Deep Learning Tool for the Automatic Detection and Segmentation of Liver Tumors

Identifier:  TFM:1882

Handle: https://hdl.handle.net/20.500.11797/TFM1882

Authors:  Balaguer Montero, Maria

Abstract:
We aim to develop a tool for the automatic detection and segmentation of liver tumors with existing Deep Learning models. To do so, we have collected 1516 computed tomography scans, coming from several institutions (open-source and private), accounting for 4790 tumors. A review of the state-of-the-art methods has been done too, as well as a benchmarking comparison to our resulting model. Moreover, the impact in segmentation performance of several scan characteristics has been evaluated, together with a sub-study to assess the inter- and intra-radiologists’ variability. Lastly, the quantification of Total Tumor Volume has been tested as a prognostic biomarker.