A deep learning system for quantitative assessment of microvascular abnormalities in nailfold capillary images.

OBJECTIVES: Nailfold capillaroscopy is key to timely diagnosis of SSc, but is often not used in rheumatology clinics because the images are difficult to interpret. We aimed to develop and validate a fully automated image analysis system to fill this gap. METHODS: We mimicked the image interpretation strategies of SSc experts, using deep learning networks to detect each capillary in the distal row of vessels and make morphological measurements. We combined measurements from multiple fingers to give a subject-level probability of SSc.We trained the system using high-resolution images from 111 subjects (group A) and tested on images from subjects not in the training set: 132 imaged at high-resolution (group B); 66 imaged with a low-cost digital microscope (group C). Roughly half of each group had confirmed SSc, and half were healthy controls or had primary RP ('normal'). We also estimated the performance of SSc experts. RESULTS: We compared automated SSc probabilities with the known clinical status of patients (SSc versus 'normal'), generating receiver operating characteristic curves (ROCs). For group B, the area under the ROC (AUC) was 97% (94-99%) [median (90% CI)], with equal sensitivity/specificity 91% (86-95%). For group C, the AUC was 95% (88-99%), with equal sensitivity/specificity 89% (82-95%). SSc expert consensus achieved sensitivity 82% and specificity 73%. CONCLUSION: Fully automated analysis using deep learning can achieve diagnostic performance at least as good as SSc experts, and is sufficiently robust to work with low-cost digital microscope images.

Nailfold capillaroscopy: a survey of current UK practice and 'next steps' to increase uptake among rheumatologists.

OBJECTIVES: To identify barriers to the use of nailfold capillaroscopy as a diagnostic tool for patients presenting with Raynaud's phenomenon in UK rheumatology centres and to obtain rheumatologists' views on a proposed internet-based standardized system for clinical reporting of nailfold capillaroscopy images. METHODS: An online survey was developed using expert opinion from clinicians, scientists and health service researchers. The survey was piloted and sent to UK-based rheumatologists using established electronic mailing lists between October 2020 and March 2021. Survey data were analysed using descriptive statistics. RESULTS: A total of 104 rheumatologists representing rheumatology centres across the UK responded to the survey. Wide variations in terms of workloads and practices were described. Thirty-four (33%) respondents reported using nailfold capillaroscopy only at their own centre, 33 (32%) referred to other centres, 9 (9%) did both and 28 (27%) did not use capillaroscopy at all. Of the 43 respondents using capillaroscopy on site, 25 (58%) used either a dermatoscope or universal serial bus microscope and 9 (21%) used videocapillaroscopy. Among the 61 respondents not undertaking capillaroscopy on site, barriers included lack of equipment (85%), lack of experience in acquiring images (69%) and lack of expertise in interpreting images (67%). Sixty-six respondents (63%) expressed interest in an internet-based, standardized automated system for reporting images. CONCLUSION: Most UK rheumatologists currently do not perform nailfold capillaroscopy on site. An internet-based nailfold capillaroscopy system for use with low-cost microscopes as well as with videocapillaroscopy could help increase uptake of capillaroscopy and thereby facilitate early diagnosis of SSc across the UK.

Quantitative nailfold capillaroscopy-update and possible next steps.

We review the exciting potential (and challenges) of quantitative nailfold capillaroscopy, focusing on its role in systemic sclerosis. Quantifying abnormality, including automated analysis of nailfold images, overcomes the subjectivity of qualitative/descriptive image interpretation. First we consider the rationale for quantitative analysis, including the potential for precise discrimination between normal and abnormal capillaries and for reliable measurement of disease progression and treatment response. We discuss nailfold image acquisition and interpretation, and describe how early work on semi-quantitative and quantitative analysis paved the way for semi-automated and automated analysis. Measurement of red blood cell velocity is described briefly. Finally we give a personal view on 'next steps'. From a clinical perspective, increased uptake of nailfold capillaroscopy by general rheumatologists could be achieved via low-cost hand-held devices with cloud-based automated analysis. From a research perspective, automated analysis could facilitate large-scale prospective studies using capillaroscopic parameters as possible biomarkers of systemic sclerosis-spectrum disorders.

