Deep learning for automatic segmentation of thigh and leg muscles.

OBJECTIVE: In this study we address the automatic segmentation of selected muscles of the thigh and leg through a supervised deep learning approach. MATERIAL AND METHODS: The application of quantitative imaging in neuromuscular diseases requires the availability of regions of interest (ROI) drawn on muscles to extract quantitative parameters. Up to now, manual drawing of ROIs has been considered the gold standard in clinical studies, with no clear and universally accepted standardized procedure for segmentation. Several automatic methods, based mainly on machine learning and deep learning algorithms, have recently been proposed to discriminate between skeletal muscle, bone, subcutaneous and intermuscular adipose tissue. We develop a supervised deep learning approach based on a unified framework for ROI segmentation. RESULTS: The proposed network generates segmentation maps with high accuracy, consisting in Dice Scores ranging from 0.89 to 0.95, with respect to "ground truth" manually segmented labelled images, also showing high average performance in both mild and severe cases of disease involvement (i.e. entity of fatty replacement). DISCUSSION: The presented results are promising and potentially translatable to different skeletal muscle groups and other MRI sequences with different contrast and resolution.

Radiomics Multifactorial in Silico Model for Spatial Prediction of Glioblastoma Progression and Recurrence: A Proof-of-Concept.

BACKGROUND: Radiomics-based prediction of glioblastoma spatial progression and recurrence may improve personalized strategies. However, most prototypes are based on limited monofactorial Gompertzian models of tumor growth. The present study consists of a proof of concept on the accuracy of a radiomics multifactorial in silico model in predicting short-term spatial growth and recurrence of glioblastoma. METHODS: A radiomics-based biomathematical multifactorial in silico model was developed using magnetic resonance imaging (MRI) data from a 53-year-old patient with newly diagnosed glioblastoma of the right supramarginal gyrus. Raw and optimized models were derived from the MRI at diagnosis and matched to the preoperative MRI obtained 28 days after diagnosis to test the accuracy in predicting the short-term spatial growth of the tumor. An additional optimized model was derived from the early postoperative MRI and matched to the MRI documenting tumor recurrence to test spatial accuracy in predicting the location of recurrence. The spatial prediction accuracy of the model was reported as an average Jaccard index. RESULTS: Optimized models yielded an average Jaccard index of 0.69 and 0.26 for short-term tumor growth and long-term recurrence site, respectively. CONCLUSIONS: The present radiomics-based multifactorial in silico model was feasible, reliable, and accurate for short-term spatial prediction of glioblastoma progression. The predictive value for the spatial location of recurrence was still low, and refinements in the description of tissue reorganization in the peritumoral and resected areas may be critical to optimize accuracy further.

Coupling solid and fluid stresses with brain tumour growth and white matter tract deformations in a neuroimaging-informed model.

Brain tumours are among the deadliest types of cancer, since they display a strong ability to invade the surrounding tissues and an extensive resistance to common therapeutic treatments. It is therefore important to reproduce the heterogeneity of brain microstructure through mathematical and computational models, that can provide powerful instruments to investigate cancer progression. However, only a few models include a proper mechanical and constitutive description of brain tissue, which instead may be relevant to predict the progression of the pathology and to analyse the reorganization of healthy tissues occurring during tumour growth and, possibly, after surgical resection. Motivated by the need to enrich the description of brain cancer growth through mechanics, in this paper we present a mathematical multiphase model that explicitly includes brain hyperelasticity. We find that our mechanical description allows to evaluate the impact of the growing tumour mass on the surrounding healthy tissue, quantifying the displacements, deformations, and stresses induced by its proliferation. At the same time, the knowledge of the mechanical variables may be used to model the stress-induced inhibition of growth, as well as to properly modify the preferential directions of white matter tracts as a consequence of deformations caused by the tumour. Finally, the simulations of our model are implemented in a personalized framework, which allows to incorporate the realistic brain geometry, the patient-specific diffusion and permeability tensors reconstructed from imaging data and to modify them as a consequence of the mechanical deformation due to cancer growth.

