ABSTRACT
The accurate segmentation of individual muscles is essential for quantitative MRI analysis of thigh images. Deep learning methods have achieved state-of-the-art results in segmentation, but they require large numbers of labeled data to perform well. However, labeling individual thigh muscles slice by slice for numerous volumes is a laborious and time-consuming task, which limits the availability of annotated datasets. To address this challenge, self-supervised learning (SSL) emerges as a promising technique to enhance model performance by pretraining the model on unlabeled data. A recent approach, called positional contrastive learning, exploits the information given by the axial position of the slices to learn features transferable on the segmentation task. The aim of this work was to propose positional contrastive SSL for the segmentation of individual thigh muscles from MRI acquisitions in a population of elderly healthy subjects and to evaluate it on different levels of limited annotated data. An unlabeled dataset of 72 T1w MRI thigh acquisitions was available for SSL pretraining, while a labeled dataset of 52 volumes was employed for the final segmentation task, split into training and test sets. The effectiveness of SSL pretraining to fine-tune a U-Net architecture for thigh muscle segmentation was compared with that of a randomly initialized model (RND), considering an increasing number of annotated volumes (S = 1, 2, 5, 10, 20, 30, 40). Our results demonstrated that SSL yields substantial improvements in Dice similarity coefficient (DSC) when using a very limited number of labeled volumes (e.g., for S $$ S $$ = 1, DSC 0.631 versus 0.530 for SSL and RND, respectively). Moreover, enhancements are achievable even when utilizing the full number of labeled subjects, with DSC = 0.927 for SSL and 0.924 for RND. In conclusion, positional contrastive SSL was effective in obtaining more accurate thigh muscle segmentation, even with a very low number of labeled data, with a potential impact of speeding up the annotation process in clinics.
ABSTRACT
BACKGROUND AND OBJECTIVE: Mental workload (MWL) is a relevant construct involved in all cognitively demanding activities, and its assessment is an important goal in many research fields. This paper aims at evaluating the reproducibility and sensitivity of MWL assessment from EEG signals considering the effects of different electrode configurations and pre-processing pipelines (PPPs). METHODS: Thirteen young healthy adults were enrolled and were asked to perform 45 min of Simon's task to elicit a cognitive demand. EEG data were collected using a 32-channel system with different electrode configurations (fronto-parietal; Fz and Pz; Cz) and analyzed using different PPPs, from the simplest bandpass filtering to the combination of filtering, Artifact Subspace Reconstruction (ASR) and Independent Component Analysis (ICA). The reproducibility of MWL indexes estimation and the sensitivity of their changes were assessed using Intraclass Correlation Coefficient and statistical analysis. RESULTS: MWL assessed with different PPPs showed reliability ranging from good to very good in most of the electrode configurations (average consistency > 0.87 and average absolute agreement > 0.92). Larger fronto-parietal electrode configurations, albeit being more affected by the choice of PPPs, provide better sensitivity in the detection of MWL changes if compared to a single-electrode configuration (18 vs. 10 statistically significant differences detected, respectively). CONCLUSIONS: The most complex PPPs have been proven to ensure good reliability (>0.90) and sensitivity in all experimental conditions. In conclusion, we propose to use at least a two-electrode configuration (Fz and Pz) and complex PPPs including at least the ICA algorithm (even better including ASR) to mitigate artifacts and obtain reliable and sensitive MWL assessment during cognitive tasks.