Detection of early subclinical lung disease in children with cystic fibrosis by lung ventilation imaging with hyperpolarised gas MRI.

Hyperpolarised 3He ventilation-MRI, anatomical lung MRI, lung clearance index (LCI), low-dose CT and spirometry were performed on 19 children (6-16 years) with clinically stable mild cystic fibrosis (CF) (FEV1>-1.96), and 10 controls. All controls had normal spirometry, MRI and LCI. Ventilation-MRI was the most sensitive method of detecting abnormalities, present in 89% of patients with CF, compared with CT abnormalities in 68%, LCI 47% and conventional MRI 22%. Ventilation defects were present in the absence of CT abnormalities and in patients with normal physiology, including LCI. Ventilation-MRI is thus feasible in young children, highly sensitive and provides additional information about lung structure-function relationships.

Indexed distribution analysis for improved significance testing of spatially heterogeneous parameter maps: application to dynamic contrast-enhanced MRI biomarkers.

PURPOSE: To develop significance testing methodology applicable to spatially heterogeneous parametric maps of biophysical and physiological measurements arising from imaging studies. THEORY: Heterogeneity can confound statistical analyses. Indexed distribution analysis (IDA) transforms a reference distribution, establishing correspondences across parameter maps to which significance tests are applied. METHODS: Well-controlled simulated and clinical K(trans) data from a dynamic contrast-enhanced magnetic resonance imaging study of bevacizumab were analyzed using conventional significance tests of parameter averages, histogram analysis, and IDA. Repeated pretreatment scans provided negative control; a post treatment scan provided positive control. RESULTS: Histogram analysis was insensitive to simulated and known effects. Simulation: conventional analysis identified treatment effect (P ≈ 5 × 10(-4)) and direction, but underestimated magnitude (relative error 67-81%); IDA identified treatment effect (P = 0.001), magnitude, direction, and spatial extent (100% accuracy). Bevacizumab: conventional analysis was sensitive to treatment effect (P = 0.01; 95% confidence interval on K(trans) decrease: 23-37%); IDA was sensitive to treatment effect (P < 0.05; K(trans) decrease approximately 25%), inferred its spatial extent to be 94-96%, and inferred that K(trans) decrease is independent of baseline value, an inference that conventional and histogram analyses cannot make. CONCLUSIONS: In the presence of heterogeneity, IDA can accurately infer the magnitude, direction, and spatial extent of between samples of parametric maps, which can be visualized spatially with respect to the original parameter maps.

Magnetic resonance transverse relaxation time T2 of knee cartilage in osteoarthritis at 3-T: a cross-sectional multicentre, multivendor reproducibility study.

OBJECTIVE: The transverse relaxation time (T2) in MR imaging has been identified as a potential biomarker of hyaline cartilage pathology. This study investigates whether MR assessments of T2 are comparable between 3-T scanners from three different vendors. DESIGN: Twelve subjects with symptoms of knee osteoarthritis and one or more risk factors had their knee scanned on each of the three vendors' scanners located in three sites in the U.K. MR data acquisition was based on the United States National Institutes of Health Osteoarthritis Initiative protocol. Measures of cartilage T2 and R2 (inverse of T2) were computed for precision error assessment. Intrascanner reproducibility was also assessed with a phantom (all three scanners) and a cohort of 5 subjects (one scanner only). RESULTS: Whole-organ magnetic resonance (WORM) semiquantitative cartilage scores ranged from minimal to advanced degradation. Intrascanner R2 root-mean-square coefficients of variation (RMSCOV) were low, within the range 2.6 to 6.3% for femoral and tibial regions. For one scanner pair, mean T2 differences ranged from -1.2 to 2.8 ms, with no significant difference observed for the medial tibia and patella regions (p < 0.05). T2 values from the third scanner were systematically lower, producing interscanner mean T2 differences within the range 5.4 to 10.0 ms. CONCLUSION: Significant interscanner cartilage T2 differences were found and should be accounted for before data from scanners of different vendors are compared.

