J. R. Clough, A. P. King, and J. A. Schnabel, “Artefact

detection in video endoscopy using retinanet and focal loss function.,”

Proceedings of the 2019 Challenge on Endoscopy Artefacts Detection

(EAD2019), Venice, Italy, vol.2366 of CEUR Workshop Proceedings.

CEUR-WS.org, 2019.

[131] M. A. Khan and J. Choo, “Multi-class artefact detection in video

endoscopy via convolution neural networks.,” Proceedings of the 2019

Challenge on Endoscopy Artefacts Detection (EAD2019), Venice, Italy,

vol.2366 of CEUR Workshop Proceedings. CEUR-WS.org, 2019.

[132] C. Zhang, N. Zhang, D. Wang, Y. Cao and B. Liu, "Artifact Detection in

Endoscopic Video with Deep Convolutional Neural Networks.,"

Proceedings of Second International Conference on Transdisciplinary AI

(TransAI), Irvine, CA, USA, pp.1-8, 2020.

[133] S. Ali, F. Zhou, B. Braden, A. Bailey, S. Yang, G. Cheng, P. Zhang, X.

Li, M. Kayser, R.D. Soberanis-Mukul, S. Albarqouni, X. Wang, C. Wang,

S. Watanabe, I. Oksuz, Q. Ning, S. Yang, M.A. Khan, X.W. Gao, S.

Realdon, M. Loshchenov, J.A. Schnabel, J.E. East, G. Wagnieres, V.B.

Loschenov, E. Grisan, C. Daul, W. Blondel, and J. Rittscher, “An objective

comparison of detection and segmentation algorithms for artefacts in

104

clinical endoscopy.,” Scientific Reports, 10, pp.1-15, 2020.

[134] X. Gao, B. Braden, S. Taylor and W. Pang, "Towards Real-Time

Detection of Squamous Pre-Cancers from Oesophageal Endoscopic

Videos.," Proceedings of 18th IEEE International Conference On Machine

Learning And Applications (ICMLA), Boca Raton, FL, USA, pp.1606-1612,

2019.

[135] L. Peng, S. Liu, D. Xie, S. Zhu and B. Zeng, "Endoscopic video deblurring

via synthesis," Proceedings of 2017 IEEE Visual Communications and

Image Processing (VCIP), St. Petersburg, FL, USA, pp.1-4, 2017.

[136] Sharib Ali, Felix Zhou, Adam Bailey, Barbara Braden, James East, Xin

Lu, and Jens Rittscher, “A deep learning framework for quality

assessment and restoration

abs/1904.07073, 2019.

in

video

endoscopy.,”

CoRR,

vol.

[137] F. Tan, S. Liu, L. Zeng and B. Zeng, "Notice of Removal: Kernel-free

video deblurring via synthesis," Proceedings of 2016 IEEE International

Conference on Image Processing (ICIP), Phoenix, AZ, USA, pp. 2683-2687,

2016.

[138] R. Girshick, J. Donahue, T. Darrell, and J. Malik, “Rich feature

hierarchies for accurate object detection and semantic segmentation,”

Proceedings of 2014 IEEE Conference on Computer Vision and Pattern

Recognition, pp. 580-587, 2014.

[139] J. Redmon, S. Divvala, R. Girshick and A. Farhadi, "You Only Look

Once: Unified, Real-Time Object Detection," Proceedings of 2016 IEEE

Conference on Computer Vision and Pattern Recognition (CVPR’16), Las

Vegas, NV, USA, pp.779-788, 2016.

[140] J. Redmon and A. Farhadi, “YOLO9000: Better, Faster, Stronger,”

Proceedings of 2017 IEEE Conference on Computer Vision and Pattern

Recognition (CVPR’17), Honolulu, HW, USA, pp.6517-6525, 2017.

