Algorithm 1: Proposed Fine-Tuning Technique

Inputs: Training and testing images. Outputs: Calculated accuracy. Select the optimal value of features for determining Fine-Tuning output. Fine-Tuning Steps (condition of output limitation). Step 1: Load pre-trained Google-Net model of CNN (replicates all model designs and their parameters on the Google-Net model, except the output layer) Step 2: Truncate the pre-trained network’s last layer (softmax layer) and replace it with our new output layer that is relevant to our problem. Step 3: Add an output layer to the target model, whose number of outputs is the number of categories in the target dataset. Step 4: Freeze the weights of the pre-trained network’s first few layers. The first few layers capture universal features such as curves and edges, which are also relevant to our new problem. Step 5: Start training the new model structure while keeping those weights intact, with the network focusing on learning dataset-specific features in the subsequent layers. Step 6: The output Layer yields one of four classes: Not-Tumour. Glioma-Tumour. Meningioma-Tumour. Pituitary-Tumour. Return accuracy