Statistical methods for analysis and processing of medical ultrasound: applications to segmentation and restoration

In this thesis two major topics inherent with medical ultrasound images are addressed: deconvolution and segmentation. In the first case a deconvolution algorithm is described allowing statistically consistent maximum a posteriori estimates of the tissue reflectivity to be restored. These estimates are proven to provide a reliable source of information for achieving an accurate characterization of biological tissues through the ultrasound echo. The second topic involves the definition of a semi automatic algorithm for myocardium segmentation in 2D echocardiographic images. The results show that the proposed method can reduce inter- and intra observer variability in myocardial contours delineation and is feasible and accurate even on clinical data.

Automated Segmentation and Non-Rigid Registration for the Quantitative Evaluation of Myocardial Perfusion from Magnetic Resonance Images

Myocardial perfusion quantification by means of Contrast-Enhanced Cardiac Magnetic Resonance images relies on time consuming frame-by-frame manual tracing of regions of interest. In this Thesis, a novel automated technique for myocardial segmentation and non-rigid registration as a basis for perfusion quantification is presented. The proposed technique is based on three steps: reference frame selection, myocardial segmentation and non-rigid registration. In the first step, the reference frame in which both endo- and epicardial segmentation will be performed is chosen. Endocardial segmentation is achieved by means of a statistical region-based level-set technique followed by a curvature-based regularization motion. Epicardial segmentation is achieved by means of an edge-based level-set technique followed again by a regularization motion. To take into account the changes in position, size and shape of myocardium throughout the sequence due to out of plane respiratory motion, a non-rigid registration algorithm is required. The proposed non-rigid registration scheme consists in a novel multiscale extension of the normalized cross-correlation algorithm in combination with level-set methods. The myocardium is then divided into standard segments. Contrast enhancement curves are computed measuring the mean pixel intensity of each segment over time, and perfusion indices are extracted from each curve. The overall approach has been tested on synthetic and real datasets. For validation purposes, the sequences have been manually traced by an experienced interpreter, and contrast enhancement curves as well as perfusion indices have been computed. Comparisons between automatically extracted and manually obtained contours and enhancement curves showed high inter-technique agreement. Comparisons of perfusion indices computed using both approaches against quantitative coronary angiography and visual interpretation demonstrated that the two technique have similar diagnostic accuracy. In conclusion, the proposed technique allows fast, automated and accurate measurement of intra-myocardial contrast dynamics, and may thus address the strong clinical need for quantitative evaluation of myocardial perfusion.

Statistical methods for biomedical signal analysis and processing

Statistical modelling and statistical learning theory are two powerful analytical frameworks for analyzing signals and developing efficient processing and classification algorithms. In this thesis, these frameworks are applied for modelling and processing biomedical signals in two different contexts: ultrasound medical imaging systems and primate neural activity analysis and modelling. In the context of ultrasound medical imaging, two main applications are explored: deconvolution of signals measured from a ultrasonic transducer and automatic image segmentation and classification of prostate ultrasound scans. In the former application a stochastic model of the radio frequency signal measured from a ultrasonic transducer is derived. This model is then employed for developing in a statistical framework a regularized deconvolution procedure, for enhancing signal resolution. In the latter application, different statistical models are used to characterize images of prostate tissues, extracting different features. These features are then uses to segment the images in region of interests by means of an automatic procedure based on a statistical model of the extracted features. Finally, machine learning techniques are used for automatic classification of the different region of interests. In the context of neural activity signals, an example of bio-inspired dynamical network was developed to help in studies of motor-related processes in the brain of primate monkeys. The presented model aims to mimic the abstract functionality of a cell population in 7a parietal region of primate monkeys, during the execution of learned behavioural tasks.

Lesion detection and classification in breast cancer: evaluation of approaches based on morphological features, tracer kinetic modelling and semi-quantitative parameters in MR functional imaging (DCE-MRI)

