MAchine LEarning for miRNA (MALERNA)

Cancer diagnosis may involve impractical or invasive methods. A less invasive monitoring option is to use cancer associated microRNA (miRNAs) biomarkers, present in biological fluids such as blood, urine, saliva, and pleural fluid.. Many studies have demonstrated that miRNAs contribute to carcinogenesis through controlling expression that facilitates tumor growth, invasion, angiogenesis, and immune evasion. In addition, the dysregulation of miRNAs is often associated with the presence of tumors in human tissue. Thus, miRNAs could be use as early biomarkers to detect cancer tumors in an early stage. For example, the survival rate for lung cancer is 5 years is <5\%, but an early diagnosis can increase it to almost 50\%.

miRNAs can function as oncogenes or tumor suppressors in all types of cancer, being considered as prognostic and diagnostic biomarkers and thus it is practicable to build databases with miRNA expressions data for cancer research. Moreover, from the analysis of this information is possible to extract certain features that could be use as cancer biomarkers. For example, hsa-mir-21 is mentioned as a marker for patients with squamous cell lung carcinoma, with astrocytoma , breast cancer, gastric cancer, and is a well-known oncogene.

Formerly, traditional lineal statistical analysis tend to manage inaccurate amounts of necessary data or incorrect quantities of measured elements, make them inadequate to extract meaningful features. For instance,next-generation sequencing (NGS) technologies, such as Applied Biosystems, SOLiD3,or HiSeq from Illumina extract thousands of components in genome sequences. A suitable solution, is to use machine learning techniques for analysis and classification of miRNA data.

Icarii

mir-10b

Human Motion as a Complex System (Brain Modeling)

The movement of the human body is the result of complex processes with several subsystems involved. The motion execution is generated with electrical impulses at the motor cortex, propagating through the nerve system until reaching the alpha neurons in charge of the specific activation of a group of muscle fibers.This study describes the development of an anatomically based computer model to describe the movement in skeletal muscle. The overall system is composed by the abstraction of the following 3 subsystems: The brain and the skeletal muscle with Bidomain formulation, and the cable equation to connect both systems.

Two scenarios are represented: a healthy subject, and a subject with a blockage in the motor pathway. To validate the simulation we compare the activity with the measurements of a patient using fNIRS and EMG. The results of the electrical propagation in both cases are shown to be consistent with the measurements.Previous methods describe the activity of the brain and muscle systems independently. However, a global system that considers the propagation of the electrical activity from the brain to the muscle is still missing in order to integrate these subsystems.The results of these models seem promising to further understanding of neuro muscular diseases. The future work involves the development of closed loop models of the motor units and muscle groups which will help with the design of modern neuro rehabilitation therapies.

 

EEG Simulator

Design of a Video Game for Rehabilitation

Many people suffer from a brain injury that requires rehabilitation. Rehabilitation might be exhausting and difficult. Therefore, it is necessary to develop mechanisms to engage the patients, e.g. a videogame. We propose the design of an immersive videogame that integrates motion capture, electromyography (EMG) sensing and Virtual Reality (VR) in one unique system using Unity engine and the design of EMG sensors. The system has as user inputs the motion capture system given by Microsoft Kinect or Yei 3-Space Sensors, and the measured effort in the muscle (biceps). The output is the control of a videogame sequence using for the visualization the Oculus Rift. We tested the system with 120 healthy subjects, where 94% of them were able to control their muscle’s signal.