Using multi-layer perceptron with Laplacian edge detector for bladder cancer diagnosis

Area of Science:

• Medical Imaging
• Computational Biology
• Oncology

Background:

• Urinary bladder cancer detection relies on complex imaging analysis.
• Simpler, effective diagnostic methods are needed to complement existing techniques like Deep Learning Convolutional Neural Networks.

Purpose of the Study:

• To evaluate the implementation possibility of a Multi-Layer Perceptron (MLP) for urinary bladder cancer detection.
• To compare the MLP method with established deep learning approaches.
• To assess the impact of image pre-processing and size on diagnostic accuracy.

Main Methods:

• Utilized a dataset of 1997 bladder cancer images and 986 non-cancer tissue images.
• Applied Laplacian edge detection for image pre-processing.
• Trained and tested a Multi-Layer Perceptron model on the processed images.
• Compared diagnostic performance across different image resolutions (50x50, 100x100).

Main Results:

• The Multi-Layer Perceptron, when pre-processed with Laplacian edge detection, achieved an Area Under the Curve (AUC) value of up to 0.99.
• Optimal results were obtained using image sizes of 50x50 and 100x100 pixels.
• The MLP demonstrated a viable and highly accurate alternative for bladder cancer detection.

Conclusions:

• A Multi-Layer Perceptron combined with Laplacian edge detection offers a highly effective and potentially simpler method for urinary bladder cancer diagnosis.
• Image pre-processing and appropriate image sizing are critical factors for maximizing the performance of this diagnostic approach.
• This method shows significant potential for integration into clinical diagnostic workflows.