105

[141] MediaTek Inc., “MediaTek Helio P60.” https://www.mediatek.jp/

products/mediatek-helio-p60

[142] MediaTek Inc., “MediaTek Helio P90.” https://www.mediatek.com/

products/smartphones/mediatek-helio-p90

[143] Qualcomm Technologies, Inc., “Snapdragon 710 Mobile Platform.”

https://www.qualcomm.com/products/snapdragon-710-mobile-platform

[144] Qualcomm Technologies, Inc., “Snapdragon 675 Mobile Platform.”

https://www.qualcomm.com/products/snapdragon-675-mobile-platform

Inc.,

[145] MediaTek

products/aiot/i500

“i500

(MT8385).”

https://www.mediatek.com/

[146] VIA Technologies, Inc., “VIA SOM-9X50”

https://www.viatech.com/en/products/boards/modules/som-9x50/

[147] A. Ignatov, R. Timofte, A. Kulik, S. Yang, K. Wang, F. Baum, M. Wu, L.

Xu, L. V. Gool, “AI Benchmark: All About Deep Learning on Smartphones

in 2019,” arXiv:1910.06663v1 [cs.PF], 2019.

[148] A.G. Howard, M. Zhu, B. Chen, D. Kalenichenko, W. Wang, T. Weyand,

M. Andreetto, H. Adam, “Mobilenets: Efficient convolutional neural

networks for mobile vision applications,” arXiv preprint arXiv:1704.04861,

2017.

[149] C. Szegedy, V. Vanhoucke, S. Ioffe, J. Shlens, Z. Wojna, “Rethinking the

inception architecture for computer vision,” Proceedings of Conference on

Computer Vision and Pattern Recognition (CVPR’16), Las Vegas, NV,

USA, pp. 2818–2826, 2016.

[150] C. Szegedy, S. Ioffe, V. Vanhoucke, A.A. Alemi, “Inception-v4, inceptionresnet and the impact of residual connections on learning,” AAAI. vol. 4,

p. 12, 2017.

106

[151] C. Dong, C. C. Loy, K. He and X. Tang, "Image Super-Resolution Using

Deep Convolutional Networks," IEEE Transactions on Pattern Analysis

and Machine Intelligence, vol. 38, no. 2, pp. 295-307, 1 Feb. 2016

107

Chapter 6. Conclusion

6.1. Conclusion

In this thesis, I described the hardware implementation of a computer-aided

diagnosis support system that supports endoscopic video using a

customizable DSP. I clarified the problem of unclear frames in endoscopic

video images and described the balancing of training dataset to build a robust

CAD system with CNN feature extraction and SVM classification. I carefully

analyzed bottlenecks when the CAD system is installed into the customizable

DSP core and clarified optimization methods so that processing can be

executed in real time in the hardware prototype system to be developed. I

evaluated the developed CAD system with these improvement approaches by

results of classification. And, I confirmed the developed CAD system with

CNN feature extraction and SVM classification can be executed in real-time

for endoscopic video images with high classification accuracy.

In Chapter 1, I explained the background of this research. First, I described

trends in colorectal cancer incidence and mortality worldwide. After that, I

explained the principle of NBI magnified endoscopy and described the

classification of colorectal cancer by the NBI endoscopy observation.

In Chapter 2, I explained how the conventional BoF-based CAD system

works, then explained the CAD system with CNN feature extraction and SVM

classification. I described the difference between outputs from CNN layers

equivalence to feature quantity and the features by D-SIFT, and further

described SVM libraries to be implemented referred the report of our research

group. I confirmed over 90% accuracy for quantitative staging classification

with pre-trained CNN instead of D-SIFT by limited training data.

In Chapter 3, I clarified the problems that affect the classification accuracy

in endoscopic video images. Also, the endoscopic image dataset used in this

study was explained, and the problems caused by the imbalanced dataset

were described. To solve the problem of imbalanced dataset, I prepared the

balancing dataset and evaluated the result of classification accuracy from

SVM trained by the balancing dataset. As a result, I confirmed that the

classification accuracy of SVM is improved by using the balancing dataset. I

described the application to endoscopic video images. I explained some issues

such as blurring, color shift etc. in the endoscopic video image. I evaluated

the result of classification accuracy when the endoscopic video image was

108

input to the developed CAD system. It was shown that the classification

accuracy was improved by using multiple size image dataset for SVM training.

