Francisco P. M. Oliveira

Medical image registration: a review

This paper presents a review of automated image registration methodologies that have been used in the medical field. The aim of this paper is to be an introduction to the field, provide knowledge on the work that has been developed and to be a suitable reference for those who are looking for registration methods for a specific application. The registration methodologies under review are classified into intensity or feature based. The main steps of these methodologies, the common geometric transformations, the similarity measures and accuracy assessment techniques are introduced and described.

Registration of pedobarographic image data in the frequency domain

Image registration has been used to support pixel-level data analysis on pedobarographic image data sets. Some registration methods have focused on robustness and sacrificed speed, but a recent approach based on external contours offered both high computational processing speed and high accuracy. However, since contours can be influenced by local perturbations, we sought more global methods. Thus, we propose two new registration methods based on the Fourier transform, cross-correlation and phase correlation which offer high computational speed. We found out that both proposed methods revealed high accuracy for the similarity measures considered, using control geometric transformations. Additionally, both methods revealed high computational processing speed which, combined with their accuracy and robustness, allows their implementation in near-real-time applications. Furthermore, we found that the current methods were robust to moderate levels of noise, and consequently, do not require noise removal procedure like the contours method does.

Registration of plantar pressure images

In this work, five computational methodologies to register plantar pressure images are compared: (1) the first methodology is based on matching the external contours of the feet; (2) the second uses the phase correlation technique; (3) the third addresses the direct maximization of cross-correlation using the Fourier transform; (4) the fourth minimizes the sum of squared differences using the Fourier transform; and (5) the fifth methodology iteratively optimizes an intensity (dis)similarity measure based on Powells method. The accuracy and robustness of the five methodologies were assessed by using images from three common plantar pressure acquisition devices: a Footscan system, an EMED system, and a light reflection system. Using the residual error as a measure of accuracy, all methodologies revealed to be very accurate even in the presence of noise. The most accurate was the methodology based on the iterative optimization, when the mean squared error was minimized. It achieved a residual error inferior to 0.01 mm and 0.6 mm for non-noisy and noisy images, respectively. On the other hand, the methodology based on image contour matching was the fastest, but its accuracy was the lowest.

Towards an efficient and robust foot classification from pedobarographic images

This paper presents a new computational framework for automatic foot classification from digital plantar pressure images. It classifies the foot as left or right and simultaneously calculates two well-known footprint indices: the Cavanaghs arch index (AI) and the modified AI. The accuracy of the framework was evaluated using a set of plantar pressure images from two common pedobarographic devices. The results were outstanding, as all feet under analysis were correctly classified as left or right and no significant differences were observed between the footprint indices calculated using the computational solution and the traditional manual method. The robustness of the proposed framework to arbitrary foot orientations and to the acquisition device was also tested and confirmed.

Computer-aided diagnosis of Parkinson?s disease based on [123I]FP-CIT SPECT binding potential images, using the voxels-as-features approach and support vector machines

OBJECTIVE The aim of the present study was to develop a fully-automated computational solution for computer-aided diagnosis in Parkinson syndrome based on [(123)I]FP-CIT single photon emission computed tomography (SPECT) images. APPROACH A dataset of 654 [(123)I]FP-CIT SPECT brain images from the Parkinsons Progression Markers Initiative were used. Of these, 445 images were of patients with Parkinsons disease at an early stage and the remainder formed a control group. The images were pre-processed using automated template-based registration followed by the computation of the binding potential at a voxel level. Then, the binding potential images were used for classification, based on the voxel-as-feature approach and using the support vector machines paradigm. MAIN RESULTS The obtained estimated classification accuracy was 97.86%, the sensitivity was 97.75% and the specificity 98.09%. SIGNIFICANCE The achieved classification accuracy was very high and, in fact, higher than accuracies found in previous studies reported in the literature. In addition, results were obtained on a large dataset of early Parkinsons disease subjects. In summation, the information provided by the developed computational solution potentially supports clinical decision-making in nuclear medicine, using important additional information beyond the commonly used uptake ratios and respective statistical comparisons. (ClinicalTrials.gov Identifier: NCT01141023).

Analysis of ground reaction force and electromyographic activity of the gastrocnemius muscle during double support.

Mechanisms associated with energy expenditure during gait have been extensively researched and studied. According to the double-inverted pendulum model energy expenditure is higher during double support, as lower limbs need to work to redirect the centre of mass velocity. This study looks into how the ground reaction force of one limb affects the muscle activity required by the medial gastrocnemius of the contralateral limb during step-to-step transition. Thirty-five subjects were monitored as to the medial gastrocnemius electromyographic activity of one limb and the ground reaction force of the contralateral limb during double support. After determination of the Pearson correlation coefficient (r), a moderate correlation was observed between the medial gastrocnemius electromyographic activity of the dominant leg and the vertical (Fz) and anteroposterior (Fy) components of ground reaction force of the non-dominant leg (r = 0.797, p < 0.0001; r = –0.807, p < 0.0001). A weak and moderate correlation was observed between the medial gastrocnemius electromyographic activity of the non-dominant leg and the Fz and Fy of the dominant leg, respectively (r = 0.442, p = 0.018; r = –0.684 p < 0.0001). The results obtained suggest that during double support, ground reaction force is associated with the electromyographic activity of the contralateral medial gastrocnemius and that there is an increased dependence between the ground reaction force of the non-dominant leg and the electromyographic activity of the dominant medial gastrocnemius.

