Günther Platsch

Optimized data preprocessing for multivariate analysis applied to 99mTc-ECD SPECT data sets of Alzheimer's patients and asymptomatic controls

Multivariate image analysis has shown potential for classification between Alzheimers disease (AD) patients and healthy controls with a high-diagnostic performance. As image analysis of positron emission tomography (PET) and single photon emission computed tomography (SPECT) data critically depends on appropriate data preprocessing, the focus of this work is to investigate the impact of data preprocessing on the outcome of the analysis, and to identify an optimal data preprocessing method. In this work, technetium-99methylcysteinatedimer (99mTc-ECD) SPECT data sets of 28 AD patients and 28 asymptomatic controls were used for the analysis. For a series of different data preprocessing methods, which includes methods for spatial normalization, smoothing, and intensity normalization, multivariate image analysis based on principal component analysis (PCA) and Fisher discriminant analysis (FDA) was applied. Bootstrap resampling was used to investigate the robustness of the analysis and the classification accuracy, depending on the data preprocessing method. Depending on the combination of preprocessing methods, significant differences regarding the classification accuracy were observed. For 99mTc-ECD SPECT data, the optimal data preprocessing method in terms of robustness and classification accuracy is based on affine registration, smoothing with a Gaussian of 12 mm full width half maximum, and intensity normalization based on the 25% brightest voxels within the whole-brain region.

Improved Anatomical Landmark Localization in Medical Images Using Dense Matching of Graphical Models

We propose a method for reliably and accurately identifying anatomical landmarks in 3D CT volumes based on dense matching of parts-based graphical models. Such a system can be used to establish reliable correspondences in medical images which can be useful on their own or as part of more complex processing e.g. atlas building. We propose and investigate novel methods for efficiently optimizing parameters of appearance models for landmark localization in 3D images. We also investigate the trade-off between the number of model parameters and registration accuracy. We present results for the localization of 22 landmarks in clinical 3D CT volumes of cancer patients and optimization of part-specific patch scales. Over-fitting is likely due to an intrinsically high variability of the data and a limited labeled training and test set, here 83 scans, so we employ a rigorous bootstrap analysis to validate the results. The average mean and maximum registration error over all landmarks is reduced by 31% and 25% for the optimized model, compared to an empirically determined baseline. Additionally, we show a significantly improved performance over standard methods as the number of free parameters increases from an isotropic patch scale shared by all parts, to specific anisotropic patch scales learnt for each part in the model.

Personalized Graphical Models for Anatomical Landmark Localization in Whole-Body Medical Images

The goal of this work is to accurately and reliably localize anatomical landmarks in 3D Computed Tomography (CT) scans of the upper bodies of cancer patients even in the presence of pathologies and imaging artifacts that may markedly change the appearances of anatomical structures. We propose a method based on dense matching of parts-based graphical models. For landmark localization, we replace population averaged models by personalized models that are adapted to each test image at runtime. We do so by jointly leveraging weighted combinations of labeled training exemplars. We report results for localizing standard anatomical landmarks in clinical 3D CT volumes, using a database of 83 lung cancer patients. We compare our method against both (baseline) population averaged graphical models and against atlas-based deformable registration and show the method is in each case able to localize landmarks with significantly improved reliability and accuracy.

Personalization of pictorial structures for anatomical landmark localization

We propose a method for accurately localizing anatomical landmarks in 3D medical volumes based on dense matching of parts-based graphical models. Our novel approach replaces population mean models by jointly leveraging weighted combinations of labeled exemplars (both spatial and appearance) to obtain personalized models for the localization of arbitrary landmarks in upper body images. We compare the method to a baseline population-mean graphical model and atlas-based deformable registration optimized for CT-CT registration, by measuring the localization accuracy of 22 anatomical landmarks in clinical 3D CT volumes, using a database of 83 lung cancer patients. The average mean localization error across all landmarks is 2.35 voxels. Our proposed method outperforms deformable registration by 73%, 93% for the most improved landmark. Compared to the baseline population-mean graphical model, the average improvement of localization accuracy is 32%; 67% for the most improved landmark.

