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.

Simulation of optical observables for space missions to small bodies

The navigation of deep space spacecraft requires accurate measurement of the probe’s state and attitude with respect to a body whose ephemerides may not be known with good accuracy. The heliocentric state of the spacecraft is estimated through radiometric techniques (ranging, Doppler, and Delta-DOR), while optical observables can be introduced to improve the uncertainty in the relative position and attitude with respect to the target body. In this study, we analyze how simulated optical observables affect the estimation of parameters in an orbit determination problem, considering the case of the ESA’s Hera mission towards the binary asteroid system composed of Didymos and Dimorphos. To this extent, a shape model and a photometric function are used to create synthetic onboard camera images. Then, using a stereophotoclinometry technique on some of the simulated images, we create a database of maplets that describe the 3D geometry of the surface around a set of landmarks. The matching of maplets with the simulated images provides the optical observables, expressed as pixel coordinates in the camera frame, which are fed to an orbit determination filter to estimate a certain number of solve-for parameters. The noise introduced in the output optical observables by the image processing can be quantified using as a metric the quality of the residuals, which is used to fine-tune the maplet-matching parameters. In particular, the best results are obtained when using small maplets, with high correlation coefficients and occupation factors.

Content-aware approach for improving biomedical image analysis: an interdisciplinary study series

Biomedicine is a highly interdisciplinary research area at the interface of sciences, anatomy, physiology, and medicine. In the last decade, biomedical studies have been greatly enhanced by the introduction of new technologies and techniques for automated quantitative imaging, thus considerably advancing the possibility to investigate biological phenomena through image analysis. However, the effectiveness of this interdisciplinary approach is bounded by the limited knowledge that a biologist and a computer scientist, by professional training, have of each other’s fields. The possible solution to make up for both these lacks lies in training biologists to make them interdisciplinary researchers able to develop dedicated image processing and analysis tools by exploiting a content-aware approach. The aim of this Thesis is to show the effectiveness of a content-aware approach to automated quantitative imaging, by its application to different biomedical studies, with the secondary desirable purpose of motivating researchers to invest in interdisciplinarity. Such content-aware approach has been applied firstly to the phenomization of tumour cell response to stress by confocal fluorescent imaging, and secondly, to the texture analysis of trabecular bone microarchitecture in micro-CT scans. Third, this approach served the characterization of new 3-D multicellular spheroids of human stem cells, and the investigation of the role of the Nogo-A protein in tooth innervation. Finally, the content-aware approach also prompted to the development of two novel methods for local image analysis and colocalization quantification. In conclusion, the content-aware approach has proved its benefit through building new approaches that have improved the quality of image analysis, strengthening the statistical significance to allow unveiling biological phenomena. Hopefully, this Thesis will contribute to inspire researchers to striving hard for pursuing interdisciplinarity.

Regularization meets GreenAI: a new framework for image reconstruction in life sciences applications

Ill-conditioned inverse problems frequently arise in life sciences, particularly in the context of image deblurring and medical image reconstruction. These problems have been addressed through iterative variational algorithms, which regularize the reconstruction by adding prior knowledge about the problem's solution. Despite the theoretical reliability of these methods, their practical utility is constrained by the time required to converge. Recently, the advent of neural networks allowed the development of reconstruction algorithms that can compute highly accurate solutions with minimal time demands. Regrettably, it is well-known that neural networks are sensitive to unexpected noise, and the quality of their reconstructions quickly deteriorates when the input is slightly perturbed. Modern efforts to address this challenge have led to the creation of massive neural network architectures, but this approach is unsustainable from both ecological and economic standpoints. The recently introduced GreenAI paradigm argues that developing sustainable neural network models is essential for practical applications. In this thesis, we aim to bridge the gap between theory and practice by introducing a novel framework that combines the reliability of model-based iterative algorithms with the speed and accuracy of end-to-end neural networks. Additionally, we demonstrate that our framework yields results comparable to state-of-the-art methods while using relatively small, sustainable models. In the first part of this thesis, we discuss the proposed framework from a theoretical perspective. We provide an extension of classical regularization theory, applicable in scenarios where neural networks are employed to solve inverse problems, and we show there exists a trade-off between accuracy and stability. Furthermore, we demonstrate the effectiveness of our methods in common life science-related scenarios. In the second part of the thesis, we initiate an exploration extending the proposed method into the probabilistic domain. We analyze some properties of deep generative models, revealing their potential applicability in addressing ill-posed inverse problems.