Comparison of model-based arterial input functions for dynamic contrast-enhanced MRI in tumor bearing rats.

When using tracer kinetic modeling to analyze dynamic contrast-enhanced MRI (DCE-MRI) it is necessary to identify an appropriate arterial input function (AIF). The measured AIF is often poorly sampled in both clinical and preclinical MR systems due to the initial rapid increase in contrast agent concentration and the subsequent large-scale signal change that occurs in the arteries. However, little work has been carried out to quantify the sensitivity of tracer kinetic modeling parameters to the form of AIF. Using a preclinical experimental data set, we sought to measure the effect of varying model forms of AIF on the extended Kety compartmental model parameters (K(trans), v(e), and v(p)) through comparison with the results of experimentally acquired high temporal resolution AIFs. The AIF models examined have the potential to be parameterized on lower temporal resolution data to predict the form of the true, higher temporal resolution AIF. The models were also evaluated through application to the population average AIF. It was concluded that, in the instance of low temporal resolution or noisy data, it may be preferable to use a bi-exponential model applied to the raw data AIF, or when individual measurements are not available a bi-exponential model of the average AIF.

Oxygen-induced changes in longitudinal relaxation times in skeletal muscle.

The objective was to measure the effect of 100% oxygen inhalation on T1 relaxation times in skeletal muscle. Healthy volunteers were scanned using three different MRI protocols while breathing medical air and 100% oxygen. Measurements of T1 were made from regions of interest (ROIs) within various skeletal muscle groups. Dynamic data of subjects breathing a sequence of air-oxygen-air allowed the calculation of characteristic wash-in and -out times for dissolved oxygen in muscle. Contrary to previous findings, a statistically significant decrease in T1 in skeletal muscle was observed due to oxygen inhalation. We report approximate baseline characteristic values for the response of skeletal muscle to oxygen inhalation. This measurement may provide new biomarkers for evaluation of oxygen delivery and consumption in normal and diseased skeletal muscle.

Investigation of genomic instability in paediatric Barrett's oesophagus using laser microdissection on paraffin-embedded endoscopic biopsies.

OBJECTIVES: To investigate the occurrence of microsatellite instability (MSI) in paediatric Barrett's oesophagus (BE) with the aim of identifying a potential marker for patients at risk for developing dysplasia or adenocarcinoma at a later stage. PATIENTS AND METHODS: Endoscopic oesophageal biopsies from 6 pediatric patients harbouring BE and 6 age-matched controls were retrospectively investigated. After all of the samples were made anonymous, a 5-microm section was cut and stained with toluidine blue. A precise cell was selected and captured using a laser microdissection microscope. Microsatellite analysis was performed on the DNA extracted from the captured cells. Genomic DNA was amplified by polymerase chain reaction using primers for 5 mononucleotide repeats and analysed by GeneMapper software on an ABI 3730. DNA extracted from a formalin-fixed colonic adenocarcinoma known to have MSI was used as a positive control. RESULTS: The median age of the patients with BE was 9 years. The relevant complaint was long-standing vomiting in 4 cases and history of dysphagia in 3 cases (1 case had both symptoms). All of the cases had a history of reflux oesophagitis with histological confirmation of oesophagitis. Reflux was associated with a hiatus hernia, obesity, and cerebral palsy in each of 3 cases, and with asthma in 2 patients. Histologically, all of the cases showed the presence of specialized intestinal metaplasia containing goblet cells replacing the squamous oesophageal epithelium. None of the cases tested showed any evidence of MSI. CONCLUSIONS: A single molecular marker that would allow recognition of those patients at risk for Barrett's adenocarcinoma has not yet been identified. The absence of MSI in our cases could be due to the need for a longer period of BE before genomic instability develops, or MSI may only arise in a proportion of patients. This does not exclude other genetic alterations, however rare they may be.

Supine posture changes lung volumes and increases ventilation heterogeneity in cystic fibrosis.