Automated segmentation of the incus and malleus ossicles in conventional tri-dimensional computed tomography images

This article proposes a fully automated computational solution to segment the incus and malleus ear ossicles in conventional tri-dimensional X-ray computed tomography images. The solution uses a registration-based segmentation paradigm, followed by image segmentation refinement. It was tested against a dataset comprising 21 computed tomography volumetric images of the ear acquired using standard protocols and with resolutions varying from 0.162 × 0.162 × 0.6 to 0.166 × 0.166 × 1.0 mm3. The images used were randomly selected from subjects who had had a computed tomography examination of the ear due to ear-related pathologies. Dice’s coefficient and the Hausdorff distance were used to compare the results of the automated segmentation against those of a manual segmentation performed by two experts. The mean agreement between automated and manual segmentations was equal to 0.956 (Dice’s coefficient), and the mean Hausdorff distance among the shapes obtained was 1.14 mm, which is approximately equal to the maximum distance between the neighbouring voxels in the dataset tested. The results confirm that the automated segmentation of the incus and malleus ossicles in tri-dimensional images acquired from patients with ear-related pathologies, using conventional computed tomography scanners and standard protocols, is feasible, robust and accurate. Thus, the solution developed can be employed efficiently in computed tomography ear examinations to help radiologists and otolaryngologists in the evaluation of bi-dimensional slices by providing the related tri-dimensional model.

Extraction, selection and comparison of features for an effective automated computer-aided diagnosis of Parkinson’s disease based on [123I]FP-CIT SPECT images

PurposeThis work aimed to assess the potential of a set of features extracted from [123I]FP-CIT SPECT brain images to be used in the computer-aided “in vivo” confirmation of dopaminergic degeneration and therefore to assist clinical decision to diagnose Parkinson’s disease.MethodsSeven features were computed from each brain hemisphere: five standard features related to uptake ratios on the striatum and two features related to the estimated volume and length of the striatal region with normal uptake. The features were tested on a dataset of 652 [123I]FP-CIT SPECT brain images from the Parkinson’s Progression Markers Initiative. The discrimination capacities of each feature individually and groups of features were assessed using three different machine learning techniques: support vector machines (SVM), k-nearest neighbors and logistic regression.ResultsCross-validation results based on SVM have shown that, individually, the features that generated the highest accuracies were the length of the striatal region (96.5%), the putaminal binding potential (95.4%) and the striatal binding potential (93.9%) with no statistically significant differences among them. The highest classification accuracy was obtained using all features simultaneously (accuracy 97.9%, sensitivity 98% and specificity 97.6%). Generally, slightly better results were obtained using the SVM with no statistically significant difference to the other classifiers for most of the features.ConclusionsThe length of the striatal region uptake is clinically useful and highly valuable to confirm dopaminergic degeneration “in vivo” as an aid to the diagnosis of Parkinson’s disease. It compares fairly well to the standard uptake ratio-based features, reaching, at least, similar accuracies and is easier to obtain automatically. Thus, we propose its day to day clinical use, jointly with the uptake ratio-based features, in the computer-aided diagnosis of dopaminergic degeneration in Parkinson’s disease.

Matching Contours in Images using Curvature Information and Optimization based on Dynamic Programming

This paper presents a novel methodology to match contours of objects represented in images. In the matching, we use sets of ordered points extracted from the external contours of the objects. Each of these points defines a vertex of a polygon to be associated to the correspondent contour. To establish the matching, we compute a cost matching matrix by comparing the amplitudes of the angles defined by the vertices of one of the contours with the amplitudes of the angles defined by the vertices of the other contour. Afterwards, the optimal global matching that preserves the contours points orders is determined using an optimization algorithm based on dynamic programming; defining the optimal global matching as the one that presents the minimum sum of the costs of all individual matches established. Based on the matching found, we present a methodology to compute the geometric transformation of similarity that best aligns the contours matched. The obtained matching results were good for contours defined by few points and the computation time was always very low.

Patient-specific gamma-index analysis to evaluate 99mTc-MAA as a predictor for 90Y glass microspheres liver radioembolisation dosimetry

ABSTRACT Radioembolisation (RE) is a minimally invasive therapy for primary/metastatic liver tumours. Yttrium-90 (90Y) microspheres (MS) are infused through the hepatic artery. Positron emission tomography (PET) allows visualisation of the 90Y-MS distribution in the liver after the RE procedure. All patients are submitted to a pretherapeutic Technetium-99m (99mTc) macroaggregated albumin (MAA) perfusion scintigraphy and single-photon emission computed tomography (SPECT) liver imaging. This work investigates the value of pre-treatment 99mTc-MAA SPECT to predict intrahepatic 90Y-MS treatment dosimetry, using the gamma-index (γ-index) analytical method. Six treatments with MAA and MS administered in the same hepatic artery branches were retrospectively selected. Multimodal images were used for semi-automatic and manual segmentation of liver and tumour volumes. Absorbed dose was calculated on SPECT and PET maps using the Imalytics Research Workstation (Philips). The γ-index calculation and analysis were performed using ‘in-house’ software with multiple distance to agreement (DTA) and dose difference (DD) tolerance criteria. γ-index passing rate values of 80% and > 90% were achieved for respectively conservative 10 mm/10% and less conservative 15 mm/15% (DTA/DD) tolerance criteria. RE treatments performed in similar conditions to planning give γ-index passing rates (several DTA/DD criteria) indicating potential predictive power of dosimetry planning for post-radioembolisation dosimetry outcome.