Subject(s)
Electroencephalography , Workload , Adult , Humans , Reproducibility of Results , Signal Processing, Computer-Assisted , ElectrodesABSTRACT
Extraction of intravoxel incoherent motion (IVIM) parameters from noisy diffusion-weighted (DW) images using a biexponential fitting model is computationally challenging, and the reliability of the estimated perfusion-related quantities represents a limitation of this technique. Artificial intelligence can overcome the current limitations and be a suitable solution to advance use of this technique in both preclinical and clinical settings. The purpose of this work was to develop a deep neural network (DNN) approach, trained on numerical simulated phantoms with different signal to noise ratios (SNRs), to improve IVIM parameter estimation. The proposed approach is based on a supervised fully connected DNN having 3 hidden layers, 18 inputs and 3 targets with standardized values. 14 × 103 simulated DW images, based on a Shepp-Logan phantom, were randomly generated with varying SNRs (ranging from 10 to 100). 7 × 103 images (1000 for each SNR) were used for training. Performance accuracy was assessed in simulated images and the proposed approach was compared with the state-of-the-art Bayesian approach and other DNN algorithms. The DNN approach was also evaluated in vivo on a high-field MRI preclinical scanner. Our DNN approach showed an overall improvement in accuracy when compared with the Bayesian approach and other DNN methods in most of the simulated conditions. The in vivo results demonstrated the feasibility of the proposed approach in real settings and generated quantitative results comparable to those obtained using the Bayesian and unsupervised approaches, especially for D and f, and with lower variability in homogeneous regions. The DNN architecture proposed in this work outlines two innovative features as compared with other studies: (1) the use of standardized targets to improve the estimation of parameters, and (2) the implementation of a single DNN to enhance the IVIM fitting at different SNRs, providing a valuable alternative tool to compute IVIM parameters in conditions of high background noise.
Subject(s)
Artificial Intelligence , Diffusion Magnetic Resonance Imaging , Algorithms , Bayes Theorem , Diffusion Magnetic Resonance Imaging/methods , Motion , Neural Networks, Computer , Reproducibility of ResultsABSTRACT
Connectivity among different areas within the brain is a topic that has been notably studied in the last decade. In particular, EEG-derived measures of effective connectivity examine the directionalities and the exerted influences raised from the interactions among neural sources that are masked out on EEG signals. This is usually performed by fitting multivariate autoregressive models that rely on the stationarity that is assumed to be maintained over shorter bits of the signals. However, despite being a central condition, the selection process of a segment length that guarantees stationary conditions has not been systematically addressed within the effective connectivity framework, and thus, plenty of works consider different window sizes and provide a diversity of connectivity results. In this study, a segment-size-selection procedure based on fourth-order statistics is proposed to make an informed decision on the appropriate window size that guarantees stationarity both in temporal and spatial terms. Specifically, kurtosis is estimated as a function of the window size and used to measure stationarity. A search algorithm is implemented to find the segments with similar stationary properties while maximizing the number of channels that exhibit the same properties and grouping them accordingly. This approach is tested on EEG signals recorded from six healthy subjects during resting-state conditions, and the results obtained from the proposed method are compared to those obtained using the classical approach for mapping effective connectivity. The results show that the proposed method highlights the influence that arises in the Default Mode Network circuit by selecting a window of 4 s, which provides, overall, the most uniform stationary properties across channels.
Subject(s)
Brain Mapping , Brain , Algorithms , Electroencephalography , Humans , Magnetic Resonance Imaging/methods , Neural PathwaysABSTRACT
Electroencephalography (EEG) and electromyography (EMG) are widespread and well-known quantitative techniques used for gathering biological signals at cortical and muscular levels, respectively. Indeed, they provide relevant insights for increasing knowledge in different domains, such as physical and cognitive, and research fields, including neuromotor rehabilitation. So far, EEG and EMG techniques have been independently exploited to guide or assess the outcome of the rehabilitation, preferring one technique over the other according to the aim of the investigation. More recently, the combination of EEG and EMG started to be considered as a potential breakthrough approach to improve rehabilitation effectiveness. However, since it is a relatively recent research field, we observed that no comprehensive reviews available nor standard procedures and setups for simultaneous acquisitions and processing have been identified. Consequently, this paper presents a systematic review of EEG and EMG applications specifically aimed at evaluating and assessing neuromotor performance, focusing on cortico-muscular interactions in the rehabilitation field. A total of 213 articles were identified from scientific databases, and, following rigorous scrutiny, 55 were analyzed in detail in this review. Most of the applications are focused on the study of stroke patients, and the rehabilitation target is usually on the upper or lower limbs. Regarding the methodological approaches used to acquire and process data, our results show that a simultaneous EEG and EMG acquisition is quite common in the field, but it is mostly performed with EMG as a support technique for more specific EEG approaches. Non-specific processing methods such as EEG-EMG coherence are used to provide combined EEG/EMG signal analysis, but rarely both signals are analyzed using state-of-the-art techniques that are gold-standard in each of the two domains. Future directions may be oriented toward multi-domain approaches able to exploit the full potential of combined EEG and EMG, for example targeting a wider range of pathologies and implementing more structured clinical trials to confirm the results of the current pilot studies.