I confirmed average value and standard deviation of the output of SVM for

quantitative staging classification (0.692 ± 0.328 -> 0.995 ± 0.015 for

Type2A/3 staging).

In Chapter4, I showed that the acceleration of convolutional layer consists

of multiply and accumulate is the key to CNN processing. I clarified the

requirement for hardware to be implemented a CAD system with CNN and

SVM. I decided to implement a CAD system on a customizable DSP, and

performed a detailed analysis of bottlenecks, and explained the optimization

method. I implemented the optimized CAD system with CNN and SVM on

the FPGA-based prototyping system and confirmed that it is possible to

classify endoscopic images in real time. I confirmed real-time processing of

CAD system with staging classification for edge device at small clinics.

(44.6 fps / 22 ms latency @ 200MHz, 66.6 mW)

In Chapter 5, And, I proposed two navigation function methods by unclear

region detection and by multiple staging region. I confirmed that it is possible

to realize the navigation function by unclear region detection using YOLO2

and staging classification by AlexNet and SVMs executed at 30 fps on the

customizable embedded DSP core. I confirmed that it is possible to realize the

navigation function by multiple staging region at 39fps on the customizable

DSP core.

I described the real-time CAD system with quantitative staging and

navigation on customizable embedded DSP. I proposed a CAD system that

provides quantitative and objective index of cancer stage to doctors based on

the standard JNET classification, not only classify cancer or not-cancer and

implemented on a customizable DSP. And, I confirmed the developed CAD

system achieved real-time quantitative staging classification for the

endoscopic video image (44.6fps throughput / 22ms latency @200MHz,

66.6mW power consumption), and sufficient classification accuracy (> 90%).

And, I proposed two navigation functions which provide unclear region

information to doctors in the CAD system, and I confirmed that it is possible

to realize the CAD system with the navigation function on the customizable

embedded DSP core.

109

Academic and Industrial impact

1) CAD system for real-time video on customizable DSP

- Memory and cycles reduction by 8-bit quantized AlexNet and SVM

implementation. (Academic)

- Avoiding system bus conflict by using hidden layer feature. (Academic)

- 44.6 fps and 22 ms latency, 66.6 mW power (Industrial)

2) Real-time CAD with over 90% accuracy

- Pre-trained CNN instead of Handcrafted feature Extraction (Academic)

- Multi-sizing and balancing in training data set (Academic)

- Quantitative staging with 90% accuracy (Industrial)

3) Real-time Navigation for effective quantitative staging

- unclear region detection for one staging region (Academic), 30fps @ 525

MHz (Industrial)

- multiple staging regions (Academic), 39fps @525 MHz (Industrial)

Figure 87 shows comparison of performance, power and diagnosis quality to

previous studies. We have achieved high performance, power consumption

and diagnostic quality compared to previous studies.

(b) Performance vs Power

(a) Performance vs Diagnosis Quality

40

with Navigation

66mW

30

Kudo2020

CNN: 500W

Quantitative

Diagnosis

Kominami2016

D-SIFT/SVM: 500W

20

100

Byrne2019

CNN: 600W

Classification

10

Detection

10

20

Performance (fps)

Kudo2020

CNN: Classification

Kominami2016

D-SIFT/SVM: Staging

10

Reference:

Jetson Nano: AlexNet

Wang2018

CNN: 800W

Wang2018

CNN: Detection

Byrne2019

CNN: Classification

1000

Our CAD

Power (W)

Diagnosis Quality

Staging

Diagnosis

30

● Server GPUs

● Customizable DSP

0.1

● CPU

● Server

Our CAD

with Navigation

10

20

30

Performance (fps)

Figure 87 Comparison of (a) performance vs diagnosis quality and (b)

performance vs power.

