A full automatic method for the soft tissues sarcoma treatment response based on fuzzy logic

Aim of this study was to develop a full automatic method for the soft tissue sarcoma (STS) identification and its evaluation during chemotherapy (CT) treatment, based on diffusion MRI. This procedure includes two main phases, the first one is a registration step in order to compose a coregistered set of images concerning morphological and functional images pre and post CT, registered to the pre-treatment T1. The second phase is a fuzzy characterization of STS diffusive parameter along with a fuzzy inference step for the evaluation of treatment response on the characterized area. The results of this procedure are two membership degree maps which measure the probability for each segmented pixel to be responding or not to CT. These two maps could assist radiologists during follow-up assessment, by automatically extracting the lesion volume, report lesion composition, and measure the uncertainty of the estimate, in order to manage the intrinsic and well-known heterogeneity of STS. Many studies demonstrated the prognostic values of diffusion MRI and the apparent diffusion coefficient (ADC) in the assessment of cellularity changes during therapy and their correlation with lesion response. In this scenario, the proposed framework, allows for a total unsupervised identification and probabilistic evaluation of STS treatment response based on diffusion MRI, mapping the local changes of ADC during therapy rather than describe the whole lesion by a statistical moment.