The Research Unit of Computer Systems and Bioinformatics deals with intelligent analysis of data, signals, images and videos with particular reference to biomedical applications.  Research activity focuses on the use and development of typical methods and techniques for elaboration of signals, machine learning and multi-parameter analysis to retrieve useful information for the final user from data, signals, images and videos. In particular, the laboratory developed competences concerning information filtering and retrieval from signals and images, supervised and unsupervised classification methods, Multi-Expert systems, learning from imbalanced data, reliability estimation of sample classification, rejection option integration,  Computer-Aided-Diagnosis (CAD) instruments, analysis of human behavior and informatics and telemedicine applications for disabled and elderly people. In this respect, the experience of some members of the research group finds its application, concerning high-performance calculation and calculator networks, especially wireless.   That allows on the one hand to apply very sophisticated methodologies and techniques – burdensome from a computational point of view – on great volumes of data, keeping response times within more than acceptable limits considering the specific application, and on the other hand to refer to distributed architectures that allow user interaction modalities to be more flexible and able to attain full usability of  results obtained in biomedical field.

More specifically, research activities concerned:

  • Innovative methods for imbalanced dataset classification able to maximize performances on minority classes without threatening the overall performances
  • Elaboration of 3D microscopy images of hundreds of Gigabytes to: improve their quality, allow an effective visualization, segment objects of biological interest, retrieve synthetic information of the quantitative kind
  • A CAD system for automatic analysis of images in Indirect  immunofluorescence on HEP-2 substrate, as to both intensity evaluation and pattern recognition, and Critidia Lucilae; currently efforts are focused on: advanced techniques for cell segmentation, support to classification of ANCA substrate IIF images, mitosis detection
  • Automatic retrieval of information from images and video sequences. Currently efforts are focused on: classification of actions and activities of monitored subjects and gait analysis of subjects under movement rehabilitation
  • Algorithms for the identification of functional couplings between different cerebral or peripheral areas in EEG-TMS analysis paradigms. The developed techniques tackle the issues of filtering (disturbance and noise removal),  feature individuation, based on time frequency techniques, and signal classification
  • Analysis of biological data, with particular reference to genomic data retrieved with sequencing techniques, to understand the role that can be played by large-dimension DNA structures (long non-coding DNA) in some epigenetic regulation mechanisms
  • Analysis of physiological data, as for instance heart rate and hemoglobin saturation, to monitor patient’s health conditions and foresee the onset of deviations from normality.  Algorithms being developed can be executed on mobile devices or dedicated servers. The applicative area is the patient affected by chronic obstructive lung disease (BPCO)