Generative Neural Networks for image iper-resolution and improvement of Optical Character Recognition's performances

The usage of Optical Character Recognition’s (OCR, systems is a widely spread technology into the world of Computer Vision and Machine Learning. It is a topic that interest many field, for example the automotive, where becomes a specialized task known as License Plate Recognition, useful for many application from the automation of toll road to intelligent payments. However, OCR systems need to be very accurate and generalizable in order to be able to extract the text of license plates under high variable conditions, from the type of camera used for acquisition to light changes. Such variables compromise the quality of digitalized real scenes causing the presence of noise and degradation of various type, which can be minimized with the application of modern approaches for image iper resolution and noise reduction. Oneclass of them is known as Generative Neural Networks, which are very strong ally for the solution of this popular problem.

Feasibility And Reproducibility Of Diffusion-weighted Magnetic Resonance Imaging Of The Fetal Brain In Twin-twin Transfusion Syndrome.

The aim of this study is to test the feasibility and reproducibility of diffusion-weighted magnetic resonance imaging (DW-MRI) evaluations of the fetal brains in cases of twin-twin transfusion syndrome (TTTS). From May 2011 to June 2012, 24 patients with severe TTTS underwent MRI scans for evaluation of the fetal brains. Datasets were analyzed offline on axial DW images and apparent diffusion coefficient (ADC) maps by two radiologists. The subjective evaluation was described as the absence or presence of water diffusion restriction. The objective evaluation was performed by the placement of 20-mm(2) circular regions of interest on the DW image and ADC maps. Subjective interobserver agreement was assessed by the kappa correlation coefficient. Objective intraobserver and interobserver agreements were assessed by proportionate Bland-Altman tests. Seventy-four DW-MRI scans were performed. Sixty of them (81.1%) were considered to be of good quality. Agreement between the radiologists was 100% for the absence or presence of diffusion restriction of water. For both intraobserver and interobserver agreement of ADC measurements, proportionate Bland-Altman tests showed average percentage differences of less than 1.5% and 95% CI of less than 18% for all sites evaluated. Our data demonstrate that DW-MRI evaluation of the fetal brain in TTTS is feasible and reproducible.

[association Between Physical And Psychosocial Impacts Of Oral Disorders And Quality Of Life Among The Elderly].

This study sought to evaluate the association between the impact of oral disorders in terms of physical/psychosocial dimensions and quality of life among the elderly. It involved a cross-sectional study conducted among the elderly (65-74 years) in 2008/2009. The social impact was assessed using the Oral Health Impact Profile (OHIP 14) and the quality of life using the SF 12 Short-Form Health Survey. Descriptive, univariate and multivariate (logistic regression) analysis was conducted with correction for the design effect, using SPSS(r)18.0 software. Of the 800 individuals approached, 736 elderly individuals participated (TR = 92%), with a mean age of 67.77 years, the majority of whom showed no impact based on the measurement of the prevalence of OHIP. The functional limitation dimension of the OHIP was associated with the physical domain of the SF12, irrespective of the other variables investigated. However, the seriousness of OHIP and its psychological discomfort and disability dimensions was associated with the mental domain of the SF12. The conclusion reached is that some impacts of oral disorders were associated with unsatisfactory quality of life in the physical and mental domains.

Applying The Maternal Near Miss Approach For The Evaluation Of Quality Of Obstetric Care: A Worked Example From A Multicenter Surveillance Study.

To assess quality of care of women with severe maternal morbidity and to identify associated factors. This is a national multicenter cross-sectional study performing surveillance for severe maternal morbidity, using the World Health Organization criteria. The expected number of maternal deaths was calculated with the maternal severity index (MSI) based on the severity of complication, and the standardized mortality ratio (SMR) for each center was estimated. Analyses on the adequacy of care were performed. 17 hospitals were classified as providing adequate and 10 as nonadequate care. Besides almost twofold increase in maternal mortality ratio, the main factors associated with nonadequate performance were geographic difficulty in accessing health services (P < 0.001), delays related to quality of medical care (P = 0.012), absence of blood derivatives (P = 0.013), difficulties of communication between health services (P = 0.004), and any delay during the whole process (P = 0.039). This is an example of how evaluation of the performance of health services is possible, using a benchmarking tool specific to Obstetrics. In this study the MSI was a useful tool for identifying differences in maternal mortality ratios and factors associated with nonadequate performance of care.

