Changes In Image Intensity Research Articles

Rapidly count crop seedling emergence based on waveform Method(WM) using drone imagery at the early stage

Effectively monitoring seedling emergence is critical to identify missing cotton seedling at early stages, allowing the prompt replenishment of the seedlings to maintain crop yield. However, traditional manual inspections have the limitations with low efficiency, poor timeliness, and large counting errors. Although UAV imagery has been applied in seedling monitoring, there are still chance in cotton seedling monitoring under complex environment conditions. Therefore, there is a urgent need for an efficient, rapid, and low-cost method for monitoring early seedling emergence in cotton crops. This study utilizes RGB images of early seedling emergence captured from an unmanned aerial vehicle (UAV) at a flight height of 10 m and a resolution of 0.33 cm in Shihezi, Xinjiang, on 7 and 12 days after seeding (DAS). The modified Hough transform method is constructed for line detection, and crop rows are subsequently extracted from the lines and masked images. After masking, the waveform is extracted from the excess green index (ExG) greyscale map and smoothed to determine the peaks. Seedlings are counted and located by numbering and positioning the peaks, meanwhile their coordinates and distance determine the number and location of missing seedlings. Images with varying resolutions (ranging from 0.33 to 2.5 cm) and intensity changes (0.8 to 1.2) verify the accuracy and efficiency of the proposed waveform method(WM). The results shown that at 7 DAS, the root mean square error (RMSE) is less than 2.0 plants/row, the relative root mean squared error (RRMSE) is less than 2 %, and the coefficient of determination (R2) is stable at 0.95–0.96 when the resolution is below 1.0 cm. The RMSE, RRMSE and R2 fluctuate more when the resolution is less than 1.0 cm. At 12 DAS, the counting accuracy is stable under different conditions, and the RMSE and RRMSE are distributed smoothly over the range of 1.76–4.39 plants/row and 1.29–5.47 % with different resolutions. The R2 is in the range of 0.88–0.97, RRMSE is 1.90–3.39 % and the RMSE is 1.65–2.90 plants/row under different intensities. The results of cotton seedling counting using the waveform method are less affected by image resolution and intensity changes. Therefore, the proposed WM provides efficient, accurate, and automatic counting for cotton seedlings in response to complex weather conditions. Furthermore, it would give a novel approach to accurately countequally spaced row-sown crops based on UAV images.

Edge-guided two-stage feature matching for infrared and visible image registration in electric power scenes

Collaborative processing of infrared and visible (IR–VS) images is essential in identifying potential equipment failures during power inspections. However, due to significant differences in spectrum, resolution, and intensity between IR–VS images, it is still challenging to obtain an accurate spatial registration for them. To address these issues, we propose the edge-guided two-stage registration strategy for various kinds of real-world applications. We first develop a heated, hollow calibration board tailored to IR–VS cameras, thus conducting offline calibration to estimate the scale and resolution relationships. The IR–VS images are pretransformed into a similar scale to improve registration accuracy. Furthermore, we propose the edge-guided feature matching network (EGFMNet) to extract feature points and build associated descriptors end-to-end. Specifically, benefiting from the excellent description ability of the network for image intensity changes, we directly adopt multimodal edge maps to learn more consistent features for accurate registration. Meanwhile, the local–global consistent features from edge maps are extracted to eliminate the nonlinear radiation distortion. By jointly learning the repeatability and reliability information, robust feature points with distinctive descriptors can be extracted and matched for homography estimation. Comparative and ablation experiments demonstrate the effectiveness and practicality of the proposed registration strategy in real-world power inspection scenarios. Extended experiments on several public datasets further demonstrate the superiority of the proposed method.