Subject(s)
Signal Processing, Computer-Assisted , Stroke , Electroencephalography , Electromyography , HumansABSTRACT
The Intra-Voxel Incoherent Motion (IVIM) model is largely adopted to estimate slow and fast diffusion coefficients of water molecules in biological tissues, which are used in cancer applications. The most reported fitting approach is a voxel-wise segmented non-linear least square, whereas Bayesian approaches with a direct fit, also considering spatial regularization, were proposed too. In this work a novel segmented Bayesian method was proposed, also in combination with a spatial regularization through a Conditional Autoregressive (CAR) prior specification. The two segmented Bayesian approaches, with and without CAR specification, were compared with two standard least-square and a direct Bayesian fitting methods. All approaches were tested on simulated images and real data of patients with head-and-neck and rectal cancer. Estimation accuracy and maps noisiness were quantified on simulated images, whereas the coefficient of variation and the goodness of fit were evaluated for real data. Both versions of the segmented Bayesian approach outperformed the standard methods on simulated images for pseudo-diffusion (D∗ ) and perfusion fraction (f), whilst the segmented least-square fitting remained the less biased for the diffusion coefficient (D). On real data, Bayesian approaches provided the less noisy maps, and the two Bayesian methods without CAR generally estimated lower values for f and D∗ coefficients with respect to the other approaches. The proposed segmented Bayesian approaches were superior, in terms of estimation accuracy and maps quality, to the direct Bayesian model and the least-square fittings. The CAR method improved the estimation accuracy, especially for D∗ .
Subject(s)
Algorithms , Diffusion Magnetic Resonance Imaging , Motion , Bayes Theorem , Computer Simulation , Humans , Image Processing, Computer-Assisted , Time FactorsABSTRACT
PURPOSE: The aim of the paper is to evaluate if advanced dMRI techniques, including diffusion kurtosis imaging (DKI) and neurite orientation dispersion and density imaging (NODDI), could provide novel insights into the subtle microarchitectural modifications occurring in the corticospinal tract (CST) of stroke patients in subacute and chronic phases. METHODS: Seventeen subjects (age 68 ± 11 years) in the subacute phase (14 ± 3 days post-stroke), 10 of whom rescanned in the chronic phase (231 ± 36 days post-stroke), were enrolled. Images were acquired using a 3-T MRI scanner with a two-shell EPI protocol (20 gradient directions, b = 700 s/mm2, 3 b = 0; 64 gradient directions, b = 2000 s/mm2, 9 b = 0). DTI-, DKI-, and NODDI-derived parameters were calculated in the posterior limb of the internal capsule (PLIC) and in the cerebral peduncle (CP). RESULTS: In the subacute phase, a reduction of FA, AD, and KA values was correlated with an increase of ODI, RD, and AK parameters, in both the ipsilesional PLIC and CP, suggesting that increased fiber dispersion can be the main structural factor. In the chronic phase, a reduction of FA and an increase of ODI persisted in the ipsilesional areas. This was associated with reduced Fic and increased MD, with a concomitant reduction of MK and increase of RD, suggesting that fiber reduction, possibly due to nerve degeneration, could play an important role. CONCLUSIONS: This study shows that advanced dMRI approaches can help elucidate the underpinning architectural modifications occurring in the CST after stroke. Further follow-up studies on bigger cohorts are needed to evaluate if DKI- and NODDI-derived parameters might be proposed as complementary biomarkers of brain microstructural alterations.
