Spatio-temporal classification for polyp diagnosis.

Colonoscopy remains the gold standard investigation for colorectal cancer screening as it offers the opportunity to both detect and resect pre-cancerous polyps. Computer-aided polyp characterisation can determine which polyps need polypectomy and recent deep learning-based approaches have shown promising results as clinical decision support tools. Yet polyp appearance during a procedure can vary, making automatic predictions unstable. In this paper, we investigate the use of spatio-temporal information to improve the performance of lesions classification as adenoma or non-adenoma. Two methods are implemented showing an increase in performance and robustness during extensive experiments both on internal and openly available benchmark datasets.

Polyp detection on video colonoscopy using a hybrid 2D/3D CNN.

Colonoscopy is the gold standard for early diagnosis and pre-emptive treatment of colorectal cancer by detecting and removing colonic polyps. Deep learning approaches to polyp detection have shown potential for enhancing polyp detection rates. However, the majority of these systems are developed and evaluated on static images from colonoscopies, whilst in clinical practice the treatment is performed on a real-time video feed. Non-curated video data remains a challenge, as it contains low-quality frames when compared to still, selected images often obtained from diagnostic records. Nevertheless, it also embeds temporal information that can be exploited to increase predictions stability. A hybrid 2D/3D convolutional neural network architecture for polyp segmentation is presented in this paper. The network is used to improve polyp detection by encompassing spatial and temporal correlation of the predictions while preserving real-time detections. Extensive experiments show that the hybrid method outperforms a 2D baseline. The proposed architecture is validated on videos from 46 patients and on the publicly available SUN polyp database. A higher performance and increased generalisability indicate that real-world clinical implementations of automated polyp detection can benefit from the hybrid algorithm and the inclusion of temporal information.

Evaluation of Automated Multiclass Fluid Segmentation in Optical Coherence Tomography Images Using the Pegasus Fluid Segmentation Algorithms.

Purpose: To evaluate the performance of the Pegasus-OCT (Visulytix Ltd) multiclass automated fluid segmentation algorithms on independent spectral domain optical coherence tomography data sets. Methods: The Pegasus automated fluid segmentation algorithms were applied to three data sets with edematous pathology, comprising 750, 600, and 110 b-scans, respectively. Intraretinal fluid (IRF), sub-retinal fluid (SRF), and pigment epithelial detachment (PED) were automatically segmented by Pegasus-OCT for each b-scan where ground truth from data set owners was available. Detection performance was assessed by calculating sensitivities and specificities, while Dice coefficients were used to assess agreement between the segmentation methods. Results: For two data sets, IRF detection yielded promising sensitivities (0.98 and 0.94, respectively) and specificities (1.00 and 0.98) but less consistent agreement with the ground truth (dice coefficients 0.81 and 0.59); likewise, SRF detection showed high sensitivity (0.86 and 0.98) and specificity (0.83 and 0.89) but less consistent agreement (0.59 and 0.78). PED detection on the first data set showed moderate agreement (0.66) with high sensitivity (0.97) and specificity (0.98). IRF detection in a third data set yielded less favorable agreement (0.46-0.57) and sensitivity (0.59-0.68), attributed to image quality and ground truth grader discordance. Conclusions: The Pegasus automated fluid segmentation algorithms were able to detect IRF, SRF, and PED in SD-OCT b-scans acquired across multiple independent data sets. Dice coefficients and sensitivity and specificity values indicate the potential for application to automated detection and monitoring of retinal diseases such as age-related macular degeneration and diabetic macular edema. Translational Relevance: The potential of Pegasus-OCT for automated fluid quantification and differentiation of IRF, SRF, and PED in OCT images has application to both clinical practice and research.

DISEASE CLASSIFICATION OF MACULAR OPTICAL COHERENCE TOMOGRAPHY SCANS USING DEEP LEARNING SOFTWARE: Validation on Independent, Multicenter Data.