From the above research, quantitative and objective staging index are

provided to the doctor more accurately in magnified NBI endoscopic

observation, which is independent from the experience of doctors, and

110

diagnostic support method is established such as a "second opinion" at

magnified NBI endoscopic observation on site.

6.2. Future Works

It has been shown that it is possible to construct a CAD system that can

process endoscopic video images in real time, however some problems still

remain.

In our research group, there are enough endoscopic images for each

pathological type which has clearly captured and trimmed the lesion by the

clinical doctor. Currently, there is a few images for unclear parts including

blur, reflection of light or color shift. Therefore, as future research subjects,

1) we accumulate the training data sets for navigation function by CNN,

classification, and verify the practical CAD system capability. Unclear region

can be detected by YOLO2, however, there are lesion part or normal mucosa

in the bounding box output by YOLO2. Thus, 2) we evaluate not only YOLO2

for unclear region detection quantitively using statistical measure such as

mAP (mean Average Precision) and implement with classification into the

customizable DSP core. In the proposed CAD system, direction to clear region

is navigated by a red arrow. 3) We consider how to provide the navigation and

improve the navigation function which allows clinical doctors to freely switch

the warning display according to their level of experience.

Expansion to endoscopic systems other than NBI (Narrow Band Imaging) is

also a future research topic. The proposed CAD system can classify NBI

images magnified from medium-magnification (1.3-1.9x) to strongmagnification (3.6x). It is considered to apply non-magnified endoscopic

images and high magnification (360x) by endocytoscopy. In addition, it is

considered to apply endoscopic system using engineering methods different

from NBI such as WLI (White Light Imaging) and BLI (Blue Laser Imaging).

Optimization of the customizable DSP core by adding user-defined

instruction set is also remained. This makes it possible to perform more

complicated processing in neural networks proposed in recent years not only

AlexNet or YOLO2. By using a neural network proposed in recent years, a

CAD system with higher classification accuracy can be expected.

Furthermore, it has been reported an endoscopic system controlled by voice

recognition [152] and voice recognition implemented on the customizable DSP

for medical devices [153]. Therefore, assistance by voice recognition in the

111

operation of a doctor’s endoscope is also conceivable.

112

Referrences

[152] S. Afonso, I. Laranjo, J. Braga, V. Alves, and J. Neves, “Endoscopic

Procedures Control Using Speech Recognition.,” Advances in Information

Science and Applications, vol.2, pp.404-409, 2014.

[153] A. Yamada, M. Tsuji, N. Tamba, M. Odagawa, and Y. Fujinaga,

“Development of intra-operative sterile image reference system

manipulated by voice recognition.,” Proceedings of the 79th Annual

Meeting of the Japan Radiological Society, Yokohama, Japan, 2020

113

Publications

(1) A Hardware Implementation on Customizable Embedded DSP Core for

Colorectal Tumor Classification with Endoscopic Video toward Real-Time

Computer-Aided Diagnosis System.

Masayuki Odagawa, Takumi Okamoto, Tetsuhi Koide, Toru Tamaki,

Bisser Raytchev, Kazufumi Kaneda, Shigeto Yoshida, Hiroshi Mieno,

Shinji Tanaka, Takayuki Sugawara, Hiroshi Toishi, Masayuki Tsuji and

Nobuo Tamba

IEICE

TRANSACTIONS

on

Fundamentals

of

Electronics,

Communications and Computer Sciences, Vol.E104-A77, No.4, pp. 691701 (2021).

DOI: 10.1587/transfun.2020EAP1069

(2) Feasibility Study for Computer-Aided Diagnosis System with Navigation

Function of Clear Region for Real-Time Endoscopic Video Image on

Customizable Embedded DSP Cores.