Subject(s)
Brain Ischemia/diagnostic imaging , Diffusion Magnetic Resonance Imaging , Image Processing, Computer-Assisted , Pyramidal Tracts/diagnostic imaging , Stroke/diagnostic imaging , Aged , Brain Ischemia/complications , Chronic Disease , Cohort Studies , Female , Humans , Male , Middle Aged , Stroke/etiology , Time FactorsABSTRACT
The main aim of this paper was to propose triggered intravoxel incoherent motion (IVIM) imaging sequences for the evaluation of perfusion changes in calf muscles before, during and after isometric intermittent exercise. Twelve healthy volunteers were involved in the study. The subjects were asked to perform intermittent isometric plantar flexions inside the MRI bore. MRI of the calf muscles was performed on a 3.0 T scanner and diffusion-weighted (DW) images were obtained using eight different b values (0 to 500 s/mm2 ). Acquisitions were performed at rest, during exercise and in the subsequent recovery phase. A motion-triggered echo-planar imaging DW sequence was implemented to avoid movement artifacts. Image quality was evaluated using the average edge strength (AES) as a quantitative metric to assess the motion artifact effect. IVIM parameters (diffusion D, perfusion fraction f and pseudo-diffusion D*) were estimated using a segmented fitting approach and evaluated in gastrocnemius and soleus muscles. No differences were observed in quality of IVIM images between resting state and triggered exercise, whereas the non-triggered images acquired during exercise had a significantly lower value of AES (reduction of more than 20%). The isometric intermittent plantar-flexion exercise induced an increase of all IVIM parameters (D by 10%; f by 90%; D* by 124%; fD* by 260%), in agreement with the increased muscle perfusion occurring during exercise. Finally, IVIM parameters reverted to the resting values within 3 min during the recovery phase. In conclusion, the IVIM approach, if properly adapted using motion-triggered sequences, seems to be a promising method to investigate muscle perfusion during isometric exercise.
Subject(s)
Exercise/physiology , Magnetic Resonance Imaging , Motion , Muscle, Skeletal/physiology , Perfusion , Adult , Artifacts , Diffusion Magnetic Resonance Imaging , Female , Humans , MaleABSTRACT
Bacterial-derived DNA fragments (BDNAs) have been shown to be present in a dialysis fluid, to pass through dialyzer membranes, and to induce interleukin 6 (IL-6) in mononuclear cells. DNA fragments are thought to be derived from microorganisms inhabiting hemodialysis water and fluid. The primary aim of the present study was to develop two degenerated TaqMan real-time quantitative-PCR (Q-PCR) for detection of a broad range of bacterial DNA that specifically detect 16S ribosomal DNA (rDNA) (862 and 241 bp) and evaluate the efficiency of the Bellco Selecta resin to capture the BDNAs in the dialysis fluid. For this purpose, we decided to compare measurements of unfragmented samples (9.8 × 105 Escherichia coli genome) with artificially fragmented DNA samples. We assessed two broad-range real-time PCR targeting bacterial 16S rRNA genes for detection of fragmented and unfragmented bacterial DNA in the dialytic fluid and demonstrated that Bellco Selecta resin is capable of retaining these types of bacterial DNA.
Subject(s)
DNA, Bacterial/genetics , DNA, Ribosomal/genetics , Escherichia coli/genetics , Genes, rRNA , RNA, Ribosomal, 16S/genetics , Real-Time Polymerase Chain Reaction/methodsABSTRACT
PURPOSE: To propose and assess a new method that automatically extracts a three-dimensional (3D) geometric model of the thoracic aorta (TA) from 3D cine phase contrast MRI (PCMRI) acquisitions. METHODS: The proposed method is composed of two steps: segmentation of the TA and creation of the 3D geometric model. The segmentation algorithm, based on Level Set, was set and applied to healthy subjects acquired in three different modalities (with and without SENSE reduction factors). Accuracy was evaluated using standard quality indices. The 3D model is characterized by the vessel surface mesh and its centerline; the comparison of models obtained from the three different datasets was also carried out in terms of radius of curvature (RC) and average tortuosity (AT). RESULTS: In all datasets, the segmentation quality indices confirmed very good agreement between manual and automatic contours (average symmetric distance < 1.44 mm, DICE Similarity Coefficient > 0.88). The 3D models extracted from the three datasets were found to be comparable, with differences of less than 10% for RC and 11% for AT. CONCLUSION: Our method was found effective on PCMRI data to provide a 3D geometric model of the TA, to support morphometric and hemodynamic characterization of the aorta.