PURPOSE: To evaluate Pegasus optical coherence tomography (OCT), a clinical decision support software for the identification of features of retinal disease from macula OCT scans, across heterogenous populations involving varying patient demographics, device manufacturers, acquisition sites, and operators. METHODS: Five thousand five hundred and eighty-eight normal and anomalous macular OCT volumes (162,721 B-scans), acquired at independent centers in five countries, were processed using the software. Results were evaluated against ground truth provided by the data set owners. RESULTS: Pegasus-OCT performed with areas under the curve of the receiver operating characteristic of at least 98% for all data sets in the detection of general macular anomalies. For scans of sufficient quality, the areas under the curve of the receiver operating characteristic for general age-related macular degeneration and diabetic macular edema detection were found to be at least 99% and 98%, respectively. CONCLUSION: The ability of a clinical decision support system to cater for different populations is key to its adoption. Pegasus-OCT was shown to be able to detect age-related macular degeneration, diabetic macular edema, and general anomalies in OCT volumes acquired across multiple independent sites with high performance. Its use thus offers substantial promise, with the potential to alleviate the burden of growing demand in eye care services caused by retinal disease.

Automated mediastinal lymph node detection from CT volumes based on intensity targeted radial structure tensor analysis.

This paper presents a local intensity structure analysis based on an intensity targeted radial structure tensor (ITRST) and the blob-like structure enhancement filter based on it (ITRST filter) for the mediastinal lymph node detection algorithm from chest computed tomography (CT) volumes. Although the filter based on radial structure tensor analysis (RST filter) based on conventional RST analysis can be utilized to detect lymph nodes, some lymph nodes adjacent to regions with extremely high or low intensities cannot be detected. Therefore, we propose the ITRST filter, which integrates the prior knowledge on detection target intensity range into the RST filter. Our lymph node detection algorithm consists of two steps: (1) obtaining candidate regions using the ITRST filter and (2) removing false positives (FPs) using the support vector machine classifier. We evaluated lymph node detection performance of the ITRST filter on 47 contrast-enhanced chest CT volumes and compared it with the RST and Hessian filters. The detection rate of the ITRST filter was 84.2% with 9.1 FPs/volume for lymph nodes whose short axis was at least 10 mm, which outperformed the RST and Hessian filters.

Geodesic patch-based segmentation.

Label propagation has been shown to be effective in many automatic segmentation applications. However, its reliance on accurate image alignment means that segmentation results can be affected by any registration errors which occur. Patch-based methods relax this dependence by avoiding explicit one-to-one correspondence assumptions between images but are still limited by the search window size. Too small, and it does not account for enough registration error; too big, and it becomes more likely to select incorrect patches of similar appearance for label fusion. This paper presents a novel patch-based label propagation approach which uses relative geodesic distances to define patient-specific coordinate systems as spatial context to overcome this problem. The approach is evaluated on multi-organ segmentation of 20 cardiac MR images and 100 abdominal CT images, demonstrating competitive results.

Hierarchical manifold learning for regional image analysis.

We present a novel method of hierarchical manifold learning which aims to automatically discover regional properties of image datasets. While traditional manifold learning methods have become widely used for dimensionality reduction in medical imaging, they suffer from only being able to consider whole images as single data points. We extend conventional techniques by additionally examining local variations, in order to produce spatially-varying manifold embeddings that characterize a given dataset. This involves constructing manifolds in a hierarchy of image patches of increasing granularity, while ensuring consistency between hierarchy levels. We demonstrate the utility of our method in two very different settings: 1) to learn the regional correlations in motion within a sequence of time-resolved MR images of the thoracic cavity; 2) to find discriminative regions of 3-D brain MR images associated with neurodegenerative disease.

Hierarchical manifold learning.

We present a novel method of hierarchical manifold learning which aims to automatically discover regional variations within images. This involves constructing manifolds in a hierarchy of image patches of increasing granularity, while ensuring consistency between hierarchy levels. We demonstrate its utility in two very different settings: (1) to learn the regional correlations in motion within a sequence of time-resolved images of the thoracic cavity; (2) to find discriminative regions of 3D brain images in the classification of neurodegenerative disease,

Groupwise combined segmentation and registration for atlas construction.

The creation of average anatomical atlases has been a growing area of research in recent years. It is of increased value to construct representations of, not only intensity atlases, but also their segmentation into required tissues or structures. This paper presents novel groupwise combined segmentation and registration approaches, which aim to simultaneously improve both the alignment of intensity images to their average shape, as well as the segmentations of structures in the average space. An iterative EM framework is used to build average 3D MR atlases of populations for which prior atlases do not currently exist: preterm infants at one- and two-years old. These have been used to quantify the growth of tissues occurring between these ages.

