We consider the problem of characterizing shape populations using highly frequent representative shapes. Framing such shapes as statistical modes – shapes that correspond to (significant) local maxima of the underlying pdfs – we develop a frequency-based, nonparametric approach for estimating sample modes. Using an elastic shape metric, we define ϵ-neighborhoods in the shape space and shortlist shapes that are central and have the most neighbors. A critical issue – How to automatically select the threshold ϵ? – is resolved using a combination of ANOVA and empirical mode distribution. The resulting modal set, in turn, helps characterize the shape population and performs better than the traditional cluster means. We demonstrate this framework using amoeba shapes from brightfield microscopy images and highlight its advantages over existing ideas.
more »
« less
Characterizing Cell Shape Distributions Using k-Mode Kernel Mixtures
This paper addresses the problem of characterizing statistical distributions of cellular shape populations using shape samples from microscopy image data. This problem is challenging because of the nonlinearity and high-dimensionality of shape manifolds. The paper develops an efficient, nonparametric approach using ideas from k-modal mixtures and kernel estimators. It uses elastic shape analysis of cell boundaries to estimate statistical modes and clusters given shapes around those modes. (Notably, it uses a combination of modal distributions and ANOVA to determine k automatically.) A population is then characterized as k-modal mixture relative to this estimated clustering and a chosen kernel (e.g., a Gaussian or a flat kernel). One can compare and analyze populations using the Fisher-Rao metric between their estimated distributions. We demonstrate this approach for classifying shapes associated with migrations of entamoeba histolytica under different experimental conditions. This framework remarkably captures salient shape patterns and separates shape data for different experimental settings, even when it is difficult to discern class differences visually.
more »
« less
- Award ID(s):
- 1955154
- PAR ID:
- 10433354
- Date Published:
- Journal Name:
- 26th International Conference on Pattern Recognition (ICPR)
- Page Range / eLocation ID:
- 2517 to 2523
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
The problem of using covariates to predict shapes of objects in a regression setting is important in many fields. A formal statistical approach, termed Geodesic regression model, is commonly used for modeling and analyzing relationships between Euclidean predictors and shape responses. Despite its popularity, this model faces several key challenges, including (i) misalignment of shapes due to pre-processing steps, (ii) difficulties in shape alignment due to imaging heterogeneity, and (iii) lack of spatial correlation in shape structures. This paper proposes a comprehensive geodesic factor regression model that addresses all these challenges. Instead of using shapes as extracted from pre-registered data, it takes a more fundamental approach, incorporating alignment step within the proposed regression model and learns them using both pre-shape and covariate data. Additionally, it specifies spatial correlation structures using low-dimensional representations, including latent factors on the tangent space and isotropic error terms. The proposed framework results in substantial improvements in regression performance, as demonstrated through simulation studies and a real data analysis on Corpus Callosum contour data obtained from the ADNI study.more » « less
-
more » « less
A continuously scanning laser Doppler vibrometer (CSLDV) system is capable of efficient and spatially dense vibration measurements by sweeping its laser spot along a scan path assigned on a structure. This paper proposes a new operational modal analysis (OMA) method based on a data processing method for CSLDV measurements of a structure, called the lifting method, under white-noise excitation and applies a baseline-free method to identify structural damage using estimated mode shapes from the OMA method. The lifting method enables transformation of raw CSLDV measurements into measurements at individual virtual measurement points, as if the latter were made by use of an ordinary scanning laser Doppler vibrometer in a step-wise manner. It is shown that a correlation function with nonnegative time delays between lifted CSLDV measurements at two virtual measurement points on a structure under white-noise excitation and its power spectrum contain modal parameters of the structure, that is, natural frequencies, modal damping ratios, and mode shapes. The modal parameters can be estimated by using a standard OMA algorithm. A major advantage of the proposed OMA method is that curvature mode shapes associated with mode shapes estimated by the method can reflect local anomaly caused by small-sized structural damage, while those estimated by other existing OMA methods that use CSLDV measurements cannot. Numerical and experimental investigations are conducted to study the OMA method and baseline-free structural damage identification method. In the experimental investigation, effects of the scan frequency of a CSLDV system on the two methods were studied. It is shown in both the numerical and experimental investigations that modal parameters can be accurately estimated by the OMA method and structural damage can be successfully identified in neighborhoods with consistently high values of curvature damage indices.
-
more » « less
Abstract The mode shape is one of the important modal parameters that enables to visualize the intrinsic behavior of a structure as well as the quantity of interest by extracting or separating modal response from measurements. In this study, a new output‐only framework is proposed to extract modes using a modal‐based Kalman filter defined in the modal space and identify the mode shape by manipulating the correlation between the separated modes and the measured responses. It is also shown that the proposed framework can be extended to estimate the mode shapes of a non‐classically damped structure in state space when the state variable is constructed from the measured responses and applied to the modal‐based Kalman filter. The mode shape estimation framework proposed in this study was verified by numerical simulations and full‐scale measurements. From the verification examples and their results, it was noted that the proposed modal identification framework is not influenced by the presence of noise, and it can be applied to identify the state‐space mode shapes of non‐classically damped systems as well as systems with very closely distributed modes such as buildings equipped with tuned mass dampers.
-
A modal decomposition is a useful tool that deconstructs the statistical dependence between two random variables by decomposing their joint distribution into orthogonal modes. Historically, modal decompositions have played important roles in statistics and information theory, e.g., in the study of maximal correlation. They are defined using the singular value decompo- sitions of divergence transition matrices (DTMs) and conditional expectation operators corresponding to joint distributions. In this paper, we first characterize the set of all DTMs, and illustrate how the associated conditional expectation operators are the only weak contractions among a class of natural candidates. While modal decompositions have several modern machine learning applications, such as feature extraction from categorical data, the sample complexity of estimating them in such scenarios has not been analyzed. Hence, we also establish some non-asymptotic sample complexity results for the problem of estimating dominant modes of an unknown joint distribution from training data.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/ICPR56361.2022.9956604