Quantification of pulmonary involvement in COVID-19 pneumonia by means of a cascade of two U-nets: training and assessment on multiple datasets using different annotation criteria.

PURPOSE: This study aims at exploiting artificial intelligence (AI) for the identification, segmentation and quantification of COVID-19 pulmonary lesions. The limited data availability and the annotation quality are relevant factors in training AI-methods. We investigated the effects of using multiple datasets, heterogeneously populated and annotated according to different criteria. METHODS: We developed an automated analysis pipeline, the LungQuant system, based on a cascade of two U-nets. The first one (U-net[Formula: see text]) is devoted to the identification of the lung parenchyma; the second one (U-net[Formula: see text]) acts on a bounding box enclosing the segmented lungs to identify the areas affected by COVID-19 lesions. Different public datasets were used to train the U-nets and to evaluate their segmentation performances, which have been quantified in terms of the Dice Similarity Coefficients. The accuracy in predicting the CT-Severity Score (CT-SS) of the LungQuant system has been also evaluated. RESULTS: Both the volumetric DSC (vDSC) and the accuracy showed a dependency on the annotation quality of the released data samples. On an independent dataset (COVID-19-CT-Seg), both the vDSC and the surface DSC (sDSC) were measured between the masks predicted by LungQuant system and the reference ones. The vDSC (sDSC) values of 0.95±0.01 and 0.66±0.13 (0.95±0.02 and 0.76±0.18, with 5 mm tolerance) were obtained for the segmentation of lungs and COVID-19 lesions, respectively. The system achieved an accuracy of 90% in CT-SS identification on this benchmark dataset. CONCLUSION: We analysed the impact of using data samples with different annotation criteria in training an AI-based quantification system for pulmonary involvement in COVID-19 pneumonia. In terms of vDSC measures, the U-net segmentation strongly depends on the quality of the lesion annotations. Nevertheless, the CT-SS can be accurately predicted on independent test sets, demonstrating the satisfactory generalization ability of the LungQuant.

Partial differential model of lactate neuro-energetics: analytic results and numerical simulations.

Interfaces play a key role on diseases development because they dictate the energy inflow of nutrients from the surrounding tissues. What is underestimated by existing mathematical models is the biological fact that cells are able to use different resources through nonlinear mechanisms. Among all nutrients, lactate appears to be a sensitive metabolic when talking about brain tumours or neurodegenerative diseases. Here we present a partial differential model to investigate the lactate exchanges between cells and the vascular network in the brain. By extending an existing kinetic model for lactate neuro-energetics, we first provide analytical proofs of the uniqueness and the derivation of precise bounds on the solutions of the problem including diffusion of lactate in a representative volume element comprising the interface between a capillary and cells. We further perform finite element simulations of the model in two test cases, discussing the relevant physical parameters governing the lactate dynamics.

In Silico Mathematical Modelling for Glioblastoma: A Critical Review and a Patient-Specific Case.

Glioblastoma extensively infiltrates the brain; despite surgery and aggressive therapies, the prognosis is poor. A multidisciplinary approach combining mathematical, clinical and radiological data has the potential to foster our understanding of glioblastoma evolution in every single patient, with the aim of tailoring therapeutic weapons. In particular, the ultimate goal of biomathematics for cancer is the identification of the most suitable theoretical models and simulation tools, both to describe the biological complexity of carcinogenesis and to predict tumor evolution. In this report, we describe the results of a critical review about different mathematical models in neuro-oncology with their clinical implications. A comprehensive literature search and review for English-language articles concerning mathematical modelling in glioblastoma has been conducted. The review explored the different proposed models, classifying them and indicating the significative advances of each one. Furthermore, we present a specific case of a glioblastoma patient in which our recently proposed innovative mechanical model has been applied. The results of the mathematical models have the potential to provide a relevant benefit for clinicians and, more importantly, they might drive progress towards improving tumor control and patient's prognosis. Further prospective comparative trials, however, are still necessary to prove the impact of mathematical neuro-oncology in clinical practice.