Toshiro Kubota

Salient closed boundary extraction with ratio contour

We present ratio contour, a novel graph-based method for extracting salient closed boundaries from noisy images. This method operates on a set of boundary fragments that are produced by edge detection. Boundary extraction identifies a subset of these fragments and connects them sequentially to form a closed boundary with the largest saliency. We encode the Gestalt laws of proximity and continuity in a novel boundary-saliency measure based on the relative gap length and average curvature when connecting fragments to form a closed boundary. This new measure attempts to remove a possible bias toward short boundaries. We present a polynomial-time algorithm for finding the most-salient closed boundary. We also present supplementary preprocessing steps that facilitate the application of ratio contour to real images. We compare ratio contour to two closely related methods for extracting closed boundaries: Elder and Zuckers method based on the shortest-path algorithm and Williams and Thornbers method based on spectral analysis and a strongly-connected-components algorithm. This comparison involves both theoretic analysis and experimental evaluation on both synthesized data and real images.

Segmentation of pulmonary nodules of various densities with morphological approaches and convexity models

Accurate segmentation of a pulmonary nodule is an important and active area of research in medical image processing. Although many algorithms have been reported in literature for this problem, those that are applicable to various density types have not been available until recently. In this paper, we propose a new algorithm that is applicable to solid, non-solid and part-solid types and solitary, vascularized, and juxtapleural types. First, the algorithm separates lung parenchyma and radiographically denser anatomical structures with coupled competition and diffusion processes. The technique tends to derive a spatially more homogeneous foreground map than an adaptive thresholding based method. Second, it locates the core of a nodule in a manner that is applicable to juxtapleural types using a transformation applied on the Euclidean distance transform of the foreground. Third, it detaches the nodule from attached structures by a region growing on the Euclidean distance map followed by a procedure to delineate the surface of the nodule based on the patterns of the region growing and distance maps. Finally, convex hull of the nodule surface intersected with the foreground constitutes the final segmentation. The performance of the technique is evaluated with two Lung Imaging Database Consortium (LIDC) data sets with 23 and 82 nodules each, and another data set with 820 nodules with manual diameter measurements. The experiments show that the algorithm is highly reliable in segmenting nodules of various types in a computationally efficient manner.

Automatic Hot Spot Detection and Segmentation in Whole Body FDG-PET Images

We present a system for automatic hot spots detection and segmentation in whole body FDG-PET images. The main contribution of our system is threefold. First, it has a novel body-section labeling module based on spatial hidden-Markov models (HMM); this allows different processing policies to be applied in different body sections. Second, the competition diffusion (CD) segmentation algorithm, which takes into account body-section information, converts the binary thresholding results to probabilistic interpretation and detects hot-spot region candidates. Third, a recursive intensity mode-seeking algorithm finds hot spot centers efficiently, and given these centers, a clinically meaningful protocol is proposed to accurately quantify hot spot volumes. Experimental results show that our system works robustly despite the large variations in clinical PET images.

Shape correspondence through landmark sliding

Motivated by improving statistical shape analysis, this paper presents a novel landmark-based method for accurate shape correspondence, where the general goal is to align multiple shape instances by corresponding a set of given landmark points along those shapes. Different from previous methods, we consider both global shape deformation and local geometric features in defining the shape-correspondence cost function to achieve a consistency between the landmark correspondence and the underlying shape correspondence. According to this cost function, we develop a novel landmark-sliding algorithm to achieve optimal landmark-based shape correspondence with preserved shape topology. The proposed method can be applied to correspond various 2D shapes in the forms of single closed curves, single open curves, self-crossing curves, and multiple curves. We also discuss the practical issue of landmark initialization. The proposed method has been tested on various biological shapes arising from medical image analysis and validated in constructing statistical shape models.

Computer aided detection via asymmetric cascade of sparse hyperplane classifiers

This paper describes a novel classification method for computer aided detection (CAD) that identifies structures of interest from medical images. CAD problems are challenging largely due to the following three characteristics. Typical CAD training data sets are large and extremely unbalanced between positive and negative classes. When searching for descriptive features, researchers often deploy a large set of experimental features, which consequently introduces irrelevant and redundant features. Finally, a CAD system has to satisfy stringent real-time requirements.This work is distinguished by three key contributions. The first is a cascade classification approach which is able to tackle all the above difficulties in a unified framework by employing an asymmetric cascade of sparse classifiers each trained to achieve high detection sensitivity and satisfactory false positive rates. The second is the incorporation of feature computational costs in a linear program formulation that allows the feature selection process to take into account different evaluation costs of various features. The third is a boosting algorithm derived from column generation optimization to effectively solve the proposed cascade linear programs.We apply the proposed approach to the problem of detecting lung nodules from helical multi-slice CT images. Our approach demonstrates superior performance in comparison against support vector machines, linear discriminant analysis and cascade AdaBoost. Especially, the resulting detection system is significantly sped up with our approach.