INTRODUCTION: Lung Clearance Index (LCI) is recognised as an early marker of cystic fibrosis (CF) lung disease. The effect of posture on LCI however is important when considering longitudinal measurements from infancy and when comparing LCI to imaging studies. METHODS: 35 children with CF and 28 healthy controls (HC) were assessed. Multiple breath washout (MBW) was performed both sitting and supine in triplicate and analysed for LCI, Scond, Sacin, and lung volumes. These values were also corrected for the Fowler dead-space to create 'alveolar' indices. RESULTS: From sitting to supine there was a significant increase in LCI and a significant decrease in FRC for both CF and HC (p<0.01). LCI, when adjusted to estimate 'alveolar' LCI (LCIalv), increased the magnitude of change with posture for both LCIalv and FRCalv in both groups, with a greater effect of change in lung volume in HC compared with children with CF. The % change in LCIalv for all subjects correlated significantly with lung volume % changes, most notably tidal volume/functional residual capacity (Vtalv/FRCalv (r = 0.54,p<0.001)). CONCLUSION: There is a significant increase in LCI from sitting to supine, which we believe to be in part due to changes in lung volume and also increasing ventilation heterogeneity related to posture. This may have implications in longitudinal measurements from infancy to older childhood and for studies comparing supine imaging methods to LCI.

A minimum description length approach to statistical shape modeling.

We describe a method for automatically building statistical shape models from a training set of example boundaries/surfaces. These models show considerable promise as a basis for segmenting and interpreting images. One of the drawbacks of the approach is, however, the need to establish a set of dense correspondences between all members of a set of training shapes. Often this is achieved by locating a set of "landmarks" manually on each training image, which is time consuming and subjective in two dimensions and almost impossible in three dimensions. We describe how shape models can be built automatically by posing the correspondence problem as one of finding the parameterization for each shape in the training set. We select the set of parameterizations that build the "best" model. We define "best" as that which minimizes the description length of the training set, arguing that this leads to models with good compactness, specificity and generalization ability. We show how a set of shape parameterizations can be represented and manipulated in order to build a minimum description length model. Results are given for several different training sets of two-dimensional boundaries, showing that the proposed method constructs better models than other approaches including manual landmarking-the current gold standard. We also show that the method can be extended straightforwardly to three dimensions.

Building 3-D statistical shape models by direct optimization.

Statistical shape models are powerful tools for image interpretation and shape analysis. A simple, yet effective, way of building such models is to capture the statistics of sampled point coordinates over a training set of example shapes. However, a major drawback of this approach is the need to establish a correspondence across the training set. In 2-D, a correspondence is often defined using a set of manually placed 'landmarks' and linear interpolation to sample the shape in between. Such annotation is, however, time-consuming and subjective, particularly when extended to 3-D. In this paper, we show that it is possible to establish a dense correspondence across the whole training set automatically by treating correspondence as an optimization problem. The objective function we use for the optimization is based on the minimum description length principle, which we argue is a criterion that leads to models with good compactness, specificity, and generalization ability. We manipulate correspondence by reparameterizing each training shape. We describe an explicit representation of reparameterization for surfaces in 3-D that makes it impossible to generate an illegal (i.e., not one-to-one) correspondence. We also describe several large-scale optimization strategies for model building, and perform a detailed analysis of each approach. Finally, we derive quantitative measures of model quality, allowing meaningful comparison between models built using different methods. Results are given for several different training sets of 3-D shapes, which show that the minimum description length models perform significantly better than other approaches.

Comparison of 3T MR scanners in regional cartilage-thickness analysis in osteoarthritis: a cross-sectional multicenter, multivendor study.

