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.

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.

Algoritmi di regolarizzazione con variazione totale nella ricostruzione di immagini di tomosintesi

La tomografia computerizzata (CT) è una metodica di diagnostica per immagini che consiste in una particolare applicazione dei raggi X. Essa, grazie ad una valutazione statistico-matematica (computerizzata) dell'assorbimento di tali raggi da parte delle strutture corporee esaminate, consente di ottenere immagini di sezioni assiali del corpo umano. Nonostante le dosi di radiazioni somministrate durante un esame siano contenute, questa tecnica rimane sconsigliata per pazienti sensibili, perciò nasce il tentativo della tomosintesi: ottenere lo stesso risultato della CT diminuendo le dosi somministrate. Infatti un esame di tomosintesi richiede poche proiezioni distribuite su un range angolare di appena 40°. Se da una parte la possibilità di una ingente diminuzione di dosi e tempi di somministrazione costituisce un grosso vantaggio dal punto di vista medico-diagnostico, dal punto di vista matematico esso comporta un nuovo problema di ricostruzione di immagini: infatti non è banale ottenere risultati validi come per la CT con un numero basso di proiezioni,cosa che va a infierire sulla mal posizione del problema di ricostruzione. Un possibile approccio al problema della ricostruzione di immagini è considerarlo un problema inverso mal posto e studiare tecniche di regolarizzazione opportune. In questa tesi viene discussa la regolarizzazione tramite variazione totale: verranno presentati tre algoritmi che implementano questa tecnica in modi differenti. Tali algoritmi verranno mostrati dal punto di vista matematico, dimostrandone ben posizione e convergenza, e valutati dal punto di vista qualitativo, attraverso alcune immagini ricostruite. Lo scopo è quindi stabilire se ci sia un metodo più vantaggioso e se si possano ottenere buoni risultati in tempi brevi, condizione necessaria per una applicazione diffusa della tomosintesi a livello diagnostico. Per la ricostruzione si è fatto riferimento a problemi-test cui è stato aggiunto del rumore, così da conoscere l'immagine originale e poter vedere quanto la ricostruzione sia ad essa fedele. Il lavoro principale dunque è stata la sperimentazione degli algoritmi su problemi test e la valutazione dei risultati, in particolare per gli algoritmi SGP e di Vogel, che in letteratura sono proposti per problemi di image deblurring.

Algoritmi per la stima del rumore nella ricostruzione di immagini

In questa tesi viene analizzato un metodo di regolarizzazione nel quale il problema di regolarizzazione è formulato come un problema di ottimizzazione vincolata. In questa formulazione è fondamentale il termine noto del vincolo, parametro per il quale dipende la bontà della ricostruzione, che corrisponde ad una stima della norma del rumore sul dato. Il contributo della tesi è l'analisi di un algoritmo proposto per stimare il rumore in applicazioni di image deblurring utilizzando la decomposizione in valori singolari dell'operatore che descrive il blur. Sono stati fatti numerosi test per valutare sia i parametri dell'algoritmo di stima che l'efficacia della stima ottenuta quando è inserita nel termine noto del vincolo del problema di ottimizzazione per la ricostruzione dell'immagine.

Tradução, adaptação cultural e validação do questionário Body image After Breast Cancer para a língua portuguesa do Brasil

Universidade Estadual de Campinas . Faculdade de Educação Física

Tradução, adaptação transcultural e validação do "Body Image Avoidance Questionnaire (BIAQ)" e do "Body Checking Questionnaire (BCQ)" para lingua portuguesa no Brasil

Universidade Estadual de Campinas . Faculdade de Educação Física

Juvenile idiopathic arthritis with involvement of the temporomandibular joint: the role of image examinations

PURPOSE: Juvenile idiopathic arthritis (JIA) has unknown etiology, and the involvement of the temporomandibular joint (TMJ) is rare in the early phase of the disease. The present article describes the use of computed tomography (CT) and magnetic resonance (MRI) images for the diagnosis of affected TMJ in JIA. CASE DESCRIPTION: A 12-year-old, female, Caucasian patient, with systemic rheumathoid arthritis and involvement of multiple joints was referred to the Imaging Center for TMJ assessment. The patient reported TMJ pain and limited opening of the mouth. The helical CT examination of the TMJ region showed asymmetric mandibular condyles, erosion of the right condyle and osteophyte-like formation. The MRI examination showed erosion of the right mandibular condyle, osteophytes, displacement without reduction and disruption of the articular disc. CONCLUSION: The disorders of the TMJ as a consequence of JIA must be carefully assessed by modern imaging methods such as CT and MRI. CT is very useful for the evaluation of discrete bone changes, which are not identified by conventional radiographs in the early phase of JIA. MRI allows the evaluation of soft tissues, the identification of acute articular inflammation and the differentiation between pannus and synovial hypertrophy.

Evaluation of follicular lymphoid depletion in the Bursa of Fabricius: an alternative methodology using digital image analysis and artificial neural networks

Fifty Bursa of Fabricius (BF) were examined by conventional optical microscopy and digital images were acquired and processed using Matlab® 6.5 software. The Artificial Neuronal Network (ANN) was generated using Neuroshell® Classifier software and the optical and digital data were compared. The ANN was able to make a comparable classification of digital and optical scores. The use of ANN was able to classify correctly the majority of the follicles, reaching sensibility and specificity of 89% and 96%, respectively. When the follicles were scored and grouped in a binary fashion the sensibility increased to 90% and obtained the maximum value for the specificity of 92%. These results demonstrate that the use of digital image analysis and ANN is a useful tool for the pathological classification of the BF lymphoid depletion. In addition it provides objective results that allow measuring the dimension of the error in the diagnosis and classification therefore making comparison between databases feasible.

