Claudia Testa

Magnetic Resonance Fingerprinting (MRF) is a fast quantitative MR Imaging technique that provides multi-parametric maps with a single acquisition. Neural Networks (NNs) accelerate reconstruction but require significant resources for training. We propose an FPGA-based NN for real-time brain parameter reconstruction from MRF data. Training the NN takes an estimated 200 seconds, significantly faster than standard CPU-based training, which can be up to 250 times slower. This method could enable real-time brain analysis on mobile devices, revolutionizing clinical decision-making and telemedicine.

Articular cartilage (AC) is a specialized connective tissue that covers the ends of long bones and facilitates the load‐bearing of joints. It consists of chondrocytes distributed throughout an extracellular matrix and organized into three zones: superficial, middle, and deep. Nuclear magnetic resonance (NMR) techniques can be used to characterize this layered structure. In this study, devoted to a better understanding of the NMR response of this complex tissue, 20 specimens excised from femoral and tibial cartilage of bovine samples were analyzed by the low‐field single‐sided NMR‐MOUSE‐PM10. A multiparametric depth‐wise analysis was performed to characterize the laminar structure of AC and investigate the origin of the NMR dependence on depth. The depth dependence of the single parameters T₁, T₂, and D has been described in literature, but their simultaneous measurement has not been fully exploited yet, as well as the extent of their variability. A novel parameter, α, evaluated by applying a double‐quantum‐like sequence, has been measured. The significant decrease in T₁, T₂, and D from the middle to the deep zone is consistent with depth‐dependent composition and structure changes of the complex matrix of fibers confining and interacting with water. The α parameter appears to be a robust marker of the layered structure with a well‐reproducible monotonic trend across the zones. The discrimination of cartilage zones was reinforced by a multivariate principal component analysis statistical analysis. The large number of samples allowed for the identification of the smallest number of parameters or their combination able to classify samples. The first two components clustered the data according to the different zones, highlighting the sensitivity of the NMR parameters to the structural and compositional variations of AC. Using two parameters, the best result was obtained by considering T₁ and α. Single‐sided NMR devices, portable and low‐cost, provide information on NMR parameters related to tissue composition and structure.

Deep learning (DL) networks have shown attractive performance in medical image processing tasks such as brain tumor classification. However, they are often criticized as mysterious “black boxes”. The opaqueness of the model and the reasoning process make it difficult for health workers to decide whether to trust the prediction outcomes. In this study, we develop an interpretable multi-part attention network (IMPA-Net) for brain tumor classification to enhance the interpretability and trustworthiness of classification outcomes. The proposed model not only predicts the tumor grade but also provides a global explanation for the model interpretability and a local explanation as justification for the proffered prediction. Global explanation is represented as a group of feature patterns that the model learns to distinguish high-grade glioma (HGG) and low-grade glioma (LGG) classes. Local explanation interprets the reasoning process of an individual prediction by calculating the similarity between the prototypical parts of the image and a group of pre-learned task-related features. Experiments conducted on the BraTS2017 dataset demonstrate that IMPA-Net is a verifiable model for the classification task. A percentage of 86% of feature patterns were assessed by two radiologists to be valid for representing task-relevant medical features. The model shows a classification accuracy of 92.12%, of which 81.17% were evaluated as trustworthy based on local explanations. Our interpretable model is a trustworthy model that can be used for decision aids for glioma classification. Compared with black-box CNNs, it allows health workers and patients to understand the reasoning process and trust the prediction outcomes.

The Enhanced-Deep-Super-Resolution (EDSR) model is a state-of-the-art convolutional neural network suitable for improving image spatial resolution. It was previously trained with general-purpose pictures and then, in this work, tested on biomedical magnetic resonance (MR) images, comparing the network outcomes with traditional up-sampling techniques. We explored possible changes in the model response when different MR sequences were analyzed. T₁w and T₂w MR brain images of 70 human healthy subjects (F:M, 40:30) from the Cambridge Centre for Ageing and Neuroscience (Cam-CAN) repository were down-sampled and then up-sampled using EDSR model and BiCubic (BC) interpolation. Several reference metrics were used to quantitatively assess the performance of up-sampling operations (RMSE, pSNR, SSIM, and HFEN). Two-dimensional and three-dimensional reconstructions were evaluated. Different brain tissues were analyzed individually. The EDSR model was superior to BC interpolation on the selected metrics, both for two- and three- dimensional reconstructions. The reference metrics showed higher quality of EDSR over BC reconstructions for all the analyzed images, with a significant difference of all the criteria in T₁w images and of the perception-based SSIM and HFEN in T₂w images. The analysis per tissue highlights differences in EDSR performance related to the gray-level values, showing a relative lack of outperformance in reconstructing hyperintense areas. The EDSR model, trained on general-purpose images, better reconstructs MR T₁w and T₂w images than BC, without any retraining or fine-tuning. These results highlight the excellent generalization ability of the network and lead to possible applications on other MR measurements.

Multiple Sclerosis (MS) is an autoimmune demyelinating disease characterised by changes in iron and myelin content. These biomarkers are detectable by Quantitative Susceptibility Mapping (QSM), an advanced Magnetic Resonance Imaging technique detecting magnetic properties. When analysed with radiomic techniques that exploit its intrinsic quantitative nature, QSM may furnish biomarkers to facilitate early diagnosis of MS and timely assessment of progression. In this work, we explore the robustness of QSM radiomic features by varying the number of grey levels (GLs) and echo times (TEs), in a sample of healthy controls and patients with MS. We analysed the white matter in total and within six clinically relevant tracts, including the cortico-spinal tract and the optic radiation. After optimising the number of GLs (n = 64), at least 65% of features were robust for each Volume of Interest (VOI), with no difference (p > .05) between left and right hemispheres. Different outcomes in feature robustness among the VOIs depend on their characteristics, such as volume and variance of susceptibility values. This study validated the processing pipeline for robustness analysis and established the reliability of QSM-based radiomics features against GLs and TEs. Our results provide important insights for future radiomics studies using QSM in clinical applications.

Myotonic dystrophy type 1 (DM1) is a genetic disorder caused by a (CTG) expansion in the DM protein kinase (DMPK) gene, representing the most common adult muscular dystrophy, characterized by a multisystem involvement with predominantly skeletal muscle and brain affection. Neuroimaging studies showed widespread white matter changes and brain atrophy in DM1, but only a few studies investigated the role of white matter metabolism in the pathophysiology of central nervous system impairment. We aim to reveal the relationship between the metabolic profile of parieto-occipital white matter (POWM) as evaluated with proton MR spectroscopy technique, with the visuoperceptual and visuoconstructional dysfunctions in DM1 patients. MR spectroscopy (3 Tesla) and neuropsychological evaluations were performed in 34 DM1 patients (19 F, age: 46.4 ± 12.1 years, disease duration: 18.7 ± 11.6 years). The content of neuro-axonal marker N-acetyl-aspartate, both relative to Creatine (NAA/Cr) and to myo-Inositol (NAA/mI) resulted significantly lower in DM1 patients compared to HC (p-values < 0.0001). NAA/Cr and NAA/mI correlated with the copy of the Rey-Osterrieth complex figure (r = 0.366, p = 0.033; r = 0.401, p = 0.019, respectively) and with Street’s completion tests scores (r = 0.409, p = 0.016; r = 0.341, p = 0.048 respectively). The proportion of white matter hyperintensities within the MR spectroscopy voxel did not correlate with the metabolite content. In this study, POWM metabolic alterations in DM1 patients were not associated with the white matter morphological changes and correlated with specific neuropsychological deficits.