Multiple discriminant analysis of SPECT data for alzheimer's disease, frontotemporal dementia and asymptomatic controls

Multiple discriminant analysis (MDA) is a generalization of the Fisher discriminant analysis (FDA) and makes it possible to discriminate more than two classes by projecting the data onto a subspace. In this work, it was applied to technetium- 99methylcysteinatedimer (99mTc-ECD) SPECT datasets of 10 Alzheimers disease (AD) patients, 11 frontotemporal dementia (FTD) patients and 11 asymptomatic controls (CTR). Principal component analysis (PCA) was used for dimensionality reduction, followed by projection of the data onto a discrimination plane via MDA. In order to separate the different groups, linear boundaries were calculated by applying FDA to two classes at a time (linear machine). By executing the F-test for different numbers of principal components and examining the corresponding classification accuracy, an optimal discrimination plane based on the first three principal components was determined. In order to further assess the method, another dataset comprising patients with early-onset AD and FTD (beginning or suspected disease) was projected by the same method onto this discrimination plane, resulting in a correct classification for most cases. The successful discrimination of another dataset on the same plane indicates that the model is well suited to account for disease-specific characteristics within the classes, even for patients with early-onset AD and FTD.

Classification of amyloid status using machine learning with histograms of oriented 3D gradients

Brain amyloid burden may be quantitatively assessed from positron emission tomography imaging using standardised uptake value ratios. Using these ratios as an adjunct to visual image assessment has been shown to improve inter-reader reliability, however, the amyloid positivity threshold is dependent on the tracer and specific image regions used to calculate the uptake ratio. To address this problem, we propose a machine learning approach to amyloid status classification, which is independent of tracer and does not require a specific set of regions of interest. Our method extracts feature vectors from amyloid images, which are based on histograms of oriented three-dimensional gradients. We optimised our method on 133 18F-florbetapir brain volumes, and applied it to a separate test set of 131 volumes. Using the same parameter settings, we then applied our method to 209 11C-PiB images and 128 18F-florbetaben images. We compared our method to classification results achieved using two other methods: standardised uptake value ratios and a machine learning method based on voxel intensities. Our method resulted in the largest mean distances between the subjects and the classification boundary, suggesting that it is less likely to make low-confidence classification decisions. Moreover, our method obtained the highest classification accuracy for all three tracers, and consistently achieved above 96% accuracy.

Semiquantitative slab view display for visual evaluation of 123I-FP-CIT SPECT.

ObjectiveDopamine transporter single-photon emission computed tomography (SPECT) with 123I-FP-CIT is used widely in the diagnosis of clinically uncertain parkinsonian syndromes. In terms of the evaluation of FP-CIT SPECT, some practice guidelines state that visual interpretation alone is generally sufficient in clinical patient care, whereas other guidelines consider semiquantitative analysis of striatal dopamine transporter availability mandatory. This discrepancy might be because of a relative lack of widely available display tools for FP-CIT SPECT. In this study, we evaluate a semiquantitative slab view display optimized for visual evaluation of FP-CIT SPECT that might resolve the discrepancy. Patients and methodsThe reconstructed FP-CIT SPECT image was stereotactically normalized and scaled voxel by voxel to the mean uptake in the entire brain without striata. From the resulting distribution volume ratio image, a 12-mm-thick transversal slice (slab) through the striata was displayed with a standard colour table with predefined fixed thresholds on the distribution volume ratio. Visual scoring of the semiquantitative slab view was performed twice by four independent readers in 235 unselected patients. The specific binding ratio in the caudate and putamen was computed by fully automated semiquantitative analysis with predefined standard regions of interest in template space. ResultsIntrarater and inter-rater agreement of binary visual categorization as ‘normal’ or ‘reduced’ was excellent (mean Cohen’s &kgr;=0.88 and 0.83, respectively). The area under the receiver–operator characteristic curve of the specific putamen-binding ratio for differentiation between visually normal and visually reduced (majority read) was 0.96. ConclusionVisual interpretation of FP-CIT SPECT on the basis of the semiquantitative slab view display provides excellent stability within and between readers as well as very high agreement with semiquantitative analysis. This suggests that the slab view display enables reliable visual interpretation of FP-CIT SPECT in clinical routine patient care.