The diagnosis, grading and classification of tumours has benefited considerably from the development of DCE-MRI which is now essential to the adequate clinical management of many tumour types due to its capability in detecting active angiogenesis. Several strategies have been proposed for DCE-MRI evaluation. Visual inspection of contrast agent concentration curves vs time is a very simple yet operator dependent procedure, therefore more objective approaches have been developed in order to facilitate comparison between studies. In so called model free approaches, descriptive or heuristic information extracted from time series raw data have been used for tissue classification. The main issue concerning these schemes is that they have not a direct interpretation in terms of physiological properties of the tissues. On the other hand, model based investigations typically involve compartmental tracer kinetic modelling and pixel-by-pixel estimation of kinetic parameters via non-linear regression applied on region of interests opportunely selected by the physician. This approach has the advantage to provide parameters directly related to the pathophysiological properties of the tissue such as vessel permeability, local regional blood flow, extraction fraction, concentration gradient between plasma and extravascular-extracellular space. Anyway, nonlinear modelling is computational demanding and the accuracy of the estimates can be affected by the signal-to-noise ratio and by the initial solutions. The principal aim of this thesis is investigate the use of semi-quantitative and quantitative parameters for segmentation and classification of breast lesion. The objectives can be subdivided as follow: describe the principal techniques to evaluate time intensity curve in DCE-MRI with focus on kinetic model proposed in literature; to evaluate the influence in parametrization choice for a classic bi-compartmental kinetic models; to evaluate the performance of a method for simultaneous tracer kinetic modelling and pixel classification; to evaluate performance of machine learning techniques training for segmentation and classification of breast lesion.

A new method for the estimation of the reverberation time from measured room impulse responses with application to Italian opera houses

The Schroeder's backward integration method is the most used method to extract the decay curve of an acoustic impulse response and to calculate the reverberation time from this curve. In the literature the limits and the possible improvements of this method are widely discussed. In this work a new method is proposed for the evaluation of the energy decay curve. The new method has been implemented in a Matlab toolbox. Its performance has been tested versus the most accredited literature method. The values of EDT and reverberation time extracted from the energy decay curves calculated with both methods have been compared in terms of the values themselves and in terms of their statistical representativeness. The main case study consists of nine Italian historical theatres in which acoustical measurements were performed. The comparison of the two extraction methods has also been applied to a critical case, i.e. the structural impulse responses of some building elements. The comparison underlines that both methods return a comparable value of the T30. Decreasing the range of evaluation, they reveal increasing differences; in particular, the main differences are in the first part of the decay, where the EDT is evaluated. This is a consequence of the fact that the new method returns a “locally" defined energy decay curve, whereas the Schroeder's method accumulates energy from the tail to the beginning of the impulse response. Another characteristic of the new method for the energy decay extraction curve is its independence on the background noise estimation. Finally, a statistical analysis is performed on the T30 and EDT values calculated from the impulse responses measurements in the Italian historical theatres. The aim of this evaluation is to know whether a subset of measurements could be considered representative for a complete characterization of these opera houses.

A skin surface characterization system based on capacitive image analysis

During the last few years, several methods have been proposed in order to study and to evaluate characteristic properties of the human skin by using non-invasive approaches. Mostly, these methods cover aspects related to either dermatology, to analyze skin physiology and to evaluate the effectiveness of medical treatments in skin diseases, or dermocosmetics and cosmetic science to evaluate, for example, the effectiveness of anti-aging treatments. To these purposes a routine approach must be followed. Although very accurate and high resolution measurements can be achieved by using conventional methods, such as optical or mechanical profilometry for example, their use is quite limited primarily to the high cost of the instrumentation required, which in turn is usually cumbersome, highlighting some of the limitations for a routine based analysis. This thesis aims to investigate the feasibility of a noninvasive skin characterization system based on the analysis of capacitive images of the skin surface. The system relies on a CMOS portable capacitive device which gives 50 micron/pixel resolution capacitance map of the skin micro-relief. In order to extract characteristic features of the skin topography, image analysis techniques, such as watershed segmentation and wavelet analysis, have been used to detect the main structures of interest: wrinkles and plateau of the typical micro-relief pattern. In order to validate the method, the features extracted from a dataset of skin capacitive images acquired during dermatological examinations of a healthy group of volunteers have been compared with the age of the subjects involved, showing good correlation with the skin ageing effect. Detailed analysis of the output of the capacitive sensor compared with optical profilometry of silicone replica of the same skin area has revealed potentiality and some limitations of this technology. Also, applications to follow-up studies, as needed to objectively evaluate the effectiveness of treatments in a routine manner, are discussed.

