Bastian Bier

Band-Pass Filter Design by Segmentation in Frequency Domain for Detection of Epithelial Cells in Endomicroscope Images

Voice hoarseness can have various reasons, one of them is a change of the vocal fold mucus. This change can be examined with micro endoscopes. Cell detection in these images is a difficult task, due to bad image quality, caused by noise and illumination variations. In previous works, it was observed that the repetitive pattern of the cell walls cause an elliptical shape in the Fourier domain [1, 2]. A manual segmentation and back transformation of this shape results in filtered images, where the cell detection is much easier [3]. The goal of this work is to automatically segment the elliptical shape in Fourier domain. Two different approaches are developed to get a suitable band-pass filter: a thresholding and an active contour method. After the band-pass filter is applied, the achieved results are superior to the manual segmentation case.

Left ventricular heart phantom for wall motion analysis

In interventional cardiology, three-dimensional anatomical and functional information of the cardiac chambers, e.g. the left ventricle, would have an important impact on diagnosis and therapy. With the technology of C-arm CT it is possible to reconstruct intraprocedural 3-D images from angiographic projection data. Due to the long acquisition time of several seconds, motion-related artifacts, like blurring or streaks, occur. Therefore, the heart dynamics need to be taken into account in order to improve the reconstruction results. When it comes to the evaluation of different motion estimation and compensation algorithms and techniques of motion analysis, there is still a lack of comparability of the final reconstructions and motion parameters between the research groups. Since the results are heavily dependent on the applied motion pattern and simulation parameters, the experiments are not reproducible. We try to overcome these problems by providing varying left heart ventricle phantom datasets, consisting of projection images as well as extracted surface meshes. Up to now, there are six different datasets available: one with a normal sinus rhythm, one with a normal sinus rhythm and a catheter, one with a lateral wall defect of the ventricle, two with a lateral contraction phase shift and one without any motion. The existing datasets are based on a phantom similar to the 4D XCAT phantom with a contrasted left ventricle, myocardium, and aorta. The geometry calibration and acquisition protocol from a real clinical C-arm scanner are used. A webpage is provided where the data and the necessary files are publicly available for download at conrad.stanford.edu/data/heart.

DeepDRR – A Catalyst for Machine Learning in Fluoroscopy-Guided Procedures

Machine learning-based approaches outperform competing methods in most disciplines relevant to diagnostic radiology. Interventional radiology, however, has not yet benefited substantially from the advent of deep learning, in particular because of two reasons: 1) Most images acquired during the procedure are never archived and are thus not available for learning, and 2) even if they were available, annotations would be a severe challenge due to the vast amounts of data. When considering fluoroscopy-guided procedures, an interesting alternative to true interventional fluoroscopy is in silico simulation of the procedure from 3D diagnostic CT. In this case, labeling is comparably easy and potentially readily available, yet, the appropriateness of resulting synthetic data is dependent on the forward model. In this work, we propose DeepDRR, a framework for fast and realistic simulation of fluoroscopy and digital radiography from CT scans, tightly integrated with the software platforms native to deep learning. We use machine learning for material decomposition and scatter estimation in 3D and 2D, respectively, combined with analytic forward projection and noise injection to achieve the required performance. On the example of anatomical landmark detection in X-ray images of the pelvis, we demonstrate that machine learning models trained on DeepDRRs generalize to unseen clinically acquired data without the need for re-training or domain adaptation. Our results are promising and promote the establishment of machine learning in fluoroscopy-guided procedures.

X-ray-transform Invariant Anatomical Landmark Detection for Pelvic Trauma Surgery

X-ray image guidance enables percutaneous alternatives to complex procedures. Unfortunately, the indirect view onto the anatomy in addition to projective simplification substantially increase the task-load for the surgeon. Additional 3D information such as knowledge of anatomical landmarks can benefit surgical decision making in complicated scenarios. Automatic detection of these landmarks in transmission imaging is challenging since image-domain features characteristic to a certain landmark change substantially depending on the viewing direction. Consequently and to the best of our knowledge, the above problem has not yet been addressed. In this work, we present a method to automatically detect anatomical landmarks in X-ray images independent of the viewing direction. To this end, a sequential prediction framework based on convolutional layers is trained on synthetically generated data of the pelvic anatomy to predict 23 landmarks in single X-ray images. View independence is contingent on training conditions and, here, is achieved on a spherical segment covering (120 x 90) degrees in LAO/RAO and CRAN/CAUD, respectively, centered around AP. On synthetic data, the proposed approach achieves a mean prediction error of 5.6 +- 4.5 mm. We demonstrate that the proposed network is immediately applicable to clinically acquired data of the pelvis. In particular, we show that our intra-operative landmark detection together with pre-operative CT enables X-ray pose estimation which, ultimately, benefits initialization of image-based 2D/3D registration.

Respiratory motion compensation in rotational angiography: Graphical model-based optimization of auto-focus measures

Non-recurrent intra-scan motion, such as respiration, corrupts rotational coronary angiography acquisitions and inhibits uncompensated 3D reconstruction. Therefore, state-of-the-art algorithms that rely on 3D/2D registration of initial reconstructions to the projection data are unfavorable as prior models of sufficient quality cannot be obtained. To overcome this limitation, we propose a compensation method that optimizes a task-based autofocus measure using graphical model-based optimization.