Subject(s)
Aorta, Thoracic/anatomy & histology , Imaging, Three-Dimensional/methods , Magnetic Resonance Imaging, Cine/methods , Adult , Algorithms , Healthy Volunteers , Hemodynamics , HumansABSTRACT
PURPOSE: To introduce and validate an automatic segmentation method for the discrimination of skeletal muscle (SM), and adipose tissue (AT) components (subcutaneous adipose tissue [SAT] and intermuscular adipose tissue [IMAT]) from T1-weighted (T1 -W) magnetic resonance imaging (MRI) images of the thigh. MATERIALS AND METHODS: Eighteen subjects underwent an MRI examination on a 1.5T Philips Achieva scanner. Acquisition was performed using a T1 -W sequence (TR = 550 msec, TE = 15 msec), pixel size between 0.81-1.28 mm, slice thickness of 6 mm. Bone, AT, and SM were discriminated using a fuzzy c-mean algorithm and morphologic operators. The muscle fascia that separates SAT from IMAT was detected by integrating a morphological-based segmentation with an active contour Snake. The method was validated on five young normal weight, five older normal weight, and five older obese females, comparing automatic with manual segmentations. RESULTS: We reported good performance in the extraction of SM, AT, and bone in each subject typology (mean sensitivity above 96%, mean relative area difference of 1.8%, 2.7%, and 2.5%, respectively). A mean distance between contours pairs of 0.81 mm and a mean percentage of contour points with distance smaller than 2 pixels of 86.2% were obtained in the muscle fascia identification. Significant correlation was also found between manual and automatic IMAT and SAT cross-sectional areas in all subject typologies (p < 0.001). CONCLUSION: The proposed automatic segmentation approach provides adequate thigh tissue segmentation and may be helpful in studies of regional composition.
Subject(s)
Adipose Tissue/diagnostic imaging , Magnetic Resonance Imaging , Muscle, Skeletal/diagnostic imaging , Thigh/diagnostic imaging , Adiposity , Adult , Age Factors , Aged , Algorithms , Body Composition , Electronic Data Processing , Fascia/diagnostic imaging , Female , Fuzzy Logic , Humans , Middle Aged , Models, Statistical , Obesity/diagnostic imaging , Obesity/physiopathology , Pattern Recognition, Automated , Reproducibility of Results , Young AdultABSTRACT
Cardiovascular disease risk has been consistently linked with particulate matter (PM) exposure. Cell-derived microvesicles (MVs) are released into plasma and transfer microRNAs (miRNAs) between tissues. MVs can be produced by the respiratory system in response to proinflammatory triggers, enter the circulatory system and remotely modify gene expression in cardiovascular tissues. However, whether PM affects MV signaling has never been investigated. In this study, we evaluated expression of microRNAs contained within plasma MVs upon PM exposure both in vivo and in vitro. In the in vivo study, we isolated plasma MVs from healthy steel plant workers before and after workplace PM exposure. We measured the expression of 88 MV-associated miRNAs by real-time polymerase chain reaction. To assess a possible source of the MV miRNAs identified in vivo, we measured their miRNA expression in PM-treated A549 pulmonary cell lines in vitro. MiRNA profiling of plasma MVs showed 5.62- and 13.95-fold increased expression of miR-128 and miR-302c, respectively, after 3 days of workplace PM exposure (P < 0.001). According to Ingenuity Pathway Analysis, miR-128 is part of coronary artery disease pathways, and miR-302c is part of coronary artery disease, cardiac hypertrophy and heart failure pathways. In vitro experiments confirmed a dose-dependent expression of miR-128 in MVs released from A549 cells after 6 h of PM treatment (P = 0.030). MiR-302c was expressed neither from A549 cells nor in reference lung RNA. These results suggest novel PM-activated molecular mechanisms that may mediate the effects of air pollution and could lead to the identification of new diagnostic and therapeutic interventions.