Non Invasive Tools for Early Detection of Autism Spectrum Disorders

Autism Spectrum Disorders (ASDs) describe a set of neurodevelopmental disorders. ASD represents a significant public health problem. Currently, ASDs are not diagnosed before the 2nd year of life but an early identification of ASDs would be crucial as interventions are much more effective than specific therapies starting in later childhood. To this aim, cheap an contact-less automatic approaches recently aroused great clinical interest. Among them, the cry and the movements of the newborn, both involving the central nervous system, are proposed as possible indicators of neurological disorders. This PhD work is a first step towards solving this challenging problem. An integrated system is presented enabling the recording of audio (crying) and video (movements) data of the newborn, their automatic analysis with innovative techniques for the extraction of clinically relevant parameters and their classification with data mining techniques. New robust algorithms were developed for the selection of the voiced parts of the cry signal, the estimation of acoustic parameters based on the wavelet transform and the analysis of the infant’s general movements (GMs) through a new body model for segmentation and 2D reconstruction. In addition to a thorough literature review this thesis presents the state of the art on these topics that shows that no studies exist concerning normative ranges for newborn infant cry in the first 6 months of life nor the correlation between cry and movements. Through the new automatic methods a population of control infants (“low-risk”, LR) was compared to a group of “high-risk” (HR) infants, i.e. siblings of children already diagnosed with ASD. A subset of LR infants clinically diagnosed as newborns with Typical Development (TD) and one affected by ASD were compared. The results show that the selected acoustic parameters allow good differentiation between the two groups. This result provides new perspectives both diagnostic and therapeutic.

Pattern-based segmentation of digital documents: model and implementation

This thesis proposes a new document model, according to which any document can be segmented in some independent components and transformed in a pattern-based projection, that only uses a very small set of objects and composition rules. The point is that such a normalized document expresses the same fundamental information of the original one, in a simple, clear and unambiguous way. The central part of my work consists of discussing that model, investigating how a digital document can be segmented, and how a segmented version can be used to implement advanced tools of conversion. I present seven patterns which are versatile enough to capture the most relevant documents’ structures, and whose minimality and rigour make that implementation possible. The abstract model is then instantiated into an actual markup language, called IML. IML is a general and extensible language, which basically adopts an XHTML syntax, able to capture a posteriori the only content of a digital document. It is compared with other languages and proposals, in order to clarify its role and objectives. Finally, I present some systems built upon these ideas. These applications are evaluated in terms of users’ advantages, workflow improvements and impact over the overall quality of the output. In particular, they cover heterogeneous content management processes: from web editing to collaboration (IsaWiki and WikiFactory), from e-learning (IsaLearning) to professional printing (IsaPress).

Supervised and weakly supervised counting-by-segmentation: the fluorescent microscopy use case

This thesis focuses on automating the time-consuming task of manually counting activated neurons in fluorescent microscopy images, which is used to study the mechanisms underlying torpor. The traditional method of manual annotation can introduce bias and delay the outcome of experiments, so the author investigates a deep-learning-based procedure to automatize this task. The author explores two of the main convolutional-neural-network (CNNs) state-of-the-art architectures: UNet and ResUnet family model, and uses a counting-by-segmentation strategy to provide a justification of the objects considered during the counting process. The author also explores a weakly-supervised learning strategy that exploits only dot annotations. The author quantifies the advantages in terms of data reduction and counting performance boost obtainable with a transfer-learning approach and, specifically, a fine-tuning procedure. The author released the dataset used for the supervised use case and all the pre-training models, and designed a web application to share both the counting process pipeline developed in this work and the models pre-trained on the dataset analyzed in this work.

Analysis of human body kinematics using a hybrid markerless video acquisition and processing

A main objective of the human movement analysis is the quantitative description of joint kinematics and kinetics. This information may have great possibility to address clinical problems both in orthopaedics and motor rehabilitation. Previous studies have shown that the assessment of kinematics and kinetics from stereophotogrammetric data necessitates a setup phase, special equipment and expertise to operate. Besides, this procedure may cause feeling of uneasiness on the subjects and may hinder with their walking. The general aim of this thesis is the implementation and evaluation of new 2D markerless techniques, in order to contribute to the development of an alternative technique to the traditional stereophotogrammetric techniques. At first, the focus of the study has been the estimation of the ankle-foot complex kinematics during stance phase of the gait. Two particular cases were considered: subjects barefoot and subjects wearing ankle socks. The use of socks was investigated in view of the development of the hybrid method proposed in this work. Different algorithms were analyzed, evaluated and implemented in order to have a 2D markerless solution to estimate the kinematics for both cases. The validation of the proposed technique was done with a traditional stereophotogrammetric system. The implementation of the technique leads towards an easy to configure (and more comfortable for the subject) alternative to the traditional stereophotogrammetric system. Then, the abovementioned technique has been improved so that the measurement of knee flexion/extension could be done with a 2D markerless technique. The main changes on the implementation were on occlusion handling and background segmentation. With the additional constraints, the proposed technique was applied to the estimation of knee flexion/extension and compared with a traditional stereophotogrammetric system. Results showed that the knee flexion/extension estimation from traditional stereophotogrammetric system and the proposed markerless system were highly comparable, making the latter a potential alternative for clinical use. A contribution has also been given in the estimation of lower limb kinematics of the children with cerebral palsy (CP). For this purpose, a hybrid technique, which uses high-cut underwear and ankle socks as “segmental markers” in combination with a markerless methodology, was proposed. The proposed hybrid technique is different than the abovementioned markerless technique in terms of the algorithm chosen. Results showed that the proposed hybrid technique can become a simple and low-cost alternative to the traditional stereophotogrammetric systems.