INTRODUCTION: Cartilage thickness from MR images has been identified as a possible biomarker in knee osteoarthritis (OA) research. The ability to acquire MR data at multiple centers by using different vendors' scanners would facilitate patient recruitment and shorten the duration of OA trials. Several vendors manufacture 3T MR scanners, including Siemens, Philips Medical Systems, and GE Healthcare. This study investigates whether quantitative MR assessments of cartilage morphology are comparable between scanners of three different vendors. METHODS: Twelve subjects with symptoms of knee OA and one or more risk factors had their symptomatic knee scanned on each of the three vendor's scanners located in three sites in the UK: Manchester (Philips), York (GE), and Liverpool (Siemens). The NIH OAI study protocol was used for the Siemens scanner, and equivalent protocols were developed for the Philips and GE scanners with vendors' advice. Cartilage was segmented manually from sagittal 3D images. By using recently described techniques for Anatomically Corresponded Regional Analysis of Cartilage (ACRAC), a statistical model was used anatomically to align all the images and to produce detailed maps of mean differences in cartilage-thickness measures between scanners. Measures of cartilage mean thickness were computed in anatomically equivalent regions for each subject and scanner image. RESULTS: The ranges of mean cartilage-thickness measures for this cohort were similar for all regions and across all scanners. Philips intrascanner root-mean-square coefficients of variation were low in the range from 2.6% to 4.6%. No significant differences were found for thickness measures of the weight-bearing femorotibial regions from the Philips and Siemens images except for the central medial femur compartment (P = 0.04). Compared with the other two scanners, the GE scanner provided consistently lower mean thickness measures in the central femoral regions (mean difference, -0.16 mm) and higher measures in the tibial compartments (mean difference, +0.19 mm). CONCLUSIONS: The OAI knee-imaging protocol, developed on the Siemens platform, can be applied to research and trials by using other vendors' 3T scanners giving comparable morphologic results. Accurate sequence optimization, differences in image postprocessing, and extremity coil type are critical factors for interscanner precision of quantitative analysis of cartilage morphology. It is still recommended that longitudinal observations on individuals should be performed on the same scanner and that assessment of intra- and interscanner precision errors is undertaken before commencement of the main study.

Change in hematologic indices over time in pediatric inflammatory bowel disease treated with azathioprine.

Azathioprine leads to changes in mean corpuscular volume (MCV) and white blood cell (WBC) indices reflecting efficacy or toxicity. Understanding the interactions between bone marrow stem cells and azathioprine could highlight abnormal response patterns as forerunners for hematologic malignancies. This study gives a statistical description of factors influencing the relationship between MCV and WBC in children with inflammatory bowel disease treated with azathioprine. We found that leukopenia preceded macrocytosis. Macrocytosis is therefore not a good predictor of leukopenia. Further studies will be necessary to determine the subgroup of patients at increased risk of malignancies based on bone marrow response.

Anatomically corresponded regional analysis of cartilage in asymptomatic and osteoarthritic knees by statistical shape modelling of the bone.

Magnetic resonance imaging (MRI) is emerging as the method of choice for measuring cartilage loss in osteoarthritis (OA), but current methods of analysis are imperfect for therapeutic clinical trials. In this paper, we present and evaluate, in two multicenter multivendor studies, a new method for anatomically corresponded regional analysis of cartilage (ACRAC) that allows analysis of knee cartilage morphology in anatomically corresponding focal regions defined on the bone surface. In our first study, 3-D knee MR Images were obtained from 19 asymptomatic female volunteers, followed by segmentations of the bone and cartilage. Minimum description length (MDL) statistical shape models (SSMs) were constructed from the segmented bone surfaces, providing mean bone shapes and a dense set of anatomically corresponding positions on each individual bone, the accuracy of which were measured using repeat images from a subset of the volunteers. Cartilage thicknesses were measured at these locations along 3-D normals to the bone surfaces, yielding corresponded cartilage thickness maps. Functional subregions of the joint were defined on the mean bone shapes, and propagated, using the correspondences, to each individual. ACRAC improved reproducibility, particularly in the central, load bearing subregions of the joint, compared with measures of volume obtained directly from the segmented cartilage surfaces. In our second study, MR Images were obtained from 31 female patient-volunteers with knee OA at baseline and six months. We obtained manual segmentations of the cartilage, and automatic segmentations of the bone using active appearance models (AAMs) built from the bone SSMs of the first study. ACRAC enabled the detection of significant thickness loss in the central, load-bearing regions of the whole femur (-5.57% p = 0.01, annualized) and the medial condyle (-13.08% , p = 0.024 Bonferroni corrected, annualized). We conclude that statistical shape modelling of bone surfaces defines correspondences invariant to individual joint size or shape, providing focal measures of cartilage with improved reproducibility compared to whole compartment measures. It permits the identification of anatomically equivalent regions, and provides the ability to identify the main load-bearing regions of the joint, based on the imputed premorbid state. The method permitted detection of tiny morphological change in cartilage thickness over six months in a small study, and may be useful for OA disease analysis and treatment monitoring.