Masayuki Odagawa, Tetsushi Koide, Toru Tamaki, Shigeto Yoshida,

Hiroshi Mieno, Shinji Tanaka

IEICE

TRANSACTIONS

on

Fundamentals

of

Electronics,

Communications and Computer Sciences, Vol.E105-A, No.1, pp.-, In Press

DOI: 10.1587/transfun.2021EAL2044

(3) Classification with CNN features and SVM on Embedded DSP Core for

Colorectal Magnified NBI Endoscopic Video Image.

Masayuki Odagawa, Takumi Okamoto, Tetsushi Koide, Toru Tamaki,

Shigeto Yoshida, Hiroshi Mieno, Shinji Tanaka

IEICE

TRANSACTIONS

on

Fundamentals

of

Electronics,

Communications and Computer Sciences, Vol.E105-A, No.1, pp.-, In Press

DOI: 10.1587/transfun.2021EAP1036

114

Presentations on Conferences

International conference

1.

First Author

A Hardware Implementation of Colorectal Tumor Classification for Endoscopic

Video on Customizable DSP toward Real-time Computer-Aided Diagnosis System.

Masayuki Odagawa, Takumi Okamoto, Tetsuhi Koide, Toru Tamaki, Bisser

Raytchev, Kazufumi Kaneda, Shigeto Yoshida, Hiroshi Mieno, Shinji Tanaka,

Takayuki Sugawara, Hiroshi Toishi, Masayuki Tsuji, and Nobuo Tamba

Proceedings of 2019 IEEE International Symposium on Circuits and Systems,

Sapporo, Japan, pp.1-5 (2019), Oral Session.

DOI: 10.1109/ISCAS.2019.8702379

2.

Classification Method with CNN features and SVM for Computer-Aided Diagnosis

System in Colorectal Magnified NBI Endoscopy.

Masayuki Odagawa, Takumi Okamoto, Tetsuhi Koide, Toru Tamaki, Bisser

Raytchev, Kazufumi Kaneda, Shigeto Yoshida, Hiroshi Mieno, and Shinji Tanaka

Proceedings of IEEE Regional 10 Conference, Online, Japan, pp.1-6 (2020), Oral

Session.

DOI： 10.1109/TENCON50793.2020.9293709

3.

Tensilica DSP cores for neural networks and an application to computer-aided

diagnosis system for colorectal tumor classification.

Masayuki Odagawa, Masayuki Tsuji, Hiroshi Toishi, Takayuki Sugawara, Nobuo

Tamba, Takumi Okamoto, Tetsushi Koide, Toru Tamaki, Bisser Raytchev,

Kazufumi Kaneda, Shigeto Yoshida, Hiroshi Mieno, and Shinji Tanaka

Proceedings of the International Workshop on Nanodevice Technologies 2018,

Higshi-Hiroshima, Japan, pp.6-7 (2018), Oral Session.

4.

Image Identification System on Rapid Prototyping Platform with Customizable

DSP Core for Biomedical Application.

Masayuki Odagawa, Masayuki Tsuji, Hiroshi Toishi, Takayuki Sugawara, Nobuo

Tamba, Takumi Okamoto, Tetsushi Koide, Toru Tamaki, Bisser Raytchev,

Kazufumi Kaneda, Shigeto Yoshida, Hiroshi Mieno, and Shinji Tanaka

The 3rd International Symposium on Biomedical Engineering, Higshi-Hiroshima,

Japan, November 2018, Poster Presentation.

115

Coauthor

1.

Implementation of Computer-Aided Diagnosis System on Customizable DSP Core

for Colorectal Endoscopic Images with CNN Features and SVM.

Takumi Okamoto, Tetsuhi Koide, Toru Tamaki, Bisser Raytchev, Kazufumi

Kaneda, Shigeto Yoshida, Hiroshi Mieno, Shinji Tanaka, Takayuki Sugawara,

Hiroshi Toishi, Masayuki Tsuji, Masayuki Odagawa, and Nobuo Tamba

Proc. of 2018 IEEE Regional 10 Conference, Jeju, pp.1663-1666 (2018), Oral

Session

DOI: 10.1109/TENCON.2018.8650331

2.