A control chart for monitoring images using jump location curves

Image monitoring is a relatively new research area in statistics and machine learning that has wide applications in various fields including medical diagnostics and subsequently disease monitoring, satellite imaging, security systems, and so forth. In the literature, a vast majority of the methods use image intensity changes to detect out of control images. However, this approach is often unreasonable in many real-life applications where a change in contrast between the background and foreground of an image should not indicate an out of control image as long as the boundaries of the image objects remain unchanged. In this article, we propose a Shewhart-type control chart to monitor grayscale images using detected edges of the images. The central idea is to monitor the Hausdorff distance between the point-set of detected edge pixels in each image from the corresponding point-set of the estimated true in-control image. The proposed monitoring procedure should be easy to execute many real-life applications. Numerical studies show that it performs well in various types of situations in comparison with a number of competing methods.

Extreme learning machine algorithm‐based model for lung cancer classification from histopathological real‐time images

AbstractIn the present scenario, developing an automatic and credible diagnostic system to analyze lung cancer type, stage, and level from computed tomography (C.T.) images is a very challenging task, even for experienced pathologists, due to the nonuniform illumination and artifacts. The nonuniform illumination and artifacts are the low‐frequency changes in image intensity that arise from the sensor and the person's movement while recording the C.T. scanned images. Although numerous machine learning techniques are used to improve the effectiveness of automatic lung cancer diagnostic systems, the classification accuracy of these systems still needs significant improvement to satisfy the real‐time requirement of the diagnostic situations. A new extreme learning machine (ELM) algorithm‐based model (hereafter called XlmNet) is proposed to classify the histopathology scans effectively. XlmNet utilizes The Cancer Imaging Archive (TCIA) dataset. After data collection, the initial stage in XlmNet is preprocessing, including noise removal, histogram equalization, and quality‐improved image. The enhanced Profuse Clustering (EPC) method is implemented for segmenting the affected regions from C.T. scans by image segment using superpixel clustering. The statistical attributes are extracted by using Principal Component Analysis (PCA). ELM classifier helps in classifying the lung nodules. The empirical results of the XlmNet model are related to some advanced classifiers concerning performance metrics. The evaluations of XlmNet on the TCIA dataset reveal that XlmNet outperforms other classification networks with the Accuracy of 0.965, a sensitivity of 0.964, a specificity of 0.865, a precision of 0.962, a Jaccard similarity score (JSS) of 0.95.

Does MR Signal Intensity Change have Prognostic Value in Multilevel Cervical Myelopathy? A Single-Center, Prospective Study.

Neurological recovery in cervical myelopathy remains unpredictable. There is contradictory literature regarding the prognostic value of magnetic resonance imaging (MRI) in such cases. The objective of the present study is to evaluate the morphological changes in the spinal cord in cervical spondylotic myelopathy and compare them with clinical outcome. This is a single-center, prospective, observational study. All patients with multilevel (two or more levels) cervical spondylotic myelopathy undergoing anterior spine surgery were included in the study. Patient demographics and radiological findings were recorded. MRI was repeated immediately post-op and at 1-year follow-up. MRI classification system based on axial images was used to evaluate presurgery and postsurgery changes and correlate them with clinical information. The study comprised 50 patients (40 males and 10 females) with a mean age of 59.5 years. Average duration of symptoms before surgery was 6.29 months. Thirty-four patients underwent two-level decompression, while 16 patients underwent more than two-level decompression. Average duration of follow-up was 26.82 months. Mean pre-op Nurick grade was 2.84, and mean recovery rate was 56.73. Most common pre-op MRI type was type 1. Analysis of data by logistic regression showed better recovery rate with lower age, lower pre-op Nurick grade, and lower pre-op MRI type. MR classification based on signal intensity changes in axial images have been found to correlate with recovery rate.

H-Scan Ultrasound Monitoring of Breast Cancer Response to Chemotherapy and Validation With Diffusion-Weighted Magnetic Resonance Imaging.

