CCRB Logo

Trial History Detail on 2018-02-07

CUHK_CCRB00582

2018-02-07

Prospective

CREC Ref No. 2017.554

Protocol No. 2017-HL-001

This is no a funded project

N/A

N/A

Not Applicable

Ka Lai Yip

Room 124010,
10/F, Lui Che Woo Clinical Sciences Building,
Price of Wales Hospital,
30-32 Ngan Shing Street,
Shatin, New Territories

3505 3856

Hannahykl@cuhk.edu.hk

Division of Neurology, The Chinese University of Hong Kong

Hong Kong

Ho Wan Leung

Division of Neurology,
Department of Medicine & Therapeutics,
Price of Wales Hospital, Shatin

3505 3856

Howanleung@cuhk.edu.hk

Division of Neurology, The Chinese University of Hong Kong

Hong Kong

Epileptic seizure detection by neural network architecture based on semiology of video recording

Epileptic seizure detection by neural network architecture based on semiology of video recording

利用(Spatial-Temporal GRU Convolutional Neural Network)深度學習的方法開發新型的智能電腦軟件以分析腦癎發作

Hong Kong

Yes

2017-11-20

Joint CUHK-NTEC Clinical Research Ethics Committee

CREC Ref. No: 2017.554

Epilepsy

Other

A artificial intelligent program

This research is an observational study using video recording to construct an artificial intelligent program for the recognition of epileptic seizures. In the present study, a novel marker-free analytical tool is used as a vision-based monitoring to the epileptic episodes. The video of seizure onsets will be confirmed by specialist or delegated research staffs and processed to the computing analysis.

Not drug trial

Not drug trial

Not drug trial

One epileptic seizure episode per subject

The 40 videos recording for the control will be obtained from the open database in the internet. The control is defined as that without seizure in that video recording.

Not drug trial

Not drug trial

Not drug trial

One normal body movement per subject

Subjects are aged or above 18 years old

Subjects are diagnosed with epilepsy

Subjects who completed a video recording

Subjects are able to give an informed consent

No captured event has been found in the video recording

18

999

Both Male and Female

Observational

Not Applicable

Not Applicable

Not Applicable

Not Applicable

Other

N/A

2018-03-01

40

Not Yet Recruiting

Model validation

The currently used detection model (Spatial-Temporal GRU Convolutional Neural Network) will be validated by the study of video from epilepsy patients.

The efficacy of distinguishing the epilepsy behavior and normal motion will be scored.

Detection latency

A loss function for the calculation of the detection latency is created based on (1) the seizures and normal behavior required separately and (2) the uniformly increasing loss function.

The existing of false positive will be eliminated by Two Stream Segmented ConvNets

No

2020-12-07

Type Access Document Published On  
No documents yet.
  • Page 1 of 1.