Subject(s)
Air Pollutants, Occupational/toxicity , Cell-Derived Microparticles/drug effects , MicroRNAs/genetics , Occupational Exposure/adverse effects , Particulate Matter/toxicity , Pulmonary Alveoli/drug effects , Adult , Cell Line , Cell-Derived Microparticles/metabolism , Humans , Male , Metallurgy , MicroRNAs/blood , Middle Aged , Occupational Exposure/analysis , Pulmonary Alveoli/metabolism , Real-Time Polymerase Chain ReactionABSTRACT
PURPOSE: To quantitatively assess the predictive power of early variations of parotid gland volume and density on final changes at the end of therapy and, possibly, on acute xerostomia during IMRT for head-neck cancer. MATERIALS AND METHODS: Data of 92 parotids (46 patients) were available. Kinetics of the changes during treatment were described by the daily rate of density (rΔρ) and volume (rΔvol) variation based on weekly diagnostic kVCT images. Correlation between early and final changes was investigated as well as the correlation with prospective toxicity data (CTCAEv3.0) collected weekly during treatment for 24/46 patients. RESULTS: A higher rΔρ was observed during the first compared to last week of treatment (-0,50 vs -0,05HU, p-value = 0.0001). Based on early variations, a good estimation of the final changes may be obtained (Δρ: AUC = 0.82, p = 0.0001; Δvol: AUC = 0.77, p = 0.0001). Both early rΔρ and rΔvol predict a higher "mean" acute xerostomia score (≥ median value, 1.57; p-value = 0.01). Median early density rate changes for patients with mean xerostomia score ≥ / < 1.57 were -0.98 vs -0.22 HU/day respectively (p = 0.05). CONCLUSIONS: Early density and volume variations accurately predict final changes of parotid glands. A higher longitudinally assessed score of acute xerostomia is well predicted by higher rΔρ and rΔvol in the first two weeks of treatment: best cut-off values were -0.50 HU/day and -380 mm(3)/day for rΔρ and rΔvol respectively. Further studies are necessary to definitively assess the potential of early density/volume changes in identifying more sensitive patients at higher risk of experiencing xerostomia.
Subject(s)
Head and Neck Neoplasms/radiotherapy , Parotid Gland/diagnostic imaging , Radiation Injuries/diagnostic imaging , Radiation Injuries/etiology , Radiotherapy, Conformal/adverse effects , Xerostomia/diagnostic imaging , Xerostomia/etiology , Absorptiometry, Photon , Acute Disease , Early Diagnosis , Female , Head and Neck Neoplasms/diagnostic imaging , Humans , Italy , Male , Organ Size/radiation effects , Parotid Gland/radiation effects , Reproducibility of Results , Sensitivity and Specificity , Tomography, X-Ray Computed , United StatesABSTRACT
BACKGROUND: Despite epidemiological findings showing increased air pollution related cardiovascular diseases (CVD), the knowledge of the involved molecular mechanisms remains moderate or weak. Particulate matter (PM) produces a local strong inflammatory reaction in the pulmonary environment but there is no final evidence that PM physically enters and deposits in blood vessels. Extracellular vesicles (EVs) and their miRNA cargo might be the ideal candidate to mediate the effects of PM, since they could be potentially produced by the respiratory system, reach the systemic circulation and lead to the development of cardiovascular effects.The SPHERE ("Susceptibility to Particle Health Effects, miRNAs and Exosomes") project was granted by ERC-2011-StG 282413, to examine possible molecular mechanisms underlying the effects of PM exposure in relation to health outcomes. METHODS/DESIGN: The study population will include 2000 overweight (25 < BMI < 30 kg/cm2) or obese (BMI ≥ 30 kg/cm2) subjects presenting at the Center for Obesity and Work (Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico, Milan, Italy).Each subject donates blood, urine and hair samples. Extensive epidemiological and clinical data are collected. Exposure to PM is assigned to each subject using both daily PM10 concentration series from air quality monitors and pollutant levels estimated by the FARM (Flexible air Quality Regional Model) modelling system and elaborated by the Regional Environmental Protection Agency.The recruitment period started in September 2010 and will continue until 2015. At December 31, 2013 we recruited 1250 subjects, of whom 87% lived in the province of Milan.Primary study outcomes include cardiometabolic and respiratory health effects. The main molecular mechanism we are investigating focuses on EV-associated microRNAs. DISCUSSION: SPHERE is the first large study aimed to explore EVs as a novel potential mechanism of how air pollution exposure acts in a highly susceptible population. The rigorous study design, the availability of banked biological samples and the potential to integrate epidemiological, clinical and molecular data will also furnish a powerful base for investigating different complementary molecular mechanisms. Our findings, if confirmed, could lead to the identification of potentially reversible alterations that might be considered as possible targets for new diagnostic and therapeutic interventions.