H-scan ultrasound (US) imaging is a novel tissue characterization technique to detect apoptosis-induced changes in cancer cells after the initiation of effective drug treatment. The objective of the proposed research was to assess the sensitivity of 3-dimensional (3D) H-scan US technique for monitoring the response of breast cancer-bearing animals to neoadjuvant chemotherapy and correlate results to diffusion-weighted magnetic resonance imaging (DW-MRI) measurements of programmed cancer cell death. Experimental studies used female mice (N=18) implanted with human breast cancer cells. Animals underwent H-scan US and DW-MRI imaging on days 0, 1, 3, 7, and 10. After imaging at day 0, breast tumor-bearing nude mice were treated biweekly with an apoptosis-inducing drug. Texture analysis of H-scan US images explored spatial relationships between local US scattering. At day 10, H-scan measurements were compared with DW-MRI-derived apparent diffusion coefficient (ADC) values and histological findings. H-scan US imaging of low and high dose cisplatin-treated cancer-bearing animals revealed changes in image intensity suggesting a progressive decrease in aggregate US scatterer size that was not observed in control animals. Longitudinal trends discovered in H-scan US result matched with texture analysis and DW-MRI (P < .01). Further, analysis of the H-scan US image intensity and corresponding DW-MRI-derived ADC values revealed a strong linear correlation (R2 =.93, P < .001). These changes were due to cancer cell apoptotic activity and consider as early detectable biomarker during treatment. The 3D H-scan technique holds promise for assisting clinicians in monitoring the early response of breast cancer tumor to neoadjuvant chemotherapy and adding value to traditional diagnostic ultrasound examinations.

Hybrid sparse monocular visual odometry with online photometric calibration

Most monocular visual Simultaneous Localization and Mapping (vSLAM) and visual odometry (VO) algorithms focus on either feature-based methods or direct methods. Hybrid (semi-direct) approach is less studied although it is equally important. In this paper, a hybrid sparse visual odometry (HSO) algorithm with online photometric calibration is proposed for monocular vision. HSO introduces two novel measures, that is, direct image alignment with adaptive mode selection and image photometric description using ratio factors, to enhance the robustness against dramatic image intensity changes and motion blur. Moreover, HSO is able to establish pose constraints between keyframes far apart in time and space by using KLT tracking enhanced with a local-global brightness consistency. The convergence speed of candidate map points is adopted as the basis for keyframe selection, which strengthens the coordination between the front end and the back end. Photometric calibration is elegantly integrated into the VO system working in tandem: (1) Photometric interference from the camera, such as vignetting and changes in exposure time, is accurately calibrated and compensated in HSO, thereby improving the accuracy and robustness of VO. (2) On the other hand, VO provides pre-calculated data for the photometric calibration algorithm, which reduces resource consumption and improves the estimation accuracy of photometric parameters. Extensive experiments are performed on various public datasets to evaluate the proposed HSO against the state-of-the-art monocular vSLAM/VO and online photometric calibration methods. The results show that the proposed HSO achieves superior performance on VO and photometric calibration in terms of accuracy, robustness, and efficiency, being comparable with the state-of-the-art VO/vSLAM systems. We open source HSO for the benefit of the community.

Deformable Groupwise Image Registration using Low-Rank and Sparse Decomposition

Groupwise image registration describes the problem of simultaneously aligning a series of more than two images through individual spatial deformations and it is a common task in the processing of medical image sequences. Variational methods with data fidelity terms based on robust PCA (RPCA) have proven successful in accounting for structural changes in image intensity stemming, e.g., from the uptake of a contrast agent in functional imaging. In this article, we investigate the drawbacks of the most commonly used RPCA data term and derive an improved replacement that employs explicit constraints instead of penalties. We further present a multilevel scheme with theoretically justified scaling to solve the underlying fully deformable registration model. Our numerical experiments on synthetic and real-life medical data confirm the advanced adaptability of RPCA-based data terms and showcase an improved registration accuracy of our algorithm when compared to related groupwise approaches.

A Multi-Dataset Evaluation of Frame Censoring for Motion Correction in Task-Based fMRI.