[diet Quality And Associated Factors Among The Elderly: A Population-based Study In Campinas, São Paulo State, Brazil].

The aim of this study was to assess the quality of diet among the elderly and associations with socio-demographic variables, health-related behaviors, and diseases. A population-based cross-sectional study was conducted in a representative sample of 1,509 elderly participants in a health survey in Campinas, São Paulo State, Brazil. Food quality was assessed using the Revised Diet Quality Index (DQI-R). Mean index scores were estimated and a multiple regression model was employed for the adjusted analyses. The highest diet quality scores were associated with age 80 years or older, Evangelical religion, diabetes mellitus, and physical activity, while the lowest scores were associated with home environments shared with three or more people, smoking, and consumption of soft drinks and alcoholic beverages. The findings emphasize a general need for diet quality improvements in the elderly, specifically in subgroups with unhealthy behaviors, who should be targeted with comprehensive strategies.

Mastication Movements And Sleep Quality Of Patients With Myofascial Pain: Occlusal Device Therapy Improvements.

Patients with myofascial pain experience impaired mastication, which might also interfere with their sleep quality. The purpose of this study was to evaluate the jaw motion and sleep quality of patients with myofascial pain and the impact of a stabilization device therapy on both parameters. Fifty women diagnosed with myofascial pain by the Research Diagnostic Criteria were enrolled. Pain levels (visual analog scale), jaw movements (kinesiography), and sleep quality (Epworth Sleepiness Scale; Pittsburgh Sleep Quality Index) were evaluated before (control) and after stabilization device use. Range of motion (maximum opening, right and left excursions, and protrusion) and masticatory movements during Optosil mastication (opening, closing, and total cycle time; opening and closing angles; and maximum velocity) also were evaluated. Repeated-measures analysis of variance in a generalized linear mixed models procedure was used for statistical analysis (α=.05). At baseline, participants with myofascial pain showed a reduced range of jaw motion and poorer sleep quality. Treatment with a stabilization device reduced pain (P<.001) and increased both mouth opening (P<.001) and anteroposterior movement (P=.01). Also, after treatment, the maximum opening (P<.001) and closing (P=.04) velocities during mastication increased, and improvements in sleep scores for the Pittsburgh Sleep Quality Index (P<.001) and Epworth Sleepiness Scale (P=.04) were found. Myofascial pain impairs jaw motion and quality of sleep; the reduction of pain after the use of a stabilization device improves the range of motion and sleep parameters.

[larynx Cancer: Quality Of Life And Voice After Treatment].

Treatments for patients with laryngeal cancer often have an impact on physical, social, and psychological functions. To evaluate quality of life and voice in patients treated for advanced laryngeal cancer through surgery or exclusive chemoradiation. Retrospective cohort study with 30 patients free from disease: ten total laryngectomy patients without production of esophageal speech (ES); ten total laryngectomy patients with tracheoesophageal speech (TES), and ten with laryngeal speech. Quality of life was measured by SF-36, Voice-Related Quality of Life (V-RQOL), and Voice Handicap Index (VHI) protocols, applied on the same day. The SF-36 showed that patients who received exclusive chemoradiotherapy had better quality of life than the TES and ES groups. The V-RQOL showed that the voice-related quality of life was lower in the ES group. In the VHI, the ES group showed higher scores for overall, emotional, functional, and organic VHI. Quality of life and voice in patients treated with chemoradiotherapy was better than in patients treated surgically. The type of medical treatment used in patients with laryngeal cancer can bring changes in quality of life and voice.

Pain Threshold And Sleep Quality In Women With Endometriosis.

Pain is a common complaint in women with endometriosis and can be influenced by many variables, including sleep disorders; however, no data are available on the sleep quality of women with endometriosis or on the correlation between sleep quality and pain. The 510 volunteers included in this study were divided into two groups: 257 women with a laparoscopic and histopathological diagnosis of endometriosis and 253 women with no history of endometriosis and no endometriosis-related symptoms. The volunteers answered two questionnaires: the Post-Sleep Inventory to evaluate sleep quality and the International Physical Activity Questionnaire to assess their level of physical activity. Pain was evaluated using a visual analogue scale (VAS) and women were also submitted to a physical examination, during which their pain threshold was assessed at 20 different body sites. Sleep quality was significantly poorer in women with endometriosis compared to women without the disease. The pain threshold was significantly lower in the greater trochanter and abdomen in women with endometriosis when compared to women without the disease; however, there was no difference in VAS pain score between the groups. The higher the VAS pain score, the lower the Post-Sleep Inventory score. Additionally, there was a significant positive correlation between the pain threshold at some body sites and sleep quality. Sleep quality was poorer and the pain threshold at certain body sites was lower in the group of women with endometriosis.