Decision theory applied to image quality control in radiology

Background: The present work aims at the application of the decision theory to radiological image quality control ( QC) in diagnostic routine. The main problem addressed in the framework of decision theory is to accept or reject a film lot of a radiology service. The probability of each decision of a determined set of variables was obtained from the selected films. Methods: Based on a radiology service routine a decision probability function was determined for each considered group of combination characteristics. These characteristics were related to the film quality control. These parameters were also framed in a set of 8 possibilities, resulting in 256 possible decision rules. In order to determine a general utility application function to access the decision risk, we have used a simple unique parameter called r. The payoffs chosen were: diagnostic's result (correct/incorrect), cost (high/low), and patient satisfaction (yes/no) resulting in eight possible combinations. Results: Depending on the value of r, more or less risk will occur related to the decision-making. The utility function was evaluated in order to determine the probability of a decision. The decision was made with patients or administrators' opinions from a radiology service center. Conclusion: The model is a formal quantitative approach to make a decision related to the medical imaging quality, providing an instrument to discriminate what is really necessary to accept or reject a film or a film lot. The method presented herein can help to access the risk level of an incorrect radiological diagnosis decision.

Chrysotile effects on human lung cell carcinoma in culture: 3-D reconstruction and DNA quantification by image analysis

Background: Chrysotile is considered less harmful to human health than other types of asbestos fibers. Its clearance from the lung is faster and, in comparison to amphibole forms of asbestos, chrysotile asbestos fail to accumulate in the lung tissue due to a mechanism involving fibers fragmentation in short pieces. Short exposure to chrysotile has not been associated with any histopathological alteration of lung tissue. Methods: The present work focuses on the association of small chrysotile fibers with interphasic and mitotic human lung cancer cells in culture, using for analyses confocal laser scanning microscopy and 3D reconstructions. The main goal was to perform the analysis of abnormalities in mitosis of fibers-containing cells as well as to quantify nuclear DNA content of treated cells during their recovery in fiber-free culture medium. Results: HK2 cells treated with chrysotile for 48 h and recovered in additional periods of 24, 48 and 72 h in normal medium showed increased frequency of multinucleated and apoptotic cells. DNA ploidy of the cells submitted to the same chrysotile treatment schedules showed enhanced aneuploidy values. The results were consistent with the high frequency of multipolar spindles observed and with the presence of fibers in the intercellular bridge during cytokinesis. Conclusion: The present data show that 48 h chrysotile exposure can cause centrosome amplification, apoptosis and aneuploid cell formation even when long periods of recovery were provided. Internalized fibers seem to interact with the chromatin during mitosis, and they could also interfere in cytokinesis, leading to cytokinesis failure which forms aneuploid or multinucleated cells with centrosome amplification.

Comparison of multispectral wide-field optical imaging modalities to maximize image contrast for objective discrimination of oral neoplasia

Multispectral widefield optical imaging has the potential to improve early detection of oral cancer. The appropriate selection of illumination and collection conditions is required to maximize diagnostic ability. The goals of this study were to (i) evaluate image contrast between oral cancer/precancer and non-neoplastic mucosa for a variety of imaging modalities and illumination/collection conditions, and (ii) use classification algorithms to evaluate and compare the diagnostic utility of these modalities to discriminate cancers and precancers from normal tissue. Narrowband reflectance, autofluorescence, and polarized reflectance images were obtained from 61 patients and 11 normal volunteers. Image contrast was compared to identify modalities and conditions yielding greatest contrast. Image features were extracted and used to train and evaluate classification algorithms to discriminate tissue as non-neoplastic, dysplastic, or cancer; results were compared to histologic diagnosis. Autofluorescence imaging at 405-nm excitation provided the greatest image contrast, and the ratio of red-to-green fluorescence intensity computed from these images provided the best classification of dysplasia/cancer versus non-neoplastic tissue. A sensitivity of 100% and a specificity of 85% were achieved in the validation set. Multispectral widefield images can accurately distinguish neoplastic and non-neoplastic tissue; however, the ability to separate precancerous lesions from cancers with this technique was limited. (C) 2010 Society of Photo-Optical Instrumentation Engineers. [DOI: 10.1117/1.3516593]

Improvements on ICA mixture models for image pre-processing and segmentation

Today several different unsupervised classification algorithms are commonly used to cluster similar patterns in a data set based only on its statistical properties. Specially in image data applications, self-organizing methods for unsupervised classification have been successfully applied for clustering pixels or group of pixels in order to perform segmentation tasks. The first important contribution of this paper refers to the development of a self-organizing method for data classification, named Enhanced Independent Component Analysis Mixture Model (EICAMM), which was built by proposing some modifications in the Independent Component Analysis Mixture Model (ICAMM). Such improvements were proposed by considering some of the model limitations as well as by analyzing how it should be improved in order to become more efficient. Moreover, a pre-processing methodology was also proposed, which is based on combining the Sparse Code Shrinkage (SCS) for image denoising and the Sobel edge detector. In the experiments of this work, the EICAMM and other self-organizing models were applied for segmenting images in their original and pre-processed versions. A comparative analysis showed satisfactory and competitive image segmentation results obtained by the proposals presented herein. (C) 2008 Published by Elsevier B.V.