Subject motion during fMRI can affect our ability to accurately measure signals of interest. In recent years, frame censoring—that is, statistically excluding motion-contaminated data within the general linear model using nuisance regressors—has appeared in several task-based fMRI studies as a mitigation strategy. However, there have been few systematic investigations quantifying its efficacy. In the present study, we compared the performance of frame censoring to several other common motion correction approaches for task-based fMRI using open data and reproducible workflows. We analyzed eight publicly available datasets representing 11 distinct tasks in child, adolescent, and adult participants. Performance was quantified using maximum t-values in group analyses, and region of interest–based mean activation and split-half reliability in single subjects. We compared frame censoring across several thresholds to the use of 6 and 24 canonical motion regressors, wavelet despiking, robust weighted least squares, and untrained ICA-based denoising, for a total of 240 separate analyses. Thresholds used to identify censored frames were based on both motion estimates (FD) and image intensity changes (DVARS). Relative to standard motion regressors, we found consistent improvements for modest amounts of frame censoring (e.g., 1–2% data loss), although these gains were frequently comparable to what could be achieved using other techniques. Importantly, no single approach consistently outperformed the others across all datasets and tasks. These findings suggest that the choice of a motion mitigation strategy depends on both the dataset and the outcome metric of interest.

Porous Silica Nanospheres with a Confined Mono(aquated) Mn(II)-Complex: A Potential T1-T2 Dual Contrast Agent for Magnetic Resonance Imaging.

Magnetic resonance imaging has emerged as an indispensable imaging modality for the early-stage diagnosis of many diseases. The imaging in the presence of a contrast agent is always advantageous, as it mitigates the low-sensitivity issue of the measurements and provides excellent contrast in the acquired images even in a short acquisition time. However, the stability and high relaxivity of the contrast agents remained a challenge. Here, molecules of a mononuclear, mono(aquated), thermodynamically stable [log KMnL = 14.80(7) and pMn = 8.97] Mn(II)-complex (1), based on a hexadentate pyridine-picolinate unit-containing ligand (H2PyDPA), were confined within a porous silica nanosphere in a noncovalent fashion to render a stable nanosystem, complex 1@SiO2NP. The entrapped complex 1 (complex 1@SiO2) exhibited r1 = 8.46 mM-1 s-1 and r2 = 33.15 mM-1 s-1 at pH = 7.4, 25 °C, and 1.41 T in N-(2-hydroxyethyl)piperazine-N'-ethanesulfonic acid buffer. The values were about 2.9 times higher compared to the free (unentrapped)-complex 1 molecules. The synthesized complex 1@SiO2NP interacted significantly with albumin protein and consequently boosted both the relaxivity values to r1 = 24.76 mM-1 s-1 and r2 = 63.96 mM-1 s-1 at pH = 7.4, 37 °C, and 1.41 T. The kinetic inertness of the entrapped molecules was established by recognizing no appreciable change in the r1 value upon challenging complex 1@SiO2NP with 30 and 40 times excess of Zn(II) ions at pH 6 and 25 °C. The water molecule coordinated to the Mn(II) ion in complex 1@SiO2 was also impervious to the physiologically relevant anions (bicarbonate, biphosphate, and citrate) and pH of the medium. Thus, it ensured the availability of the inner-coordination site of complex 1 for the coordination of water molecules in the biological media. The concentration-dependent changes in image intensities in T1- and T2-weighted phantom images and uptake of the nanoparticles by the HeLa cell put forward the biocompatible complex 1@SiO2NP as a potential dual-mode MRI contrast agent, an alternative to Gd(III)-containing contrast agents.

A novel bio-inspired texture descriptor based on biodiversity and taxonomic measures

Texture can be defined as the change of image intensity that forms repetitive patterns resulting from the physical properties of an object’s roughness or differences in a reflection on the surface. Considering that texture forms a system of patterns in a non-deterministic way, biodiversity concepts can help its characterization from an image. This paper proposes a novel approach to quantify such a complex system of diverse patterns through species diversity, richness, and taxonomic distinctiveness. The proposed approach considers each image channel as a species ecosystem and computes species diversity and richness as well as taxonomic measures to describe the texture. Furthermore, the proposed approach takes advantage of ecological patterns’ invariance characteristics to build a permutation, rotation, and translation invariant descriptor. Experimental results on three datasets of natural texture images and two datasets of histopathological images have shown that the proposed texture descriptor has advantages over several texture descriptors and deep methods.