Feature Extraction of Colorectal Endoscopic Images for Computer-Aided Diagnosis

with CNN.

Takumi Okamoto, Masayuki Odagawa, Tetsuhi Koide, Toru Tamaki, Bisser

Raytchev, Kazufumi Kaneda, Shigeto Yoshida, Hiroshi Mieno and Shinji Tanaka

Proc. of 2019 2nd International Symposium on Devices, Circuits and Systems,

Higashi-Hiroshima, pp.1-4 (2019), Oral Session

DOI: 10.1109/ISDCS.2019.8719104

3.

Real-time processing of computer-aided diagnosis system for colorectal tumor

classification in NBI endoscopy using CNN features by implementing to

Tensilica Vision P6 DSP.

Masayuki Tsuji, Hiroshi Toishi, Takayuki Sugawara, Masayuki Odagawa, Nobuo

Tamba, Takumi Okamoto, Tetsushi Koide, Toru Tamaki, Bisser Raytchev,

Kazufumi Kaneda, Shigeto Yoshida, Hiroshi Mieno, and Shinji Tanaka

Proc. of the 2nd International Symposium on Biomedical Engineering, Tokyo,

Japan, pp.2425-2426 (2017), Oral Session

4.

Implementation of Computer-Aided Diagnosis System on Customizable DSP Core

for Colorectal Endoscopic Images with CNN features and SVM.

Takumi Okamoto, Tetsuhi Koide, Toru Tamaki, Bisser Raytchev, Kazufumi

Kaneda, Shigeto Yoshida, Hiroshi Mieno, Shinji Tanaka, Masayuki Odagawa,

Takayuki Sugawara, Hiroshi Toishi, Masayuki Tsuji, and Nobuo Tamba

The 2nd International Symposium on Biomedical Engineering, Tokyo, Japan,

November 2017, Poster Presentation.

116

5.

An Improvement of Real-Time Computer-Aided Diagnosis System for Colorectal

Endoscopic Video.

Hiroki Iwata, Masaya Ueda, Guan Juangang, Tetsushi Koide, Toru Tamaki, Bisser

Raytchev, Kazufumi Kaneda, Shigeto Yoshida, Hiroshi Mieno, Shinji Tanaka,

Masayuki Odagawa, Hiroshi Toishi, Takayuki Sugawara, Masayuki Tsuji, and

Nobuo Tamba

The 4th International Symposium on Biomedical Engineering, Hamamatsu, Japan,

November 2019, Poster Presentation

6.

A Method to Improve Recognition Rate of Computer-Aided Diagnosis System for

Colorectal Endoscopic Images.

Masaya Ueda, Hiroki Iwata, Masayuki Odagawa, Guan Juangang, Tetsushi Koide,

Toru Tamaki, Bisser Raytchev, Kazufumi Kaneda, Shigeto Yoshida, Hiroshi

Mieno, and Shinji Tanaka

The 4th International Symposium on Biomedical Engineering, Hamamatsu, Japan,

November 2019, Poster Presentation

Domestic conference

1.

Coauthor

CNN 特徴と SVM 分類を適用した大腸内視鏡画像がん診断支援システムの

カスタマイザブル DSP コアへの実装(Implementation of Computer-Aided

Diagnosis System for Colorectal Endoscopic Images with CNN features and SVM

on Customizable DSP Core.)Takumi Okamoto, Tetsuhi Koide, Toru Tamaki, Bisser

Raytchev, Kazufumi Kaneda, Shigeto Yoshida, Hiroshi Mieno, Shinji Tanaka,

Takayuki Sugawara, Hiroshi Toishi, Masayuki Tsuji, Masayuki Odagawa and

Nobuo Tamba

Proc. of Design Automation Symposium 2017, Kaga, pp.33-38 (2017), Oral

Session

2.

大腸がん診断支援のためのカスタマイザブル DSP による内視鏡動画リアル

タイム CNN 特徴抽出と SVM 分類(Implementation of Computer-Aided

Diagnosis System for Colorectal Endoscopic Images with CNN features and SVM

on Customizable DSP Core.)

