A Deep Learning Framework for Pediatric TLE Detection Using 18F-FDG-PET Imaging

Symmetricity-Driven Learning Framework for Pediatric Temporal Lobe Epilepsy Detection Using 18F-FDG-PET Imaging

Status

Completed

Phases

Unknown

Study type

Observational

Source

ClinicalTrials.gov

Registry ID

NCT04169581

Enrollment

201

Registered

2019-11-20

Start date

2018-06-01

Completion date

2019-04-30

Last updated

2020-01-02

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

Epilepsy, Temporal Lobe

Keywords

Deep Learning, Temporal Lobe Epilepsy, Positron-Emission Tomography

Brief summary

This study aims to use radiomics analysis and deep learning approaches for seizure focus detection in pediatric patients with temporal lobe epilepsy (TLE). Ten positron emission tomograph (PET) radiomics features related to pediatric temporal bole epilepsy are extracted and modelled, and the Siamese network is trained to automatically locate epileptogenic zones for assistance of diagnosis.

Detailed description

Purpose:The key to successful epilepsy control involves locating epileptogenic focus before treatment. 18F-FDG PET has been considered as a powerful neuroimaging technology used by physicians to assess patients for epilepsy. However, imaging quality, viewing angles, and experiences may easily degrade the consistency in epilepsy diagnosis. In this work, the investigators develop a framework that combines radiomics analysis and deep learning techniques to a computer-assisted diagnosis (CAD) method to detect epileptic foci of pediatric patients with temporal lobe epilepsy (TLE) using PET images. Methods:Ten PET radiomics features related to pediatric temporal bole epilepsy are first extracted and modelled. Then a neural network called Siamese network is trained to quanti-fy the asymmetricity and automatically locate epileptic focus for diagnosis.The performance of the proposed framework was tested and compared with both the state-of-art clinician software tool and human physicians with different levels of experiences to validate the accuracy and consistency.

Interventions

None listed

Study design

Observational model

COHORT

Time perspective

RETROSPECTIVE

Eligibility

Sex/Gender

ALL

Age

6 Years to 18 Years

Healthy volunteers

Inclusion criteria

1. Clinical diagnosis of temporal lobe epilepsy. 2. Age range from six to eighteen years old. 3. Underwent PET, EEG, computed tomography (CT) and MRI.

Exclusion criteria

1. Image quality is unsatisfactory (e.g. severe image artifacts due to head movement). 2. 18F-FDG PEG examination is negative. 3. Clinical data is incomplete. 4. EEG or MRI report is missing.

Design outcomes

Primary

Measure	Time frame	Description
The 'area under curve' (AUC ) of our model in detection performance	Through study completion, about 1 year	To evaluate the performance of our model, the investigators calculated the AUC of our model for normal or abnormal classification campared with different methods and and physicians with different levels.

Secondary

Measure	Time frame	Description
The 'dice similarity coefficient' (DSC) of our model in detection performance	Through study completion, about 3 months	The accuracy of focus lesion detection is quantitatively measured through the metric of 'dice similarity coefficient' (DSC) by comparing the spatial overlap between the marked regions between the reference standard and the subject method under test.

Countries

China

Outcome results