Similarity metrics for groupwise non-rigid registration.

The use of groupwise registration techniques for average atlas construction has been a growing area of research in recent years. One particularly challenging component of groupwise registration is finding scalable and effective groupwise similarity metrics; these do not always extend easily from pairwise metrics. This paper investigates possible choices of similarity metrics and additionally proposes a novel metric based on Normalised Mutual Information. The described groupwise metrics are quantitatively evaluated on simulated and 3D MR datasets, and their performance compared to equivalent pairwise registration.

Early growth in brain volume is preserved in the majority of preterm infants.

OBJECTIVE: Preterm infants have reduced cerebral tissue volumes in adolescence. This study addresses the question: Is reduced global brain growth in the neonatal period inevitable after premature birth, or is it associated with specific medical risk factors? METHODS: Eighty-nine preterm infants at term equivalent age without focal parenchymal brain lesions were studied with 20 full-term control infants. Using a deformation-based morphometric approach, we transformed images to a reference anatomic space, and we used the transformations to calculate whole-brain volume and ventricular volume for each subject. Patterns of volume difference were correlated with clinical data. RESULTS: Cerebral volume is not reduced compared with term born control infants (p = 0.765). Supplemental oxygen requirement at 28 postnatal days is associated with lower cerebral tissue volume at term (p < 0.001), but there were no significant differences in cerebral volumes attributable to perinatal sepsis (p = 0.515) and quantitatively defined diffuse white matter injury (p = 0.183). As expected, the ventricular system is significantly larger in preterm infants at term equivalent age compared with term control infants (p < 0.001). INTERPRETATION: Cerebral volume is not reduced during intensive care for the majority of preterm infants, but prolonged supplemental oxygen dependence is a risk factor for early attenuation of global brain growth. The reduced cerebral tissue volume seen in adolescents born preterm does not appear to be an inevitable association of prematurity, but rather caused by either specific disease during intensive care or factors operating beyond the neonatal period.

Abnormal deep grey matter development following preterm birth detected using deformation-based morphometry.

Preterm birth is a leading risk factor for neurodevelopmental and cognitive impairment in childhood and adolescence. The most common known cerebral abnormality among preterm infants at term equivalent age is a diffuse white matter abnormality seen on magnetic resonance (MR) images. It occurs with a similar prevalence to subsequent impairment, but its effect on developing neural systems is unknown. MR images were obtained at term equivalent age from 62 infants born at 24-33 completed weeks gestation and 12 term born controls. Tissue damage was quantified using diffusion-weighted imaging, and deformation-based morphometry was used to make a non-subjective survey of the whole brain to identify significant cerebral morphological alterations associated with preterm birth and with diffuse white matter injury. Preterm infants at term equivalent age had reduced thalamic and lentiform volumes without evidence of acute injury in these regions (t = 5.81, P < 0.05), and these alterations were more marked with increasing prematurity (t = 7.13, P < 0.05 for infants born at less than 28 weeks) and in infants with diffuse white matter injury (t = 6.43, P < 0.05). The identification of deep grey matter growth failure in association with diffuse white matter injury suggests that white matter injury is not an isolated phenomenon, but rather, it is associated with the maldevelopment of remote structures. This could be mediated by a disturbance to corticothalamic connectivity during a critical period in cerebral development. Deformation-based morphometry is a powerful tool for modelling the developing brain in health and disease, and can be used to test putative aetiological factors for injury.

Generalised overlap measures for assessment of pairwise and groupwise image registration and segmentation.

Effective validation techniques are an essential pre-requisite for segmentation and non-rigid registration techniques to enter clinical use. These algorithms can be evaluated by calculating the overlap of corresponding test and gold-standard regions. Common overlap measures compare pairs of binary labels but it is now common for multiple labels to exist and for fractional (partial volume) labels to be used to describe multiple tissue types contributing to a single voxel. Evaluation studies may involve multiple image pairs. In this paper we use results from fuzzy set theory and fuzzy morphology to extend the definitions of existing overlap measures to accommodate multiple fractional labels. Simple formulas are provided which define single figures of merit to quantify the total overlap for ensembles of pairwise or groupwise label comparisons. A quantitative link between overlap and registration error is established by defining the overlap tolerance. Experiments are performed on publicly available labeled brain data to demonstrate the new measures in a comparison of pairwise and groupwise registration.