A deep learning algorithm based on convolutional neural networks to detect glandular cells in digitalized biopsies of the prostate. Performed as final degree work for the degree in computer engineering.
Cancer is constituted as the second cause of global death and specifically that of prostate, it is the second type with the highest number of new cases identified each year in Spain and the most deadly among the male popula- tion. An effective way to diagnose it is through the pathological anatomy: a sample of tissue is removed from the organ in question (biopsy) that is sub- sequently analyzed under a microscope by a specialist. However, advances in the field of digital image processing have allowed the emergence of a new technique: digital pathology. Biopsies are processed by powerful scanners that generate high resolution images that can be analyzed through softwa- re. The creation of an algorithm (a diagnostic aid tool) is proposed to detect, in these images, prostate glandular cells, the biological structures in which cancer becomes visible, using convolutional neural networks and computer vision techniques, while studying its performance in large images.