Extensive clonal spread and extreme longevity in saw palmetto, a foundation clonal plant

The lack of effective tools has hampered out ability to assess the size, growth and ages of clonal plants. With Serenoa repens (saw palmetto) as a model, we introduce a novel analytical framework that integrates DNA fingerprinting and mathematical modelling to simulate growth and estimate ages of clonal plants. We also demonstrate the application of such life‐history information of clonal plants to provide insight into management plans. Serenoa is an ecologically important foundation species in many Southeastern United States ecosystems; yet, many land managers consider Serenoa a troublesome invasive plant. Accordingly, management plans have been developed to reduce or eliminate Serenoa with little understanding of its life history. Using Amplified Fragment Length Polymorphisms, we genotyped 263 Serenoa and 134 Sabal etonia (a sympatric non‐clonal palmetto) samples collected from a 20 × 20 m study plot in Florida scrub. Sabal samples were used to assign small field‐unidentifiable palmettos to Serenoa or Sabal and also as a negative control for clone detection. We then mathematically modelled clonal networks to estimate genet ages. Our results suggest that Serenoa predominantly propagate via vegetative sprouts and 10 000‐year‐old genets may be common, while showing no evidence of clone formation by Sabal. The results of this and our previous studies suggest that: (i) Serenoa has been part of scrub associations for thousands of years, (ii) Serenoa invasion are unlikely and (ii) once Serenoa is eliminated from local communities, its restoration will be difficult. Reevaluation of the current management tools and plans is an urgent task.

Estimating diameters of pulmonary nodules with competition-diffusion and robust ellipsoid fit

We propose a new technique to extract a pulmonary nodule from helical thoracic CT scans and estimate its diameter. The technique is based on a novel segmentation, or label-assignment, framework called competition-diffusion (CD), combined with robust ellipsoid fitting (EF). The competition force defined by replicator equations draws one dominant label at each voxel, and the diffusion force encourages spatial coherence in the segmentation map. CD is used to reliably extract foreground structures, and nodule like objects are further separated from attached structures using EF. Using ground-truth measured manually over 1300 nodules taken from more than 240 CT volumes, the performance of the proposed approach is evaluated in comparison with two other techniques: Local Density Maximum algorithm and the original EF. The results show that our approach provides the most accurate size estimates.

Computation of Orientational Filters for Real-time Computer Vision Problems III: Steerable System and VLSI Architecture

Abstract Orientational filters have been used frequently for computer vision problems. Despite their strength in various vision problems, their use has been limited in real-time applications since they are computationally intensive. Part I presented separable approximation as a way to implement a real-time orientational filter operation with a small amount of hardware. Part II presented an efficient computation scheme for 2D multi-resolution decomposition. This paper extends the results of the previous papers to construct a multi-resolution filter system where the orientation of each filter can be adaptively controlled. Such a system is called steerable and is useful in many image analysis/processing applications. This paper also presents a VLSI architecture for various orientational filter systems. The architecture is scalable in terms of the input image size, the filter size, the number of orientational filters, and the approximation order.

From fragments to salient closed boundaries: an in-depth study

This paper conducts an in-depth study on a classical perceptual-organization problem: finding salient closed boundaries from a set of boundary fragments detected in a noisy image. In this problem, a saliency boundary is formed by identifying and connecting a subset of fragments according to the simple Gestalt laws of closure, continuity, and proximity. Our specific interest is focused on the methods that aim to achieve boundary closure, an important global property of perceptual salient boundaries. In this paper, we analyze and compare three such methods that are developed in recent years: (a) Elder and Zuckers method based on the shortest-path algorithm, (b) Williams and Thornbers method combining the spectral-analysis and the strongly-connected-component algorithms, and (c) Wang, Kubota, and Siskinds method based on ratio-contour algorithm. Both theoretic analysis and experimental study show that, with a unified setting of fragment saliency, Wang, Kubota and Siskinds method more appropriately constrains the search space for the closed boundaries, and usually produces better performance than or at least comparable performance as the other two methods. Particularly, Wang, Kubota, and Siksinds method can always guarantee the boundary closure and simplicity, which may not be always hold in the other two methods. We construct and collect a variety of synthesized and real images for this comparison.

Reaction-diffusion systems for hypothesis propagation

Describes a technique for determining a classification map or a hypothesis selection map using a system of reaction-diffusion equations. The reaction term of the equations is a replicator equation and provides a mutually exclusive solution. The diffusion term exploits the local consistency of the solution. The technique is efficient for problems with a small labeling space compared to annealing based relaxation techniques. The paper gives experimental results of the technique for target classification and stereo correspondence.