• 096279 - BIOMEDICAL SIGNAL PROCESSING AND MEDICAL IMAGES [1] - BIOE 440

The course is intended to recall general signal processing aspects, to introduce new methods beyond those considered in the bachelor degree and examine various clinical and research application fields. Methods. From deterministic filtering to stochastic parametric analysis: mono and multi-variate AR/MA/ARMA models and non parametric as well as parametric spectral analysis. Principal component analysis. Entropy in signal processing. Applications. Automatic analysis and classification of the electrocardiographic signal (ECG). The autonomic nervous system, cardiovascular variability signals, and cardiorespiratory interactions. Fetal ECG signal. High resolution ECG and late ventricular potentials. The central nervous system: processing of the electroencephalographic signal (EEG) and of evoked (EP) and event-related potentials (ERP).

The learning objective is that the student takes possession of the signal processing methods in biomedical signals. At the end of the Course students should be able to work autonomously in different practical cases also different from those addressed in the lessons and/or in the laboratory practical exercises.

Part [1] Biomedical Signal processing

Introduction to biomedical signal processing. General block diagram of biomedical signal processing operations from analog preprocessing to digital conversion of the signals. Recall of analog filters. Numerical signals, numerical filters (FIR and IIR), and their application in biomedical field.

Cardiovascular System. ECG signal: additive noise model and main superimposed noises. Most significant configurations from clinical standpoint. High resolution ECG. Ventricular late potentials: pathophysiological aspects and processing methods.

Heart Rate Variability (HRV). Study of Autonomic Nervous System (ANS) by signal processing in short- and long-term bases. Analysis of ANS branches to HRV regulation. Pathophysiological aspects: neural control mechanisms. Fetal ECG: processing methods and enhancement of useful clinical parameters. Fetal monitoring.

Arterial blood pressure signal: detection systems and main clinical parameters. Interaction models encompassing relationship between several signals related to autonomic nervous system: ECG, arterial blood pressure and respiration; open- and closed-loop models. Pathophysiological interpretations.

Neurosensorial System. EEG (electroencephalographic) signal analysis, evoked potentials (EPs) and event-related potentials (ERPs). Review on traditional processing methods with main applications in clinical and research environments: cerebral activity detected on the scalp, at cortical level and in deep brain stimulation (DBS).

Signal processing and parametric identification. Time series analysis extracted from biological signals. Examples in ECG, EEG, EP in pre-processing, filtering, prediction, extraction and estimation of parameters.

Diagnostic Classification and Modelling. Deterministic and stochastic identification. Non parametric and parametric models. FFT and DFT power spectrum analysis. Review on stochastic identification approaches. Bayes classifiers. Model families: AR/MA/ARMA (autoregressive and moving average). Models with exogenous input X. Parametric spectral analysis: backward and forward, maximum entropy methods, Pisarenko, Prony. Comparison with traditional non-parametric techniques: examples on HRV signals, EEGs and EPs. 

Classification methods: Principal Component Analysis (PCA): general introduction and applications. Independent Component Analysis (ICA).

Entropy Analysis: methods and applications. Approximate and Sample Entropy.

Laboratory activities: Exercises with Matlab implementing signal analysis tools. Application to biomedical signals. Homework exercises with evaluation. Students will develop knowledge about the most important methods and applications for biomedical signal processing and analysis working on biomedical signal samples.

Part [1] Biomedical Signal processing (short summary)

The course is intended to recall general signal processing aspects, to introduce new methods beyond those considered in the bachelor degree and cover various clinical and research application fields. Methods. From deterministic filtering to stochastic parametric analysis: mono and multi-variate AR/MA/ARMA models and non parametric as well as parametric spectral analysis. Principal component analysis. Entropy in signal processing. Applications. Automatic analysis and classification of the electrocardiographic signal (ECG). The autonomic nervous system, cardiovascular variability signals, and cardiorespiratory interactions. Fetal ECG signal. High resolution ECG and late ventricular potentials. The central nervous system: processing of the electroencephalographic signal (EEG) and of evoked (EP) and event-related potentials (ERP).

it is advisable to be aware of the fundamentals of digital signal processing.

Assessment

The assessment will be based on a written exam with exercises and open questions covering all the Course topics and on a project developed in the Laboratory part of the Course. The project will be presented to the instructor at the end of the course. Practical assessments given as lab activity (not mandatory) and made by students will complete the exam.

The score of the written exam, the project part and the practical assessments will be summed to compute the total score. 30 cum laude will be assigned when the total score is equal or higher than 16.

The written exam will cover all course topics both in signal and image processing. It will assign up to 12 points and will be considered sufficient when the score will be equal or higher than 7. The project part will assign up to 3 points. Two Lab assignments will assign up to 2 points. These Lab assignments are not mandatory. Students can take the written part at any exam session during the year. No mid-term assessments will be scheduled.