Efficient Segmentation Network for Real-Time Blood Vessel Detection of Ultrasound Image on Mobile Devices
- 주제(키워드) Deep Learning , On-Device Machine Learning , Lightweight Network , Segmentation
- 발행기관 서강대학교 일반대학원
- 지도교수 송태경
- 발행년도 2020
- 학위수여년월 2020. 2
- 학위명 석사
- 학과 및 전공 일반대학원 전자공학과
- UCI I804:11029-000000064855
- 본문언어 영어
- 저작권 서강대학교 논문은 저작권보호를 받습니다.
초록/요약
Needle guide technology is in the spotlight for safer and more accurate injections. Ultrasound imaging systems are widely used in needle guide applications because they can show real-time monitoring and help precise injections. It is important to detect the blood vessels that the needle will inject for accurate injection. This paper uses a deep learning method to detect blood vessels. With the advances in deep learning, the need for On-Device Machine Learning which applies to it to embedded systems using has increased. However, the low calculation power and low amount of memory are the limitations for On-Device Machine Learning in mobile devices. This study suggests an efficient segmentation network and implementation optimization for On-Device Machine Learning in mobile devices. Evaluations of human forearm data show that this method can segment blood vessels with real-time processing in mobile devices.
more