Glioblastoma (GBM) is an intracranial tumor composed of infiltrating necrotic masses. Automatic contouring of GBM is an open challenging topic, since GBM is an intrinsically heterogeneous (in appearance, shape, and histology) brain tumor [1, 2]. Since 2012 a yearly challenge is organized by the MICCAI Conference, namely the Brain Tumor Image Segmentation Benchmark (BraTS). Although the actual trend is the use of machine learning to solve this problem, legal aspects about the accountability and the explainability of decisions may arise, especially in radiotherapy (RT). We present a logic-based approach using VoxLogica [3], a tool for declarative image analysis that provides powerful building blocks to develop concise, human-readable imaging algorithms.
An explainable algorithm for automatic segmentation of glioblastoma
2019
Abstract
Glioblastoma (GBM) is an intracranial tumor composed of infiltrating necrotic masses. Automatic contouring of GBM is an open challenging topic, since GBM is an intrinsically heterogeneous (in appearance, shape, and histology) brain tumor [1, 2]. Since 2012 a yearly challenge is organized by the MICCAI Conference, namely the Brain Tumor Image Segmentation Benchmark (BraTS). Although the actual trend is the use of machine learning to solve this problem, legal aspects about the accountability and the explainability of decisions may arise, especially in radiotherapy (RT). We present a logic-based approach using VoxLogica [3], a tool for declarative image analysis that provides powerful building blocks to develop concise, human-readable imaging algorithms.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