Takumi Okamoto, Tetsuhi Koide, Toru Tamaki, Bisser Raytchev, Kazufumi

Kaneda, Shigeto Yoshida, Hiroshi Mieno, Shinji Tanaka, Takayuki Sugawara,

Hiroshi Toishi, Masayuki Tsuji, Masayuki Odagawa and Nobuo Tamba

117

Proc. of Design Automation Symposium 2018, Kaga, pp.39-44 (2018), Oral

Session

3.

機械学習による内視鏡動画像リアルタイム診断支援システムのプロトタイ

ピング(Prototyping of Real-time Computer-Aided Diagnosis System for Colorectal

Endoscopic Movies and Images with Machine Learning.)

Takumi Okamoto, Masayuki Odagawa, Koujirou Takebayashi, Mikihisa Nagano,

Tetsushi Koide, Toru Tamaki, Bisser Raytchev, Kazufumi Kaneda, Shigeto

Yoshida, Hiroshi Mieno, Shinji Tanaka, Takayuki Sugawara, Hiroshi Toishi,

Masayuki Tsuji and Nobuo Tamba

デザインガイア 2018, Hiroshima, Japan

IEICE Technical Report, vol.118, no.334, VLD2018-42, pp.13-18 (2018), Oral

Session

118

Acknowledgement

I’m working for Cadence

joined Koide laboratory

mentorships, advices and

until today. Therefore, I

Design Systems (Japan) B.V. In October 2018, I

as a doctoral student. I have received many

supports from many people in working this study

would like to take this opportunity to show my

greatest appreciation to all of them.

First and foremost, I would like to show my deepest appreciation to my

supervisor Associate Professor Tetsushi Koide for immense mentorships and

supports. I appreciate all his contributions of time, advices, ideas and

stimulating my experiences.

I would like to express again my deepest appreciation to my sub-supervisors,

Professor Shin-Ichiro Kuroki, Professor Akinobu Teramoto, Professor Minoru

Fujishima and Professor Suguru Kameda for reviewing my doctoral thesis

and advising on my research.

In addition, I would like to express my deep gratitude to Professor Seiichiro

Higashi, Professor Masakazu Iwasaka, Associate Professor Anri Nakajima,

Associate Professor Shuhei Amakawa, Associate Professor Takeshi Yoshida,

Associate Professor Masataka Miyake, Associate Professor Hiroaki Hanafusa,

Associate Professor Mamoru Sasaki, at Department of Semiconductor

Electronics and Integration Science, Graduate School of Advanced Sciences

of Matter, Hiroshima University. I also would like to express my deep

gratitude to the professors and the staff at Research Institute for Nanodevice

and Bio Systems, Hiroshima University.

I would like to show my grateful appreciation to Professor Kazufumi Kaneda,

and Associate Professor Bisser Raytchev, Graduate School of Engineering,

Hiroshima University for cooperation of research. I would like to show my

grateful appreciation to Associate Professor Toru Tamaki, Department of

Computer Science, Nagoya Institute of Technology. I would like to show my

grateful appreciation to Professor Shigeto Yoshida, and Dr. Hiroshi Mineno,

JR Hiroshima Hospital for cooperation of research. I would like to show my

grateful appreciation to Professor Shinji Tanaka, Department of Endoscopy,

Hiroshima University for cooperation of research.

I’m deeply grateful to Mr. Takumi Okamoto, Cadence Design Systems

(Japan) B.V., for many discussions, advices and supports since he was a

doctoral student in Hiroshima University.

I would like to express my special thanks to Mr. Toshifumi Kaneko,

119

President of Cadence Design Systems (Japan) B.V., for permission to study

in a doctoral course and assistance in a study.

I would like to express deepest thanks to Dr. Nobuo Tamba, CEO of

T2Laboratory, for an encouragement to enter a doctoral course and enormous

suggestions and advices since he was my boss in Cadence Design Systems

(Japan) B.V.