Subject(s)
Air Pollution/adverse effects , Cardiovascular Diseases/etiology , Disease Susceptibility , Obesity , Respiratory Tract Diseases/etiology , Air Pollutants/analysis , Cardiovascular Diseases/blood , Cardiovascular Diseases/urine , Environmental Monitoring , Exosomes/chemistry , Female , Humans , Italy , Male , MicroRNAs/analysis , Middle Aged , Models, Theoretical , Respiratory Tract Diseases/blood , Respiratory Tract Diseases/urineABSTRACT
BACKGROUND: Automatic segmentation techniques based on Convolutional Neural Networks (CNNs) are widely adopted to automatically identify any structure of interest from a medical image, as they are not time consuming and not subject to high intra- and inter-operator variability. However, the adoption of these approaches in clinical practice is slowed down by some factors, such as the difficulty in providing an accurate quantification of their uncertainty. PURPOSE: This work aims to evaluate the uncertainty quantification provided by two Bayesian and two non-Bayesian approaches for a multi-class segmentation problem, and to compare the risk propensity among these approaches, considering CT images of patients affected by renal cancer (RC). METHODS: Four uncertainty quantification approaches were implemented in this work, based on a benchmark CNN currently employed in medical image segmentation: two Bayesian CNNs with different regularizations (Dropout and DropConnect), named BDR and BDC, an ensemble method (Ens) and a test-time augmentation (TTA) method. They were compared in terms of segmentation accuracy, using the Dice score, uncertainty quantification, using the ratio of correct-certain pixels (RCC) and incorrect-uncertain pixels (RIU), and with respect to inter-observer variability in manual segmentation. They were trained with the Kidney and Kidney Tumor Segmentation Challenge launched in 2021 (Kits21), for which multi-class segmentations of kidney, RC, and cyst on 300 CT volumes are available. Moreover, they were tested considering this and other two public renal CT datasets. RESULTS: Accuracy results achieved large differences across the structures of interest for all approaches, with an average Dice score of 0.92, 0.58, and 0.21 for kidney, tumor, and cyst, respectively. In terms of uncertainties, TTA provided the highest uncertainty, followed by Ens and BDC, whereas BDR provided the lowest, and minimized the number of incorrect certain pixels worse than the other approaches. Again, large differences were seen across the three structures in terms of RCC and RIU. These metrics were associated with different risk propensity, as BDR was the most risk-taking approach, able to provide higher accuracy in its prediction, but failing to assign uncertainty on incorrect segmentation in every case. The other three approaches were more conservative, providing large uncertainty regions, with the drawback of giving alert also on correct areas. Finally, the analysis of the inter-observer segmentation variability showed a significant variation among the four approaches on the external dataset, with BDR reporting the lowest agreement (Dice = 0.82), and TTA obtaining the highest score (Dice = 0.94). CONCLUSIONS: Our outcomes highlight the importance of quantifying the segmentation uncertainty and that decision-makers can choose the approach most in line with the risk propensity degree required by the application and their policy.
ABSTRACT
Muscular dystrophies present diagnostic challenges, requiring accurate classification for effective diagnosis and treatment. This study investigates the efficacy of deep learning methodologies in classifying these disorders using skeletal muscle MRI scans. Specifically, we assess the performance of the Swin Transformer (SwinT) architecture against traditional convolutional neural networks (CNNs) in distinguishing between healthy individuals, Becker muscular dystrophy (BMD), and limb-girdle muscular Dystrophy type 2 (LGMD2) patients. Moreover, 3T MRI scans from a retrospective dataset of 75 scans (from 54 subjects) were utilized, with multiparametric protocols capturing various MRI contrasts, including T1-weighted and Dixon sequences. The dataset included 17 scans from healthy volunteers, 27 from BMD patients, and 31 from LGMD2 patients. SwinT and CNNs were trained and validated using a subset of the dataset, with the performance evaluated based on accuracy and F-score. Results indicate the superior accuracy of SwinT (0.96), particularly when employing fat fraction (FF) images as input; it served as a valuable parameter for enhancing classification accuracy. Despite limitations, including a modest cohort size, this study provides valuable insights into the application of AI-driven approaches for precise neuromuscular disorder classification, with potential implications for improving patient care.
