About Dataset

This dataset contains a total of 1528 prostate MRI images in the transverse plane. The images and classification were provided by PROSTATEx Dataset and Documentation. The objective of the dataset is to train a convolutional neural network called Small VGG Net and classify new images into clinically significant and clinically non-significant for a systems engineering undergraduate thesis at the Autonomous University of Bucaramanga (UNAB).

Data Selection and Manipulation

A total of 64 patient images were taken. These patients should have a single prostate MRI finding for more accurate training. We then converted all images from DICOM to JPEG. Finally, we separated the images into two groups following the retention method. 30% of the images were from the validation group and the rest from the training group. As a result, we have two groups (significant and non-significant) divided into training (70%) and validation (30%) groups.

Thesis group

Director
Leonardo Hernán Talero Sarmiento
ltalero@unab.edu.co

Students

Juan Felipe Consuegra Rodríguez
jconsuegra869@unab.edu.co
Yeison Omar Hernández Suárez
yhernandes557@unab.edu.co

Citation of data

Geert Litjens, Oscar Debats, Jelle Barentsz, Nico Karssemeijer, and Henkjan Huisman. "ProstateX Challenge data," The Cancer Imaging Archive (2017). DOI: 10.7937/K9TCIA.2017.MURS5CL