Yrjö Häme

Adaptive Quantification and Longitudinal Analysis of Pulmonary Emphysema with a Hidden Markov Measure Field Model

The extent of pulmonary emphysema is commonly estimated from CT scans by computing the proportional area of voxels below a predefined attenuation threshold. However, the reliability of this approach is limited by several factors that affect the CT intensity distributions in the lung. This work presents a novel method for emphysema quantification, based on parametric modeling of intensity distributions and a hidden Markov measure field model to segment emphysematous regions. The framework adapts to the characteristics of an image to ensure a robust quantification of emphysema under varying CT imaging protocols, and differences in parenchymal intensity distributions due to factors such as inspiration level. Compared to standard approaches, the presented model involves a larger number of parameters, most of which can be estimated from data, to handle the variability encountered in lung CT scans. The method was applied on a longitudinal data set with 87 subjects and a total of 365 scans acquired with varying imaging protocols. The resulting emphysema estimates had very high intra-subject correlation values. By reducing sensitivity to changes in imaging protocol, the method provides a more robust estimate than standard approaches. The generated emphysema delineations promise advantages for regional analysis of emphysema extent and progression.

Robust quantification of pulmonary emphysema with a Hidden Markov Measure Field model

Determining the extent of pulmonary emphysema with quantitative computed tomography commonly relies on fixed intensity threshold values. However, the reliability of such measures is limited due to variability in parenchymal intensities and noise levels in CT images. In this work, we present a novel method for emphysema quantification, based on a lung tissue segmentation with a Hidden Markov Measure Field model. By adapting to the intensity distribution present in the input image, the method provides a more robust emphysema index than standard densitometric approaches. The focus of this study is to show robustness between imaging protocols, enabling the comparison of emphysema measures between studies. The method can have a significant impact in longitudinal analysis and prediction of emphysema. In addition, the method shows promise in delineating emphysematous regions, potentially facilitating subtyping of the disease.

Sparse sampling and unsupervised learning of lung texture patterns in pulmonary emphysema: MESA COPD study

Pulmonary emphysema is defined morphologically by enlargement of alveolar airspaces and manifests as textural differences on thoracic computed tomography (CT). This work presents an unsupervised approach to extract the most dominant local lung texture patterns on CT scans. Since the method does not use manually annotated labels restricted to predefined emphysema subtypes, it can be used for discovery of novel image-based phenotypes with greater efficiency and reliability. This study demonstrates the applicability of the learned patterns for content-based image retrieval.

High-resolution contrast enhanced multi-phase hepatic Computed Tomography data fromaporcine Radio-Frequency Ablation study

Data below 1 mm voxel size is getting more and more common in the clinical practice but it is still hard to obtain a consistent collection of such datasets for medical image processing research. With this paper we provide a large collection of Contrast Enhanced (CE) Computed Tomography (CT) data from porcine animal experiments and describe their acquisition procedure and peculiarities. We have acquired three CE-CT phases at the highest available scanner resolution of 57 porcine livers during induced respiratory arrest. These phases capture contrast enhanced hepatic arteries, portal venous veins and hepatic veins. Therefore, we provide scan data that allows for a highly accurate reconstruction of hepatic vessel trees. Several datasets have been acquired during Radio-Frequency Ablation (RFA) experiments. Hence, many datasets show also artificially induced hepatic lesions, which can be used for the evaluation of structure detection methods.

Level set-based tracking of the endocardium without a shape prior from 3D ultrasound images

A level set-based method for tracking the endocardium from 3D ultrasound images is presented. The tracking process is initialized by a manual delineation and performed automatically. The method does not use a shape prior model, removing the requirement of acquiring large sets of training data, and providing adaptivity for abnormal cases. At each image frame, the edge strength is estimated and locations with strong edges are used to compute the region-based surface deformation, which is then propagated to the rest of the image volume. The approach is evaluated with 10 sequences of canine subjects including 5 ischemic image sequences. The results show that the method performs with error values below the interobserver variability of manual segmentation, and is able to robustly follow the cardiac wall movement.

Analysis and classification of optical tomographic images of rheumatoid fingers with ANOVA and discriminate analysis

We present a study on the effectiveness of computer-aided diagnosis (CAD) of rheumatoid arthritis (RA) from frequency-domain diffuse optical tomographic (FDOT) images. FDOT is used to obtain the distribution of tissue optical properties. Subsequently, the non-parametric Kruskal-Wallis ANOVA test is employed to verify statistically significant differences between the optical parameters of patients affected by RA and healthy volunteers. Furthermore, quadratic discriminate analysis (QDA) of the absorption (μa) and scattering (μa or μs) distributions is used to classify subjects as affected or not affected by RA. We evaluate the classification efficiency by determining the sensitivity (Se), specificity (Sp), and the Youden index (Y). We find that combining features extracted from μa and μa or μs images allows for more accurate classification than when μa or μa or μs features are considered individually on their own. Combining μa and μa or μs features yields values of up to Y = 0.75 (Se = 0.84 and Sp = 0.91). The best results when μa or μs features are considered individually are Y = 0.65 (Se = 0.85 and Sp = 0.80) and Y = 0.70 (Se = 0.80 and Sp = 0.90), respectively.