Studio geomorfologico dell'area carsica attorno alla Grotta Palombara (Melilli, Siracusa)

The following thesis attempts to study and analyse the geomorphological evolution of a relatively small coastal area located to the North of Syracuse (Southeastern Sicily). The presently inactive Palombara Cave is located in this area. The 800 metres of passages in this cave show an evolution in some way linked to the local topographic and environmental changes. This portion of coastline was affected more or less constantly by the tectonic uplift during the Pleistocene, which simultaneously to the eustatic variations have played a key role in the genesis of the marine terraces and the cave. Starting from a DTM made from Lidar data, using a GIS procedure several marine terraces have been mapped. These informations combinated with a geomorphological study of the area, allowed to identify and recognise the different orders of the Middle Pleistocene terraced surfaces. Four orders of terraces between 180-75 m a.s.l have been observed, illustrated and described. Furthermore, two other supposed terrace edges located respectively at 60 and 35 m, which would indicate the presence of two more orders, have been recognised. All these marine terraces appear to have formed in the last million years. The morphological data of the Palombara cave, highlights a genesis related to the rising of CO2 rich waters coming from the depths through the fractures of the rock mass, that ranks it as a hypogenic cave. The development has been influenced by the changes in the water table, in turn determined by the fluctuations in the sea level. In fact, the cave shows a speleogenetic evolution characterised by phases of karstification in phreatic and epiphreatic environment and fossilization stages of the upper branches in vadose conditions. These observations indicate that the cave probably started forming around 600 Ky ago, contemporary to the start of volcanic processes in the area.

A novel marker for heart failure: Cross sectional area assessment in a large vein by conductance catheter.

The present study investigates the feasibility of a new application able to check the heart failure status in a patient through the estimation of the venous distension. In this way it would be possible to follow up patients, avoiding invasive or expensive exams such as cardiac catheterization and echocardiography. Moreover, the devices would also be able to diagnose the decline of the disease, in order to allow a new adaptation to therapy, and vice versa to check the improvement in the patient’s conditions after the CRT device implant. This thesis is essentially divided into three parts: an analytical model was used to obtain an estimation of the error committed for the calculation of the CSA and to understand how the accuracy and sensitivity depend on the different configurations of the electrodes and the catheter position inside the vein; secondly, an in-vitro experiment was carried out in order to verify the practical feasibility for these kinds of measurements, in a very simplified model; in the end, several animal experiments were done to test the in-vivo practicability of the proposed method. The obtained results showed the feasibility of this approach. In fact, the error committed in the estimation of CSA, during the animal experiments, can be considered acceptable (CSAerror_max ≈ -14%). Moreover, it has been demonstrated that the conductance catheter allows assessing, not only the vein CSA, but also the breathing of the animal.

Depth Estimation in Stereo Biomedical Images via Proxy-Supervised Deep Learning

In order to estimate depth through supervised deep learning-based stereo methods, it is necessary to have access to precise ground truth depth data. While the gathering of precise labels is commonly tackled by deploying depth sensors, this is not always a viable solution. For instance, in many applications in the biomedical domain, the choice of sensors capable of sensing depth at small distances with high precision on difficult surfaces (that present non-Lambertian properties) is very limited. It is therefore necessary to find alternative techniques to gather ground truth data without having to rely on external sensors. In this thesis, two different approaches have been tested to produce supervision data for biomedical images. The first aims to obtain input stereo image pairs and disparities through simulation in a virtual environment, while the second relies on a non-learned disparity estimation algorithm in order to produce noisy disparities, which are then filtered by means of hand-crafted confidence measures to create noisy labels for a subset of pixels. Among the two, the second approach, which is referred in literature as proxy-labeling, has shown the best results and has even outperformed the non-learned disparity estimation algorithm used for supervision.

Mitigating non-Lambertian surfaces issues in Stereo Matching with Neural Radiance Fields

Depth estimation from images has long been regarded as a preferable alternative compared to expensive and intrusive active sensors, such as LiDAR and ToF. The topic has attracted the attention of an increasingly wide audience thanks to the great amount of application domains, such as autonomous driving, robotic navigation and 3D reconstruction. Among the various techniques employed for depth estimation, stereo matching is one of the most widespread, owing to its robustness, speed and simplicity in setup. Recent developments has been aided by the abundance of annotated stereo images, which granted to deep learning the opportunity to thrive in a research area where deep networks can reach state-of-the-art sub-pixel precision in most cases. Despite the recent findings, stereo matching still begets many open challenges, two among them being finding pixel correspondences in presence of objects that exhibits a non-Lambertian behaviour and processing high-resolution images. Recently, a novel dataset named Booster, which contains high-resolution stereo pairs featuring a large collection of labeled non-Lambertian objects, has been released. The work shown that training state-of-the-art deep neural network on such data improves the generalization capabilities of these networks also in presence of non-Lambertian surfaces. Regardless being a further step to tackle the aforementioned challenge, Booster includes a rather small number of annotated images, and thus cannot satisfy the intensive training requirements of deep learning. This thesis work aims to investigate novel view synthesis techniques to augment the Booster dataset, with ultimate goal of improving stereo matching reliability in presence of high-resolution images that displays non-Lambertian surfaces.