Search

Machine learning approaches for image analysis require large amounts of training imaging data. As an alternative, the use of realistic synthetic data reduces the high cost associated to medical image acquisition, as well as avoiding confidentiality and privacy issues, and consequently allows the creation of public data repositories for scientific purposes. Within the context of fetal imaging, we adopt an auto-encoder based Generative Adversarial Network for synthetic fetal MRI generation. The proposed architecture features a balanced power of the discriminator against the generator during training, provides an approximate convergence measure, and enables fast and robust training to generate high-quality fetal MRI in axial, sagittal and coronal planes. We demonstrate the feasibility of the proposed approach quantitatively and qualitatively by segmenting relevant fetal structures to assess the anatomical fidelity of the simulation, and performing a clinical verisimilitude study distinguishing the simulated data from the real images. The results obtained so far are promising, which makes further investigation on this new topic worthwhile.

This work presents a multi-scale agent-based model of emphysema progression that includes both the slow action of the immune system and the fast action of force redistribution and fracture propagation of the biological tissue. The two scales are coupled because the immune response causes inflammation and adaptation, which affects the biomechanical parameters of the tissue such as his elasticity. During repeated inflammation and breathing cycles, the tissue weakens and breaks down. We found that macrophages lifespan and cytokynes diffusion ratio are the parameters that influence the outcome of the model the most.

Cochlear implantation is a surgical technique which aims to restore hearing in people with deep hearing loss. However, outcomes of the surgery still exhibit a large variability between patients. Among the factors that contribute to variability the most important are morphological differences in anatomical structures between patients and incorrect implant placements. In order to address these issues, it would be desirable to have a functional model of the cochlea which incorporates inter-patients variability and simulate electrode placement. To this end, we present a finite element model which captures the interaction between the cochlear partition, modeled as an elastic solid with finite deformation, and the perilymph fluid, modeled as a compressible, viscous fluid. Numerical results show that the membrane responds to changes in the stimulation frequencies.

Agar plates are widely used in the biomedical field as a medium in which to artificially grow bacteria, algae or fungi. Agar plates (Petri dishes) are used routinely in microbiology laboratories in order to identify the type of micro-organism responsible for infections. Such diagnoses are based on counting the number and type of bacterial colonies growing in the Petri dish. The count of bacterial colonies is a time consuming task prone to human error, so interest in automated counting systems has increased in the recent years. One of the difficulties of automatizing the counting process is the presence of markers and annotations made in the lower part of the agar plate. Efficient removal of such markers can increase the accuracy of the bacterial counting system. This article introduces a fast method for detection, segmentation and removal of annotations in agar plates that improves the results of existing bacterial colony counting algorithms.

Modern biomolecular high-throughput technologies can produce experimental results made of lists of hundreds of candidate interesting genes in the condition under study. These lists need to be biologically interpreted to achieve a better knowledge of the patho-physiological phenomena involved in the studied conditions. To reach this goal, several functional, structural, and phenotypic annotations are available within heterogeneous and widely distributed databanks. Among them, gene expression information is a useful resource to better understand gene functions. We previously developed GFINDer, a Web server that aggregates genomic annotations sparsely available in numerous databanks accessible via the Internet and allows performing statistical analysis of functional and phenotypic annotations of gene lists. To take full advantage of gene expression information provided by eVOC ontologies, we imported them in the GFINDer system. For this purpose and to keep updated such information in the GFINDer database when new releases of them are available, we designed and implemented specific parsing and updating procedures. Moreover, we developed new GFINDer modules that allow annotating human nucleotide sequences with the imported eVOC controlled information on their expression features in order to explore and statistically analyze them.