[Seminario Línea de Matemáticas grupo GAUNAL, 1er. sem. 2016] «Magnetic Resonance Image preprocessing and dimensionality reduction in Alzheimer disease diagnosis» / «Estimation and quantification of dose uncertainties for the bladder in prostate cancer...
La línea de matemáticas avanzada para el control y los sistemas dinámicos del Grupo de Automática de la Universidad Nacional (Gaunal) de la Facultad de Minas de la UN sede Medellín invita a las exposiciones:
Magnetic Resonance Images are an essential tool to detect and diagnose pathologies in brain like Alzheimer disease since they have high spatial resolution. However, brain morphology variability in the population makes the extraction of characteristics a challenge. In addition, these images represent volumes of millions of voxels that leads to high dimensional systems which may exceed or overwhelm the storage capabilities before being able to manipulate them. In this way, we present two pre-processing steps to deal with those challenges. The first pre-processing is therefore to normalize all the patients in the same framework. The normalization process is made by: a) Realignment, b) Register, c) Segmentation, d) Smoothing and, e) Normalization. For the second problem, it is therefore necessary to reduce the number of variables to represent each Gray Matter image before carrying out population analysis using tools as PCA. With this, we propose to use image compression techniques in the frequency domain, such as Discrete Fourier Transform and Discrete Cosine Transform, in order to reduce the number of parameters to represent each gray matter observation in a population database.
Expositora: María Clara Salazar Londoño
Fecha y hora: viernes 17 de junio de 2016, 10:00-10:45
Lugar: salón M8-116
In prostate cancer radiotherapy, bladder motion and deformation between fractions introduce geometrical uncertainties that make difficult to determine the real delivered dose during treatment. The aim of this work is to estimate mean and variance of the dose delivered to the bladder using Gaussian models of bladder motion and deformation. Based on a training database of patients treated for prostate cancer, individual and population-based models were previously trained based on dominant eigenmodes. For a new patient, a Gaussian distribution can thus be adapted to each mode by using solely the planning CT, and weighted sums of a these eigenmodes can then used to estimate bladders with their probability of occurrence. These statistical modes are therefore used to quantify dose and dose-volume histogram uncertainties produced by bladder motion and deformation between fractions.
Expositor: Richard Ríos Patiño
Fecha y hora: viernes 17 de junio de 2016, 11:00-11:45
Lugar: salón M8-116
Ambos seminarios son dirigidos por el profesor Jairo José Espinosa Oviedo
Información adicional: Magda Pinto Vargas, firstname.lastname@example.org , teléfono (+574)4255295
[Boletín UN Investiga 301, 9 de junio de 2016]