I would like to show grateful appreciation to Mr. Hiroshi Toishi for enormous

cooperation of hardware implementation of the computer-aided diagnosis

system in Chapter 4 when he was my colleague in Cadence Design Systems

(Japan) B.V.

I would like to express deepest appreciation to my colleagues, Mr. Takayuki

Sugawara and Dr. Masayuki Tsuji, Cadence Design Systems (Japan) B.V., for

grateful cooperation of implementation and optimization onto a customizable

DSP cores in Chapter 4.

I would like to show my special thanks to R&D teams in Cadence Design

Systems Inc. and colleagues in Cadence Design Systems (Japan) B.V. for

many supports and cooperation.

Finally, I would like to express deepest appreciation to my family for their

cooperation and patience. It took long time to write this thesis since I decided

to enter a doctoral course. I would like to apologize my family for placing large

strain.

Masayuki Odagawa

Yokohama, Kanagawa, Japan

July, 2021

120

公表論文

（Articles）

(1) A Hardware Implementation on Customizable Embedded DSP Core for

Colorectal Tumor Classification with Endoscopic Video toward Real-Time

Computer-Aided Diagnosis System.

Masayuki Odagawa, Takumi Okamoto, Tetsuhi Koide, Toru Tamaki, Bisser

Raytchev, Kazufumi Kaneda, Shigeto Yoshida, Hiroshi Mieno, Shinji Tanaka,

Takayuki Sugawara, Hiroshi Toishi, Masayuki Tsuji and Nobuo Tamba

IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and

Computer Sciences, Vol.E104.A, No.4, pp.691-701 (2021).

DOI: 10.1587/transfun.2020EAP1069

(2) Feasibility Study for Computer-Aided Diagnosis System with Navigation

Function of Clear Region for Real-Time Endoscopic Video Image on

Customizable Embedded DSP Cores.

Masayuki Odagawa, Tetsushi Koide, Toru Tamaki, Shigeto Yoshida, Hiroshi

Mieno, Shinji Tanaka

IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and

Computer Sciences, Vol.E105-A, No.1, pp.-, In press.

DOI: 10.1587/transfun.2021EAL2044

(3) Classification with CNN features and SVM on Embedded DSP Core for

Colorectal Magnified NBI Endoscopic Video Image.

Masayuki Odagawa, Takumi Okamoto, Tetsushi Koide, Toru Tamaki, Shigeto

Yoshida, Hiroshi Mieno, Shinji Tanaka

IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and

Computer Sciences, Vol.E105-A, No.1, pp.-, In press.

DOI: 10.1587/transfun.2021EAP1036

参 考 論 文

（Thesis Supplements）

(1) A Hardware Implementation of Colorectal Tumor Classification for Endoscopic

Video on Customizable DSP toward Real-time Computer-Aided Diagnosis

System.

Masayuki Odagawa, Takumi Okamoto, Tetsuhi Koide, Toru Tamaki, Bisser

Raytchev, Kazufumi Kaneda, Shigeto Yoshida, Hiroshi Mieno, Shinji Tanaka,

Takayuki Sugawara, Hiroshi Toishi, Masayuki Tsuji and Nobuo Tamba

Proceedings of 2019 IEEE International Symposium on Circuits and Systems,

Sapporo, Japan, pp.1-5 (2019).

DOI: 10.1109/ISCAS.2019.8702379

(2) Classification Method with CNN features and SVM for Computer-Aided

Diagnosis System in Colorectal Magnified NBI Endoscopy.

Masayuki Odagawa, Takumi Okamoto, Tetsuhi Koide, Toru Tamaki, Bisser

Raytchev, Kazufumi Kaneda, Shigeto Yoshida, Hiroshi Mieno and Shinji Tanaka

Proceedings of IEEE Regional 10 Conference, Online, Japan, pp.1-6 (2020)

DOI： 10.1109/TENCON50793.2020.9293709

...