ABSTRACT
BACKGROUND: Radiomics can be applied on parametric maps obtained from IntraVoxel Incoherent Motion (IVIM) MRI to characterize heterogeneity in diffusion and perfusion tissue properties. The purpose of this work is to assess the accuracy and reproducibility of radiomic features computed from IVIM maps using different fitting methods. METHODS: 200 digitally simulated IVIM-MRI images with various SNR containing different combinations of texture patterns were generated from ground truth maps of true diffusion D, pseudo-diffusion D* and perfusion fraction f. Four different methods (segmented least-square LSQ, Bayesian, supervised and unsupervised deep learning DL) were adopted to quantify IVIM maps from simulations and from two real images of liver tumor. Radiomic features were computed from ground truth and estimated maps. Accuracy and reproducibility among quantification methods were assessed. RESULTS: Almost 50% of radiomic features computed from D maps using DL approaches, 36% using Bayes and 27% using LSQ presented errors lower than 50%. Radiomic features from f and D* maps were accurate only if computed using DL methods from histogram. High reproducibility (ICC>0.8) was found only for D maps among DL and Bayes methods, whereas features from f and D* maps were less reproducible, with LSQ approach in lower agreement with the others. CONCLUSIONS: Texture patterns were preserved and correctly estimated only on D maps, except for LSQ approach. We suggest limiting radiomic analysis only to histogram and some texture features from D maps, to histogram features from f maps, and to avoid it on D* maps.
Subject(s)
Diffusion Magnetic Resonance Imaging , Image Processing, Computer-Assisted , Bayes Theorem , Diffusion Magnetic Resonance Imaging/methods , Reproducibility of Results , Image Processing, Computer-Assisted/methods , Algorithms , Magnetic Resonance Imaging , MotionABSTRACT
In clinical scenarios, the use of biomedical sensors, devices and multi-parameter assessments is fundamental to provide a comprehensive portrait of patients' state, in order to adapt and personalize rehabilitation interventions and support clinical decision-making. However, there is a huge gap between the potential of the multidomain techniques available and the limited practical use that is made in the clinical scenario. This paper reviews the current state-of-the-art and provides insights into future directions of multi-domain instrumental approaches in the clinical assessment of patients involved in neuromotor rehabilitation. We also summarize the main achievements and challenges of using multi-domain approaches in the assessment of rehabilitation for various neurological disorders affecting motor functions. Our results showed that multi-domain approaches combine information and measurements from different tools and biological signals, such as kinematics, electromyography (EMG), electroencephalography (EEG), near-infrared spectroscopy (NIRS), and clinical scales, to provide a comprehensive and objective evaluation of patients' state and recovery. This multi-domain approach permits the progress of research in clinical and rehabilitative practice and the understanding of the pathophysiological changes occurring during and after rehabilitation. We discuss the potential benefits and limitations of multi-domain approaches for clinical decision-making, personalized therapy, and prognosis. We conclude by highlighting the need for more standardized methods, validation studies, and the integration of multi-domain approaches in clinical practice and research.
ABSTRACT
An ontology-supported e-knowledge base aimed to the evaluation of obesity and related co-morbidities is presented. The main goal of such a clinical profiling tool is to help determine the health status of a subject, supporting the knowledge transfer between medical researchers and general practitioners.
Subject(s)
Decision Support Systems, Clinical , Decision Support Techniques , Diagnosis, Computer-Assisted/methods , Obesity/diagnosis , Obesity/epidemiology , HumansABSTRACT
The use of MRI radiomic models for the diagnosis, prognosis and treatment response prediction of tumors has been increasingly reported in literature. However, its widespread adoption in clinics is hampered by issues related to features stability. In the MRI radiomic workflow, the main factors that affect radiomic features computation can be found in the image acquisition and reconstruction phase, in the image pre-processing steps, and in the segmentation of the region of interest on which radiomic indices are extracted. Deep Neural Networks (DNNs), having shown their potentiality in the medical image processing and analysis field, can be seen as an attractive strategy to partially overcome the issues related to radiomic stability and mitigate their impact. In fact, DNN approaches can be prospectively integrated in the MRI radiomic workflow to improve image quality, obtain accurate and reproducible segmentations and generate standardized images. In this review, DNN methods that can be included in the image processing steps of the radiomic workflow are described and discussed, in the light of a detailed analysis of the literature in the context of MRI radiomic reliability.