Mitochondrial inheritance and germ line determination in a bivalve species with Doubly Uniparental Inheritance (DUI)

Mitochondria are inherited maternally in most metazoans. However, in some bivalves, two mitochondrial lineages are present: one transmitted through eggs (F), the other through sperm (M). This is called Doubly Uniparental Inheritance (DUI). During male embryo development, spermatozoon mitochondria aggregate and end up in the primordial germ cells, while they are dispersed in female embryos. The molecular mechanisms of segregation patterns are still unknown. In the DUI species Ruditapes philippinarum, I examined sperm mitochondria distribution by MitoTracker, microtubule staining and TEM, and I localized germ line determinants with immunocytochemical analysis. I also analyzed the gonad transcriptome, searching for genes involved in reproduction and sex determination. Moreover, I analyzed an M-type specific open reading frame that could be responsible for maintenance/degradation of M mitochondria during embryo development. These transcripts were also localized in tissues using in situ hybridization. As in Mytilus, two distribution patterns of M mitochondria were detected in R. philippinarum, supporting that they are related to DUI. Moreover, the first division midbody concurs in positioning aggregated M mitochondria on the animal-vegetal axis of the male embryo: in organisms with spiral segmentation this zone is not involved in further cleavages, so aggregation is maintained. Moreover, sperm mitochondria reach the same embryonic area where germ plasm is transferred, suggesting their contribution in male germ line formation. The finding of reproduction and ubiquitination transcripts led to formulate a model in which ubiquitination genes stored in female oocytes during gametogenesis would activate sex-gene expression in the early embryonic developmental stages (preformation). Only gametogenetic cells were labeled by in situ hybridization, proving their specific transcription in developing gametes. Other than having a role in sex determination, some ubiquination factors could also be involved in mitochondrial inheritance, and their differential expression could be responsible for the different fate of sperm mitochondria in the two sexes.

Methods for objective and interpretative evaluation of function and motor control: gait in perturbed condition and in the aquatic therapy

Gait analysis allows to characterize motor function, highlighting deviations from normal motor behavior related to an underlying pathology. The widespread use of wearable inertial sensors has opened the way to the evaluation of ecological gait, and a variety of methodological approaches and algorithms have been proposed for the characterization of gait from inertial measures (e.g. for temporal parameters, motor stability and variability, specific pathological alterations). However, no comparative analysis of their performance (i.e. accuracy, repeatability) was available yet, in particular, analysing how this performance is affected by extrinsic (i.e. sensor location, computational approach, analysed variable, testing environmental constraints) and intrinsic (i.e. functional alterations resulting from pathology) factors. The aim of the present project was to comparatively analyze the influence of intrinsic and extrinsic factors on the performance of the numerous algorithms proposed in the literature for the quantification of specific characteristics (i.e. timing, variability/stability) and alterations (i.e. freezing) of gait. Considering extrinsic factors, the influence of sensor location, analyzed variable, and computational approach on the performance of a selection of gait segmentation algorithms from a literature review was analysed in different environmental conditions (e.g. solid ground, sand, in water). Moreover, the influence of altered environmental conditions (i.e. in water) was analyzed as referred to the minimum number of stride necessary to obtain reliable estimates of gait variability and stability metrics, integrating what already available in the literature for over ground gait in healthy subjects. Considering intrinsic factors, the influence of specific pathological conditions (i.e. Parkinson’s Disease) was analyzed as affecting the performance of segmentation algorithms, with and without freezing. Finally, the analysis of the performance of algorithms for the detection of gait freezing showed how results depend on the domain of implementation and IMU position.

