Download presentation
Presentation is loading. Please wait.
Published byRoderick Nelson Modified over 6 years ago
1
A Brief Example MRI data and metadata on PTSD patients
Hippocampal volume from structural MRI Intracranial volume from structural MRI NAA levels from MRSI Age Education Level CAPS score (estimated PTSD severity based on psychological exam)
2
A Brief Example I - Get Data Into Readable Format - WEKA recognizes quite a few – ARFF (primary), CSV…
4
A Brief Example Open Weka…
8
A Brief Example Explore the data…
16
A Brief Example Based on exploring the data, some thoughts on a preliminary analysis Reduce number of variables (there are also formal ways of “subset selection” – stepwise regression, criteria based (AIC), the lasso, pace regression,…) “Disctretize” CAPS (as measure of PTSD severity) and try classifying on discretized CAPS score
17
A Brief Example Select varible subset, discretize CAPS score in Weka and save result as an ARFF file
22
A Brief Example Do some classifications in the Weka Epxplorer…
24
A Brief Example C 4.5 Decision Tree: Age Education <= 13 > 13
Rt Hipp Vol Lt Hipp Vol PTSD- Age PTSD+ Intracr Vol >3.01 <= 13 > 13 <=3.01. <=43 >43 <=2.66 >2.66 <=2.43 >2.43 <= 15 > 15 <= 1403 > 1403 <= 33 > 33
25
A Brief Example Then can automate application of multiple algorithms on multiple data sets (e.g. discretization of CAPS scores into different numbers of bins) using the Weka Experimenter
27
A Brief Example Results using Weka Experimenter (10 times, 10 fold corss validation) – prediction accuracy:
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.