| | Step 1: import MRI from medical database |
| | Step 2: linear filter via Gaussian filter |
| | Step 3: normalization using histogram processing |
| | Step 4: initiate patch extraction process |
| (i) | Patches obtained |
| (ii) | RGB channel separation |
| | Step 5: initiate feature extraction process |
| (i) | Color mapping and obtaining threshold value |
| (ii) | LBP process to attain binary image |
| (a) | Image is converted to a grayscale representation. |
| (b) | For each pixel (gp) in the image, select the P neighbouring pixels. gp’s coordinates are specified by |
| (c) | Set the pixel in the centre (gc) as the threshold for its P neighbours. |
| (d) | Set to 1 if the adjacent pixel’s value is larger than or equal to the centre pixel’s value, and 0 otherwise. |
| (e) | Compute the LBP value now. First, write a binary number comprising digits next to the centre pixel in a counterclockwise direction. This binary integer (or its decimal counterpart) is referred to as the LBP-central pixel code and is employed as a distinctive local texture. |
| | Step 6: initiate classification process |
| (i) | Multi-SVM process |
| (i) | Use training set, group train, and test set as variables for function |
| (ii) | Classify test cases and map the training data into kernel space |
| (ii) | CNN classification |
| (i) | Load train and test data |
| (ii) | Iterate the process with 100 epochs which yields less error value of 1.2% |
| (iii) | Create layers and subsampling layers for CNN for varied kernel sizes |
| (iv) | Classify the data and predict the final output |