Robotic Sensing and Manipulation of Deformable Linear Objects with Learning-based methods

Nowadays robotic applications are widespread and most of the manipulation tasks are efficiently solved. However, Deformable-Objects (DOs) still represent a huge limitation for robots. The main difficulty in DOs manipulation is dealing with the shape and dynamics uncertainties, which prevents the use of model-based approaches (since they are excessively computationally complex) and makes sensory data difficult to interpret. This thesis reports the research activities aimed to address some applications in robotic manipulation and sensing of Deformable-Linear-Objects (DLOs), with particular focus to electric wires. In all the works, a significant effort was made in the study of an effective strategy for analyzing sensory signals with various machine learning algorithms. In the former part of the document, the main focus concerns the wire terminals, i.e. detection, grasping, and insertion. First, a pipeline that integrates vision and tactile sensing is developed, then further improvements are proposed for each module. A novel procedure is proposed to gather and label massive amounts of training images for object detection with minimal human intervention. Together with this strategy, we extend a generic object detector based on Convolutional-Neural-Networks for orientation prediction. The insertion task is also extended by developing a closed-loop control capable to guide the insertion of a longer and curved segment of wire through a hole, where the contact forces are estimated by means of a Recurrent-Neural-Network. In the latter part of the thesis, the interest shifts to the DLO shape. Robotic reshaping of a DLO is addressed by means of a sequence of pick-and-place primitives, while a decision making process driven by visual data learns the optimal grasping locations exploiting Deep Q-learning and finds the best releasing point. The success of the solution leverages on a reliable interpretation of the DLO shape. For this reason, further developments are made on the visual segmentation.

Variational and learning models for image and time series inverse problems

Inverse problems are at the core of many challenging applications. Variational and learning models provide estimated solutions of inverse problems as the outcome of specific reconstruction maps. In the variational approach, the result of the reconstruction map is the solution of a regularized minimization problem encoding information on the acquisition process and prior knowledge on the solution. In the learning approach, the reconstruction map is a parametric function whose parameters are identified by solving a minimization problem depending on a large set of data. In this thesis, we go beyond this apparent dichotomy between variational and learning models and we show they can be harmoniously merged in unified hybrid frameworks preserving their main advantages. We develop several highly efficient methods based on both these model-driven and data-driven strategies, for which we provide a detailed convergence analysis. The arising algorithms are applied to solve inverse problems involving images and time series. For each task, we show the proposed schemes improve the performances of many other existing methods in terms of both computational burden and quality of the solution. In the first part, we focus on gradient-based regularized variational models which are shown to be effective for segmentation purposes and thermal and medical image enhancement. We consider gradient sparsity-promoting regularized models for which we develop different strategies to estimate the regularization strength. Furthermore, we introduce a novel gradient-based Plug-and-Play convergent scheme considering a deep learning based denoiser trained on the gradient domain. In the second part, we address the tasks of natural image deblurring, image and video super resolution microscopy and positioning time series prediction, through deep learning based methods. We boost the performances of supervised, such as trained convolutional and recurrent networks, and unsupervised deep learning strategies, such as Deep Image Prior, by penalizing the losses with handcrafted regularization terms.

Embedded systems and advanced signal processing for Acousto- Ultrasonic Inspections

Non Destructive Testing (NDT) and Structural Health Monitoring (SHM) are becoming essential in many application contexts, e.g. civil, industrial, aerospace etc., to reduce structures maintenance costs and improve safety. Conventional inspection methods typically exploit bulky and expensive instruments and rely on highly demanding signal processing techniques. The pressing need to overcome these limitations is the common thread that guided the work presented in this Thesis. In the first part, a scalable, low-cost and multi-sensors smart sensor network is introduced. The capability of this technology to carry out accurate modal analysis on structures undergoing flexural vibrations has been validated by means of two experimental campaigns. Then, the suitability of low-cost piezoelectric disks in modal analysis has been demonstrated. To enable the use of this kind of sensing technology in such non conventional applications, ad hoc data merging algorithms have been developed. In the second part, instead, imaging algorithms for Lamb waves inspection (namely DMAS and DS-DMAS) have been implemented and validated. Results show that DMAS outperforms the canonical Delay and Sum (DAS) approach in terms of image resolution and contrast. Similarly, DS-DMAS can achieve better results than both DMAS and DAS by suppressing artefacts and noise. To exploit the full potential of these procedures, accurate group velocity estimations are required. Thus, novel wavefield analysis tools that can address the estimation of the dispersion curves from SLDV acquisitions have been investigated. An image segmentation technique (called DRLSE) was exploited in the k-space to draw out the wavenumber profile. The DRLSE method was compared with compressive sensing methods to extract the group and phase velocity information. The validation, performed on three different carbon fibre plates, showed that the proposed solutions can accurately determine the wavenumber and velocities in polar coordinates at multiple excitation frequencies.