A unified information-theoretic approach to groupwise non-rigid registration and model building.

The non-rigid registration of a group of images shares a common feature with building a model of a group of images: a dense, consistent correspondence across the group. Image registration aims to find the correspondence, while modelling requires it. This paper presents the theoretical framework required to unify these two areas, providing a groupwise registration algorithm, where the inherently groupwise model of the image data becomes an integral part of the registration process. The performance of this algorithm is evaluated by extending the concepts of generalisability and specificity from shape models to image models. This provides an independent metric for comparing registration algorithms of groups of images. Experimental results on MR data of brains for various pairwise and groupwise registration algorithms is presented, and demonstrates the feasibility of the combined registration/modelling framework, as well as providing quantitative evidence for the superiority of groupwise approaches to registration.

Improved quantitative dynamic regional oxygen-enhanced pulmonary imaging using image registration.

Oxygen-enhanced MR imaging has been demonstrated in a number of recent studies as a potential method to visualize regional ventilation in the lung. A quantitative pixel-by-pixel analysis is hampered by motion and volume change due to breathing. In this study, image registration via active shape modeling is shown to produce significant improvements in the regional analysis of both static and dynamic oxygen-enhanced pulmonary MRI for five normal volunteers. The method enables the calculation of regional change in relaxation rate between breathing air and 100% oxygen, which is proportional to the change in regional oxygen concentration, and regional oxygen wash-in and wash-out time constants. Registration-corrected mapping of these parameters is likely to provide improved information in the regional assessment of a range of lung diseases.

Quantitative analysis of regional airways obstruction using dynamic hyperpolarized 3He MRI-preliminary results in children with cystic fibrosis.

PURPOSE: To investigate regional airways obstruction in patients with cystic fibrosis (CF) with quantitative analysis of dynamic hyperpolarized (HP) (3)He MRI. MATERIALS AND METHODS: Dynamic radial projection MRI of HP (3)He gas was used to study respiratory dynamics in a group of eight children with CF. Signal kinetics in a total of seven regions of interest (ROIs; three in each lung, and one in the trachea) were compared with the results of spirometric pulmonary function tests (PFTs). The tracheal signal intensity was used as a form of "input function" to normalize for input flow effects. RESULTS: A pattern of low flow rate in the upper lobes was observed. When the flow measurements from the peripheral ROIs were averaged to obtain an index of flow in the peripheral lung, a good correlation was found (P = 3.74 x 10(-5)) with the forced expired volume in one second (FEV1). CONCLUSION: These results suggest that a quantitative measurement of localized airways obstruction in the early stages of CF may be obtained from dynamic (3)He MRI by using the slope of the signal rise as a measure of air flow into the peripheral lung. This study also demonstrates that children can cooperate well with the (3)He MRI technique.

Shape discrimination in the hippocampus using an MDL model.

We extend recent work on building 3D statistical shape models, automatically, from sets of training shapes and describe an application in shape analysis. Using an existing measure of model quality, based on a minimum description length criterion, and an existing method of surface re-parameterisation, we introduce a new approach to model optimisation that is scalable, more accurate, and involves fewer parameters than previous methods. We use the new approach to build a model of the right hippocampus, using a training set of 82 shapes, manually segmented from 3D MR images of the brain. We compare the results with those obtained using another previously published method for building 3D models, and show that our approach results in a model that is significantly more specific, general, and compact. The two models are used to investigate the hypothesis that there are differences in hippocampal shape between age-matched schizophrenic and normal control subgroups within the training set. Linear discriminant analysis is used to find the combination of shape parameters that best separates the two subgroups. We perform an unbiased test that shows there is a statistically significant shape difference using either shape model, but that the difference is more significant using the model built using our approach. We show also that the difference between the two subgroups can be visualised as a mode of shape variation.