Model card
Model Name: Brain Tumor Classifier V1.0
Model Framework: PyTorch
Model Architecture: VGG16
Training Algorithm: Stochastic Gradient Descent (SGD) with a learning rate scheduler
This model is intended to classify brain tumor types using histopathological data. It is designed to be used in a research context or to assist medical professionals in diagnosing brain tumors. The model is not intended to replace professional medical advice. Factors
The model was trained on the Digital Brain Tumour Atlas dataset, which contains information about a variety of tumor types. The dataset is histopathological, so the model is dependent on the quality of these images and their accurate classification. The model was trained to distinguish between 127 different tumor types.
Model performance was assessed using multiple metrics: loss (using Cross-Entropy Loss), accuracy, precision, recall, and F1 score. These metrics were computed on both a validation dataset and the training dataset itself to monitor and prevent overfitting.
The model was evaluated using a hold-out validation set derived from the same Digital Brain Tumour Atlas dataset. This evaluation set was not used during model training to ensure an unbiased estimation of the model's generalization ability.
The model was trained on the Digital Brain Tumour Atlas dataset. The dataset was divided into a training set and a validation set. The training process lasted for 300 epochs, with the learning rate decreasing by a factor of 1/e every 25 epochs using a predefined scheduler.
This model should only be used to assist in brain tumor classification and should not be relied upon as the sole diagnostic tool. The model may not perform with the same level of accuracy across all demographics due to biases inherent in the training dataset. Professional medical consultation and imaging tests should always be prioritized.
The model has shown signs of overfitting during training and requires further improvements and more diverse data to enhance its generalization ability. The performance may be affected by the overfitting, and caution should be exercised when interpreting the results. The effectiveness of the model can be influenced by the quality and resolution of input images. It may not perform well on low-resolution or noisy images that differ significantly from the histopathological images it was trained on. Users should ensure the images they provide for classification are clear and of high quality for the best results.