Comparison of methods for classification of Alzheimer's disease, frontotemporal dementia and asymptomatic controls

Single photon emission computed tomography (SPECT) and positron emission tomography (PET) are commonly used for the study of neurodegenerative diseases such as Alzheimers disease (AD) and frontotemporal dementia (FTD). Many methods have been proposed to identify different types of dementia based on PET and SPECT images. However, an extensive evaluation and comparison of different methods for feature extraction and classification of such image data has not been performed yet. In this work, two commonly used feature extraction methods, principal component analysis (PCA) and partial least squares analysis (PLS), were used for dimensionality reduction, and three classification methods comprising multiple discriminant analysis (MDA), elastic-net logistic regression (ENLR) and support-vector machine (SVM) were used for classification of SPECT image data of asymptomatic controls (CTR), AD and FTD participants. Hence, six image classification procedures were evaluated and compared. The results indicate that PCA-based procedures have more robust and reliable performance than PLS-based procedures, and PCA-ENLR has the best estimated predictive accuracy among all three PCA-based procedures.

Quantile-based classification of Alzheimer's disease, frontotemporal dementia and asymptomatic controls from SPECT data

Nuclear imaging techniques, namely single photon emission computed tomography (SPECT) and positron emission tomography (PET), are commonly used for the study of neurodegenerative diseases such as Alzheimers disease (AD) and frontotemporal dementia (FTD). Many methods have been proposed to identify different types of dementia based on SPECT and PET images. In order to cope with the low number of datasets compared to the high number of independent variables (voxels of the dataset), they either perform a dimensionality reduction prior to classification, which implies identical influence of all available datasets, or try to extract the relevant variables for the prediction, which may be affected by statistical fluctuation resulting from mislabeled data or intrinsic noise within data samples In order to overcome these limitations, this paper presents an alternative method for classification of SPECT image data of asymptomatic controls (HC), AD and FTD participants. The proposed method produces a voxel mask that weights or ignores voxels according to their relevance for classification. The algorithm is based on quantiles and is less sensitive to the non-Gaussian statistical distribution of the classes to separate, which is a very desirable in case of dementia classification. Special care is taken to assess the robustness of the proposed approach. The classification accuracy assessed with bootstrap resampling is presented and the robustness against outliers and misdiagnosed training samples is investigated and compared with a PCA-MVA based approach. As a result, the proposed approach shows comparable results with respect to robustness, but better classification accuracy than PCA-based approaches.

A rib-specific multimodal registration algorithm for fused unfolded rib visualization using PET/CT

Respiratory motion affects the alignment of PET and CT volumes from PET/CT examinations in a non-rigid manner. This becomes particularly apparent if reviewing fine anatomical structures such as ribs when assessing bone metastases, which frequently occur in many advanced cancers. To make this routine diagnostic task more efficient, a fused unfolded rib visualization for 18F-NaF PET/CT is presented. It allows to review the whole rib cage in a single image. This advanced visualization is enabled by a novel rib-specific registration algorithm that rigidly optimizes the local alignment of each individual rib in both modalities based on a matched filter response function. More specifically, rib centerlines are automatically extracted from CT and subsequently individually aligned to the corresponding bone-specific PET rib uptake pattern. The proposed method has been validated on 20 PET/CT scans acquired at different clinical sites. It has been demonstrated that the presented rib- specific registration method significantly improves the rib alignment without having to run complex deformable registration algorithms. At the same time, it guarantees that rib lesions are not further deformed, which may otherwise affect quantitative measurements such as SUVs. Considering clinically relevant distance thresholds, the centerline portion with good alignment compared to the ground truth improved from 60:6% to 86:7% after registration while approximately 98% can be still considered as acceptably aligned.