Scatter correction using a primary modulator on a clinical angiography C‐arm CT system

Purpose Cone beam computed tomography (CBCT) suffers from a large amount of scatter, resulting in severe scatter artifacts in the reconstructions. Recently, a new scatter correction approach, called improved primary modulator scatter estimation (iPMSE), was introduced. That approach utilizes a primary modulator that is inserted between the X‐ray source and the object. This modulation enables estimation of the scatter in the projection domain by optimizing an objective function with respect to the scatter estimate. Up to now the approach has not been implemented on a clinical angiography C‐arm CT system. Methods In our work, the iPMSE method is transferred to a clinical C‐arm CBCT. Additional processing steps are added in order to compensate for the C‐arm scanner motion and the automatic X‐ray tube current modulation. These challenges were overcome by establishing a reference modulator database and a block‐matching algorithm. Experiments with phantom and experimental in vivo data were performed to evaluate the method. Results We show that scatter correction using primary modulation is possible on a clinical C‐arm CBCT. Scatter artifacts in the reconstructions are reduced with the newly extended method. Compared to a scan with a narrow collimation, our approach showed superior results with an improvement of the contrast and the contrast‐to‐noise ratio for the phantom experiments. In vivo data are evaluated by comparing the results with a scan with a narrow collimation and with a constant scatter correction approach. Conclusions Scatter correction using primary modulation is possible on a clinical CBCT by compensating for the scanner motion and the tube current modulation. Scatter artifacts could be reduced in the reconstructions of phantom scans and in experimental in vivo data.

Comparison of Different Approaches for Measuring Tibial Cartilage Thickness

Abstract Osteoarthritis is a degenerative disease affecting bones and cartilage especially in the human knee. In this context, cartilage thickness is an indicator for knee cartilage health. Thickness measurements are performed on medical images acquired in-vivo. Currently, there is no standard method agreed upon that defines a distance measure in articular cartilage. In this work, we present a comparison of different methods commonly used in literature. These methods are based on nearest neighbors, surface normal vectors, local thickness and potential field lines. All approaches were applied to manual segmentations of tibia and lateral and medial tibial cartilage performed by experienced raters. The underlying data were contrast agent-enhanced cone-beam C-arm CT reconstructions of one healthy subject’s knee. The subject was scanned three times, once in supine position and two times in a standing weight-bearing position. A comparison of the resulting thickness maps shows similar distributions and high correlation coefficients between the approaches above 0.90. The nearest neighbor method results on average in the lowest cartilage thickness values, while the local thickness approach assigns the highest values. We showed that the different methods agree in their thickness distribution. The results will be used for a future evaluation of cartilage change under weight-bearing conditions.

Epipolar Consistency Conditions for Motion Correction in Weight-Bearing Imaging

Recent C-arm CT systems allow for the examination of a patient’s knees under weight-bearing conditions. The standing patient tends to show involuntary motion, which introduces motion artifacts in the reconstruction. The state-of-the-art motion correction approach uses fiducial markers placed on the patients’ skin to estimate rigid leg motion. Marker placement is tedious, time consuming and associated with patient discomfort. Further, motion on the skin surface does not reflect the internal bone motion. We propose a purely projection based motion estimation method using consistency conditions of X-ray projections. The epipolar consistency between all pairs of projections is optimized over various motion parameters. We validate our approach by simulating motion from a tracking system in forward projections of clinical data. We visually and numerically assess reconstruction image quality and show an improvement in Structural Similarity from 0.912 for the uncorrected case to 0.943 using the proposed method with a 3D translational motion model. Initial experiments showed promising results encouraging further investigation of practical applicability.

Truncation correction for VOI C-arm CT using scattered radiation

In C-arm computed tomography, patient dose reduction by volume-of-interest (VOI) imaging is of increasing interest for many clinical applications. A remaining limitation of VOI imaging is the truncation artifact when reconstructing a 3D volume. It can either be cupping towards the boundaries of the field-of-view (FOV) or an incorrect offset in the Hounsfield values of the reconstructed voxels. In this paper, we present a new method for correction of truncation artifacts in a collimated scan. When axial or lateral collimation are applied, scattered radiation still reaches the detector and is recorded outside of the FOV. If the full area of the detector is read out we can use this scattered signal to estimate the truncated part of the object. We apply three processing steps: detection of the collimator edge, adjustment of the area outside the FOV, and interpolation of the collimator edge. Compared to heuristic truncation correction methods we were able to reconstruct high contrast structures like bones outside of the FOV. Inside the FOV we achieved similar reconstruction results as with water cylinder truncation correction. These preliminary results indicate that scattered radiation outside the FOV can be used to improve image quality and further research in this direction seems beneficial.

JOINT calibration and motion estimation in weight-bearing cone-beam CT of the knee joint using fiducial markers

Recently, C-arm cone-beam CT systems have been used to acquire knee joints under weight-bearing conditions. For this purpose, the C-arm acquires images on a horizontal trajectory around the standing patient, who shows involuntary motion. The current state-of-the-art reconstruction approach estimates motion based on fiducial markers attached to the knee. A drawback is that this method requires calibration prior to each scan, since the horizontal trajectory is not reproducible. In this work, we propose a novel method, which does not need a calibration scan. For comparison, we extended the state-of-the-art method with an iterative scheme and we further introduce a closed-form solution of the compensated projection matrices. For evaluation, a numerical phantom and clinical data are used. The novel approach and the extended state-of-the-art method achieve a reduction of the reprojection error of 94% for the phantom data. The improvement for the clinical data ranged between 10% and 80%, which is followed by the visual impression. Therefore, the novel approach and the extended state-of-the-art method achieve superior results compared to the state-of-the-art method.