Transcutaneous Tibial Nerve Stimulation In The Treatment Of Lower Urinary Tract Symptoms And Its Impact On Health-related Quality Of Life In Patients With Parkinson Disease: A Randomized Controlled Trial.

A randomized controlled trial study was performed to evaluate the efficacy of transcutaneous tibial nerve stimulation (TTNS) and sham TTNS, in patients with Parkinson disease (PD) with lower urinary tract symptoms (LUTS). Randomized controlled trial. Thirteen patients with a diagnosis of PD and bothersome LUTS were randomly allocated to one of the following groups: Group I: TTNS group (n = 8) and group II: Sham group (n = 5). Both groups attended twice a week during 5 weeks; each session lasted 30 minutes. Eight patients received TTNS treatment and 5 subjects allocated to group II were managed with sham surface electrodes that delivered no electrical stimulation. Assessments were performed before and after the treatment; they included a 3-day bladder diary, Overactive Bladder Questionnaire (OAB-V8), and the International Consultation on Incontinence Quality of Life Questionnaire Short Form (ICIQ-SF), and urodynamic evaluation. Following 5 weeks of treatment, patients allocated to TTNS demonstrated statistically significant reductions in the number of urgency episodes (P = .004) and reductions in nocturia episodes (P < .01). Participants allocated to active treatment also showed better results after treatment in the OAB-V8 and ICIQ-SF scores (P < .01, respectively). Urodynamic testing revealed that patients in the active treatment group showed improvements in intravesical volume at strong desire to void (P < .05) and volume at urgency (P < .01) when compared to subjects in the sham treatment group. These findings suggest that TTNS is effective in the treatment of LUTS in patients with PD, reducing urgency and nocturia episodes and improving urodynamic parameters as well as symptom scores measured by the OAB-V8 and health-related quality-of-life scores measured by the ICIQ-SF.

Nausea, Vomiting And Quality Of Life Of Patients With Cancer Undergoing Antineoplastic Treatment: An Evaluation By Pharmacists.

This study aims to evaluate the frequency and severity of nausea and vomiting using two different instruments and relate them to quality of life (QOL) in patients with cancer receiving antineoplastic treatment. Severity of chemotherapy-induced nausea and vomiting (CINV) was measured by Common Terminology Criteria for Adverse Events (CTCAE) and a numerical scale. QOL was assessed using the Functional Assessment of Cancer Therapy-General questionnaire. Of the 50 patients studied, 60.0% reported nausea (40.0% CTCAE grade 1; 66.7% moderate intensity on numerical scale) and 30.0% reported vomiting (46.7% CTCAE grades 1 and 2, each; 66.7% moderate intensity on numerical scale). CINV did not influence overall QOL. The frequency of CINV was high. There was no association between nausea/vomiting and overall QOL.

Medication Adherence And Quality Of Life Among The Elderly With Diabetic Retinopathy.

one hundred (n=100) elderly outpatients with diabetic retinopathy taking antihypertensives and/or oral antidiabetics/insulin were interviewed. Adherence was evaluated by the adherence proportion and its association with the care taken in administrating medications and by the Morisky Scale. The National Eye Institute Visual Functioning Questionnaire (NEI VFQ-25) was used to evaluate HRQoL. most (58%) reported the use of 80% or more of the prescribed dose and care in utilizing the medication. The item stopping the drug when experiencing an adverse event, from the Morisky Scale, explained 12.8% and 13.5% of the variability of adherence proportion to antihypertensives and oral antidiabetics/insulin, respectively. there was better HRQoL in the Color Vision, Driving and Social Functioning domains of the NEI VFQ-25. Individuals with lower scores on the NEI VFQ-25 and higher scores on the Morisky Scale presented greater chance to be nonadherent to the pharmacological treatment of diabetes and hypertension.