Keywords:
CNS, Computer applications, Neuroradiology brain, MR, Neural networks, Computer Applications-Detection, diagnosis, Diagnostic procedure, Cancer
Authors:
P. Saponara, C. Detheridge, J. Bhangu, B. N. Bloch, K. Thomas; Boston/US
DOI:
10.1594/ecr2018/C-2825
Methods and materials
MRI images of ten primary CNS tumors were obtained from the Radiopaedia (copyright) nervous system library. Single 3T MRI images were chosen to train tumor identification.
The system analyzed 2190 images depicting Craniopharyngioma,
Ependymoma,
Ganglioglioma,
Medullablastoma,
Meningioma,
Pylocytic Astrocytoma,
Acoustic Neuroma,
and Schwannoma.
An AI system using a Convolutional Neural Network (CNN) was built through Python scripting language to accurately identify specific imaging characteristics.
Accuracy efficiency was determined in seconds per slice.
Using this data,
an AI system using a CNN was built through Python scripting language to accurately identify specific imaging characteristics.
The framework used was Google’s™ TensorFlow© 1.3.0,
built to utilize the Graphical Processing Unit (GPU).
A gradient descent algorithm was used to train the algorithm and minimize the loss function.
Validation accuracy was calculated as the percentage of calculated classifications that matched pathological diagnosis,
using a 10% holdout threshold.
Images were randomly shuffled and parsed into either training (90%) or testing (10%) data groups.
Images that were used to train the algorithm were not used to test the algorithm.
Efficiency was determined in seconds per slice.
Three anatomical planes (axial,
sagittal and coronal) were included.