Presentation is loading. Please wait.

Presentation is loading. Please wait.

Accurate classification of Tourette Syndrome children using SVM with Tract-Based Spatial Statistics Wen Hongwei State Key Laboratory of Management and.

Similar presentations


Presentation on theme: "Accurate classification of Tourette Syndrome children using SVM with Tract-Based Spatial Statistics Wen Hongwei State Key Laboratory of Management and."— Presentation transcript:

1 Accurate classification of Tourette Syndrome children using SVM with Tract-Based Spatial Statistics
Wen Hongwei State Key Laboratory of Management and Control for Complex Systems Institute of Automation, Chinese Academy of Sciences

2 Contents Background Materials Methods Results Discussion

3 Background What is Tourette Syndrome?
A chronic disorder that has both motor and vocal tics. Often tics are accompanied by comorbidities such OCD, ADHD. How is it diagnosed? Criteria according to the Diagnostic and Statistical Manual of Mental Disorders (DSM) Computer aided diagnosis (CAD) Diffusion Tensor Imaging (DTI) 抽动—秽语综合症又称多发性抽动症,也是临床较为常见的儿童行为障碍综合征,以面部、四肢、躯干部肌肉不自主抽动伴喉部异常发音及猥秽语言为特征的综合症侯群。伴随抽动症的并发症,如强迫症,注意力缺陷多动障碍 医生通过观察病人的抽动、抽搐或者通过病人家长或病人描述, 本病诊断可参照美国《精神疾病诊断统计手册》 来确认是否罹患这种病。目前没有任何检验能够确认本病的诊 断。有些医生要求做一些检查以便排除其他疾病对确认TS的干 扰。诊断具有一定的主观性并时间消耗。 因此引入医学影像方法来辅助TS的诊断具有潜在的意义。

4 Diffusion Tensor Imaging (DTI)
Mobility in a given direction is described by ADC The tissue diffusivity is described by the tensor D The diffusion equation Diffusion is represented by a 33 tensor*

5 DTI Metrics Axial diffusivity: λ// Radial diffusivity: λ⊥
Mean diffusivity Fractional anisotropy: FA

6 Contents Background Materials Methods Results Discussion

7 Materials Subjects Data Acquisition Philips 3.0 Tesla
44 TS (age: 8.98±3.114 years, range: 3–16 years; 11 female). 48 controls (age: 11.00±3.495 years; range: 3–17 years; 17 female). Clinical characteristics: Yale Global Tic Severity Scale(YGTSS) & Duration Data Acquisition Philips 3.0 Tesla DTI:30 non-collinear directions with a b value of 1000 s/mm2 T1:3D Fast Field Echo (FFE) sequence

8 Contents Background Materials Methods Results Discussion

9 Methods (Pipeline)

10 Methods (Tract-Based Spatial Statistics)

11 Methods(Relief algorithm)
Algorithm Relief Input: for each training instance a vector of attribute values and the class value Output: the vector W of estimations of the qualities of attributes Set all weights W[Ai]=0.0, i=1,2,…,p ; for j=1 to m do begin randomly select an instance Xj; find nearest hit Hj and nearest miss Mj; for k=1 to p do begin W[Ak]=W[Ak]-diff(Ak, Xj, Hj)/m+diff(Ak, Xj, Mj)/m; end;

12 Methods(ReliefF Feature Selection)

13 Methods (Classifications)

14 Contents Background Materials Methods Results Discussion

15 Results (Standard TBSS)
FA was decreased significantly RD was increased significantly MD was increased significantly no significant AD change

16 Results (criteria) Accuracy =(TP+TN)/(TP+TN+FN+FP)
Sensitivity = TP/(TP+FN) Specificity = TN/(FP+TN).

17 Results 接收者操作特征曲线(receiver operating characteristic curve,或者叫ROC曲线)

18 Results ROC curve for control and TS children classification under the peak performance of the SVM classifier which occurs with 2000 voxels AD features subset.

19 Results The coordinates of the local maxima and cluster size of AD skeleton clusters selected by the ReliefF algorithm as the most salient for group classification.

20 Contents Background Materials Methods Results Discussion

21 Discussion Our results achieved a highest accuracy of 94.6% better than some previous work(74% accuracy)(Deanna Greene, Jessica Church et al. Classification of children with Tourette syndrome using resting state functional connectivity MRI Flux Congress & SRCD Meeting.(abstract)) Methodological differences between this method and standard TBSS Follow-up work: Multi-Modal, Multi kernel, TS severity regression analysis & prediction

22 Thank you!


Download ppt "Accurate classification of Tourette Syndrome children using SVM with Tract-Based Spatial Statistics Wen Hongwei State Key Laboratory of Management and."

Similar presentations


Ads by Google