Research on the Ability to Resist Grayscale Distribution Change of the Training Strategy by Involving Knowledge From Neighbored Organs for Segmenting Small One Using Convolution Neural Network in Thoracic Radiotherapy

Automatic small-volume organ segmentation using convolution neural network is sensitive to the distribution change of image intensity. To resist it, we propose to train a neural network by involving knowledge from spatially-neighbored organs. The investigated small organ was spinal cord in the thoracic region.The proposed network (PN) architecture was based on Unet, a popular structure for medical segmentation. Its last layer consisted of three convolution ones to delineate lung, heart and spinal cord respectively. Through this way, the shared layers would involve knowledge from lung and heart when segmenting spinal cord. This design is also known as multi-task learning. There were 4240 computed tomography (CT) images from our private dataset in this work. All of them encompassed the three organs and corresponded to supine position. Among them, 1240 were used for training (856/384 for training/validation) and the remaining were for test. To evaluate the ability of the proposed training strategy to resist gray-value distribution change, we rotated all testing scans at a random angle ranging from -30° to 30° to simulate various spatial distributions in grayscale. After training, PN was applied on the test set and assessed using dice score coefficient (DSC, a metric to measure overlap between two sets) and the number of DSC = 0 (abbreviated as N). To further validate the proposed training strategy, we trained another network without knowledge from other two organs. The comparative network (CN) had the same structure with PN, but its last layer only consisted of one convolution relating to the spinal cord segmentation. With the same optimization algorithm and training set, CN was also evaluated on the same test set using DSC and N.N was 669 for CN and 34 for the proposed one. It means that CN loses the ability of recognizing spinal cord in 22.3% (669/3000) test cases, but 1.1% (34/3000) for PN. The statistical results of DSC were summarized in Table 1. This table illustrates that CN and PN achieves comparable performances in 65% of test images, since they show similar P35 ∼ P90. For the rest 35% of test scans, PN outperforms CN, because the P30 and P25 of CN are significantly lower than PN, and P35 of CN equals to P30 of PN.The proposed training strategy was demonstrated superior in resisting grayscale distribution change. Besides, the test set was never "seen" during training. It proved that the proposed one could improve a network's generalization.

딥러닝 성능 향상을 위한 3D 의료 영상 증강법

This paper proposes a 3D image augmentation method for improving the generalization performance of deep neural networks. It allows us to enrich the diversity of training data samples that is essential in medical image segmentation tasks, thus reducing the data overfitting problem caused by the fact the scale of medical image dataset is typically smaller. It also enables us to predict medical segmentation surfaces in Euclidean space without additional labeled datasets. This method includes image transformation functions, which are comprised of a spatial deformation and image intensity change, enabling the synthesis of complex effects such as variations in anatomy and image acquisition procedures. Our numerical experiments demonstrate that the proposed approach provides significant improvements over state-of-the-art methods for 3D medical image segmentation.

Variation of skeletal muscle ultrasound imaging intensity in horses after treadmill exercise: a proof of concept for glycogen content estimation

BackgroundGlycogen in skeletal muscle is a major source of energy during exercise and an important determinant of endurance capacity, so that its measurement may provide a meaningful marker of athletes’ preparation and a possible predictor of performance, both in humans and in equines. Gold standard of glycogen concentration measurement is the histochemical and biochemical analysis of biopsy-derived muscle tissue, an invasive and potentially injuring procedure. Recently, high-frequency ultrasound (US) technology is being exploited in human sports medicine to estimate muscle glycogen content. Therefore, aim of the present study is to evaluate the feasibility of US assessment of muscle glycogen in equines.ResultsUS images of gluteus medius (GL) and semitendinosus (ST) muscles were obtained on eight healthy horses (3–10 years) before and after a steady-state exercise on treadmill (velocity: 4.0–12.5 m/s; duration: 2–20 min; heart rate: 137–218 b/min). Average image greyscale intensity was significantly different between GL and ST, both before and after exercise (p < 0.001). Comparing baseline and post-exercise US images, significant increase in greyscale intensity has been observed in ST (p < 0.001), but not in GL (p = 0.129). The volume of the exercise was significantly correlated with exercise-dependent change in image intensity (R2 = 0.891), consistent with a reduction of glycogen muscle stores resulting from aerobic activity.ConclusionsUS technique evidences also in horses muscle changes possibly associated to glycogen utilisation during exercise. Present results on a small sample need to be further confirmed and provide preliminary data warranting future validation by direct glycogen measurement through biopsy technique.

