Computer aided diagnosis of acoustic neuroma: A neural network perspective
Research output: Contribution to journal › Review Article › Scientific › peer-review
|Number of pages||7|
|Journal||JOURNAL OF MEDICAL IMAGING AND HEALTH INFORMATICS|
|Publication status||Published - 1 Apr 2017|
|Publication type||A2 Review article in a scientific journal|
Acoustic neuroma is a benign brain tumour, if not identified and diagnosed at early stages can result in loss of hearing. In this study Multilayer Perceptron (MLP) a feedforward Artificial Neural Network model (ANN), has been investigated for segmentation and detection of acoustic neuroma at early stages using medical resonance imaging (MRI). The proposed methodology comprises of two phases. In first phase, regions in the MRI images were classified and segmented as affected (neuroma) or normal areas. The classification was performed on pixel level using MLP. During this phase Region of Interest (ROI) was created through domain-specific knowledge. The MLP was then trained using 1490 random samples. In the second phase acoustic neuroma was detected in MRI images. The statistical analyses were subsequently performed on the pixel level to calculate the current size of detected neuroma and to provide suggestive deadlines for treatment using the growth rate of the tumour. The methodology presented in this research is more efficient compared to state-of-The art practices. The proposed technique not only detects acoustic neuroma but also highlights its location and boundaries and estimates the size of neuroma. Moreover, the computational time associated with detection and size prediction is much less in comparison to the other existing techniques. Exceptional results were obtained with an over-All accuracy of 94.12% in the detection of Acoustic Neuroma during final experimentation.
- Acoustic Neuroma, Image Processing, Machine Vision, Magnetic Resonance Image (MRI), Multi-Layer Perceptron, Segmentation