ANN predicted outcome from acute image set. The last slices in both columns present the least
reliable prediction of the ANN.
As shown in this Figure, the optimal sensitivity and specificity of the ANN are calculated at the half angle and ROC curve intercept point (Point Opt.; Specificity Optimal = 0.84, Sensitivity Optimal = 0.85) which implies that the ANN can predict lesion and tissue with a sensitivity of 85% and a specificity of the 84%.
This figure presents DWI, T2WI-Chronic, ANN, and ISODATA maps for 3 different patients
In ISODATA maps the lesions are clustered into 2 to 3 different signatures [from top to bottom: (5, 6, and 8), (2, 3, and 4), (3 and 4)]. Signature number indicates the chance of tissue recovery. The higher the signature values, the lower the chance of recovery.
No comments:
Post a Comment