Implementation of HSV- based Thresholding Method for Iris Detection

Image thresholding is one of the most frequently used methods in image processing to perform digital image processing. Image thresholding has a technique that can separate the image object from its background. This is a technique that is quite good and effective for segmenting love. In this study, the threshold method used will be combined with the HSV mode for color detection. The threshold method will separate the object and the image background, while HSV will help improve the segmentation results based on the Hue, Saturation, Value values to be able to detect objects more accurately. Segmentation is carried out using the original input image without pre-processing or direct segmentation. As we know that in digital image processing, there are steps that are usually done to get a good input image, namely pre-processing. In this pre-processing stage, processes such as image conversion and image intensity changes are carried out so that the input image is better. Therefore, even though the input image is used without going through the pre-processing stage, the object can be segmented properly based on the color type of the object. The results of this segmentation can later be used for recognition and identification of image objects. The results of the test method for object segmentation achieved a color similarity level of 25%, with an accuracy rate of 75% in detecting uniform color objects. So that this method can be one of the most effective methods in segmenting image objects without pre-processing or direct thresholding

A dynamic lesion model for differentiation of malignant and benign pathologies

Malignant lesions have a high tendency to invade their surrounding environment compared to benign ones. This paper proposes a dynamic lesion model and explores the 2nd order derivatives at each image voxel, which reflect the rate of change of image intensity, as a quantitative measure of the tendency. The 2nd order derivatives at each image voxel are usually represented by the Hessian matrix, but it is difficult to quantify a matrix field (or image) through the lesion space as a measure of the tendency. We conjecture that the three eigenvalues contain important information of the Hessian matrix and are chosen as the surrogate representation of the Hessian matrix. By treating the three eigenvalues as a vector, called Hessian vector, which is defined in a local coordinate formed by three orthogonal Hessian eigenvectors and further adapting the gray level occurrence computing method to extract the vector texture descriptors (or measures) from the Hessian vector, a quantitative presentation for the dynamic lesion model is completed. The vector texture descriptors were applied to differentiate malignant from benign lesions from two pathologically proven datasets: colon polyps and lung nodules. The classification results not only outperform four state-of-the-art methods but also three radiologist experts.

What Role Does Trabecular Bone Score Play in Chronic Inflammatory Rheumatic Diseases?

Patients suffering from rheumatic inflammatory diseases, e.g., systemic sclerosis, rheumatoid arthritis, and ankylosing spondylitis, are at risk of low bone mass. Dual-energy X-ray Absorptiometry (DXA) is the traditional radiological measurement technique for bone mineral density (BMD). The recently developed trabecular bone score (TBS) enhances the skeletal information provided by standard BMD. It re-analyzes the spatial dynamics of pixel intensity changes in lumbar spine DXA images, defining a quantitative index, characterizing trabecular bone microarchitecture. It has been demonstrated that low TBS values are associated with an increased incidence of fractures in patients with rheumatic diseases. These methods used together for bone damage evaluation can be of value to identify individuals who will potentially fracture. The main scientific literature on the clinical aspects of osteoporosis, including the use of TBS in evaluating this pathology, are herein reported aimed at shedding light on the role trabecular bone score plays in chronic inflammatory rheumatic diseases.

Selection of a Similarity Measure Combination for a Wide Range of Multimodal Image Registration Cases

Many similarity measures (SMs) were proposed to measure the similarity between multimodal remote sensing (RS) images. Each SM is efficient to a different degree in different registration cases (we consider visible-to-infrared, visible-to-radar, visible-to-digital elevation model (DEM), and radar-to-DEM ones), but no SM was shown to outperform all other SMs in all cases. In this article, we investigate the possibility of deriving a more powerful SM by combining two or more existing SMs. This combined SM relies on a binary linear support vector machine (SVM) classifier trained using real RS images. In the general registration case, we order SMs according to their impact on the combined SM performance. The three most important SMs include two structural SMs based on modality independent neighborhood descriptor (MIND) and scale-invariant feature transform-octave (SIFT-OCT) descriptors and one area-based logarithmic likelihood ratio (logLR) SM: the former ones are more robust to structural changes of image intensity between registered modes, the latter one is to image noise. Importantly, we demonstrate that a single combined SM can be applied in the general case as well as in each particular considered registration case. As compared to existing multimodal SMs, the proposed combined SM [based on five existing SMs, namely, MIND, logLR, SIFT-OCT, phase correlation (PC), histogram of orientated phase congruency (HOPC)] increases the area under the curve (AUC) by from 1% to 21%. From a practical point of view, we demonstrate that complex multimodal image pairs can be successfully registered with the proposed combined SM, while existing single SMs fail to detect enough correspondences for registration. Our results demonstrate that MIND, SIFT, and logLR SMs capture essential aspects of the similarity between RS modes, and their properties are complementary for designing a new more efficient multimodal SM.

Artificial intelligence in detecting early RA

To prevent chronicity of Rheumatoid Arthritis (RA) by early treatment, detecting inflammatory signs in an early phase is essential. Since Magnetic Resonance Imaging (MRI) of the wrist, hand and foot can detect inflammation before it is clinically detectable, this modality may play an important role in achieving very early diagnoses. By collecting large amounts of MRI data from healthy controls and patients with arthralgia suspicious for progression to RA, patterns can be studied that are most specific for early development of RA. Furthermore, MRI can be used as outcome parameter for randomized placebo-controlled trials on early RA treatment, by detecting subtle changes in image intensities originating from natural progression or treatment effects. Very large amounts of MRI data, however, make manual quantification impractical and the coarse scale used in visual scoring systems (i.e. whole values between 0 and 3) limits its sensitivity to detect changes that are likely to be very subtle in such an early phase. In recent years, advances in artificial intelligence and especially ‘deep learning’ in interpreting medical images have shown that -in specific areas- a computerized analysis can outperform human observers. Therefore, research has been initiated into applying these artificial intelligence techniques to the quantification of early RA from MRI data. In this paper, an overview is given on the background and history of artificial intelligence, with a special focus on recent developments in ‘deep learning’, and how these techniques could be applied to detect subtle inflammatory changes in MRI data.

Segmentation Guided Registration for 3D Spectral-Domain Optical Coherence Tomography Images

Medical image registration can be used for combining information from multiple imaging modalities, monitoring changes in size, shape or image intensity over time intervals. However, the development of such technique can be challenging for 3D spectral-domain optical coherence tomography (SD-OCT) imaging, because SD-OCT image is inherently noisy and its high resolution leads to high complexity of non-rigid registration. In this paper, a new segmentation guided approach is reported for registration of retinal OCT data. The proposed method models the 3D registration as a two-stage registration including x-y direction registration and z direction registration. In x-y direction registration, the vessel maps of OCT projection images between the template and the subject are registered to find out x-y direction displacement. The multi-scale vessel enhancement filter and morphological thinning methods are used to extract the vessel maps from the projection image of 3D OCT scans. And then x-y direction displacement is estimated by matching Speeded-Up Robust Features of the vessel maps. In z direction registration, using the tissue map instead of the original intensity image, A-scans are aligned to get the local displacements in z direction. The proposed method was evaluated on 45 longitudinal retinal OCT scans from 15 subjects. Experimental results show that the proposed method is accurate and very efficient.