# US 8,542,898 B2 · Bayesian inference of particle motion and dynamics from single particle tracking and fluorescence correlation spectroscopy > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 8,542,898 B2 - Document ID: US8542898B2 - Status: Granted - Kind: B2 - Application number: 13/328,879 - Filed: 2011-12-16 - Published: 2012-06-21 - Granted: 2013-09-24 - Earliest claimed: 2010-12-16 - Assignees: Massachusetts Institute of Technology - Applicants: Mark Bathe; Jun He; Syuan-Ming Guo; Nilah Monnier - Inventors: Mark Bathe; Jun He; Syuan-Ming Guo; Nilah Monnier - Web page: https://patentgrep.com/patent/US8542898B2 ## CPC - G06T 7/20: Analysis of motion (motion estimation for coding, decoding, compressing or decompressing digital video signals H04N19/43, H04N19/51) - G01N 21/6408: {with measurement of decay time, time resolved fluorescence} - G06T 7/0016: {involving temporal comparison} - G06V 20/695: {Preprocessing, e.g. image segmentation} - G06T 2207/10064: Fluorescence image - G06T 2207/30024: Cell structures in vitro; Tissue sections in vitro - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL - G01N: INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES (measuring or testing apparatus or processes other than immunoassay, involving enzymes or microorganisms C12M, C12Q) - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING ## Abstract Techniques for inferring particle dynamics from certain data include determining multiple models for motion of particles in a biological sample. Each model includes a corresponding set of one or more parameters. Measured data is obtained based on measurements at one or more voxels of an imaging system sensitive to motion of particles in the biological sample; and, determining noise correlation of the measured data. Based at least in part on the noise correlation, a marginal likelihood is determined of the measured data given each model of the multiple models. A relative probability for each model is determined based on the marginal likelihood. Based at least in part on the relative probability for each model, a value is determined for at least one parameter of the set of one or more parameters corresponding to a selected model of the multiple models. ## Claims ### Claim 1 A method comprising: determining a plurality of models for motion of particles in a biological sample, each model including a corresponding set of one or more parameters; obtaining measured data based on measurements at one or more voxels of an imaging system sensitive to motion of particles in the biological sample; determining, at least in part on a first hardware processor, noise correlation of the measured data; determining, at least in part on a second hardware processor, based at least in part on the noise correlation, a marginal likelihood of the measured data for each model of the plurality of models; determining a relative probability for each model of the plurality of models based on the marginal likelihood; and determining, based at least in part on the relative probability for each model, a value for a parameter of the set of one or more parameters corresponding to a selected model of the plurality of models. ### Claim 2 (depends on 1) A method as recited in claim 1, wherein determining noise correlation of the measured data further comprises determining a covariance matrix of the measured data. ### Claim 3 (depends on 2) A method as recited in claim 2, wherein determining the noise correlation further comprises regularizing the covariance matrix of the measured data using a shrinkage estimator. ### Claim 4 (depends on 1) A method as recited in claim 1, wherein determining the marginal likelihood of the measured data for each model further comprises determining, based on the noise correlation, a multivariate Gaussian function that describes probability P (y|β, M) of the measured data y given a set of values β for the set of one or more parameters corresponding to each model M. ### Claim 5 (depends on 4) A method as recited in claim 4, wherein determining the marginal likelihood of the measured data for each model further comprises determining, based on the noise correlation, a posterior probability P (β|y, M) for values of a set of one or more parameters corresponding to each model M given the data y using generalized least squares, wherein the posterior probability indicates a most likely set of values βpost and a covariance matrix Σβ. ### Claim 6 (depends on 5) A method as recited in claim 5, wherein determining the marginal likelihood of the measured data for each model further comprises; determining, based on βpost and Σβ, a prior probability P (β|M) for values of a set of one or more parameters corresponding to each model; and determining the marginal likelihood P(y|M) of the measured data y for each model M by performing Monte Carlo integration, over a range of particular parameter values for the set of one or more parameters, of a product of a probability P (y|β, M) of the measured data for one set of particular parameter values for the set of one or more parameters corresponding to each model and a prior probability P (β|M) of the one set of particular parameter value for each model. ### Claim 7 (depends on 5) A method as recited in claim 5, wherein determining the marginal likelihood of the measured data for each model further comprises; determining that the posterior probability P (β|y, M) for a model M approaches a Gaussian distribution; and determining the marginal likelihood P(y|M) of the measured data y for each model M by evaluating the Laplace approximation based on βpost and Σβ. ### Claim 8 (depends on 6) A method as recited in claim 6, wherein determining the prior probability P (β|M) further determining a constant probability density function normalized to 1.0 in a range centered at βpost and extending to extreme values separated by more than about ten times a standard deviation based on Σβ. ### Claim 9 (depends on 6) A method as recited in claim 6, further comprising determining the prior probability of a parameter of a model by determining equal probability per decade for the parameter normalized to 1.0 in a range centered at βpost and extending to extreme values separated by more than about ten times a standard deviation based on Σβ. ### Claim 10 (depends on 1) A method as recited in claim 1, wherein the selected model is the one most likely model. ### Claim 11 (depends on 1) A method as recited in claim 1, wherein determining the value for the parameter further comprises determining an average of values of the parameter over all models wherein each value is weighted by the relative probability for each model. ### Claim 12 (depends on 1) A method as recited in claim 1, wherein determining the value for the parameter of the set of one or more parameters corresponding to the selected model further comprises selecting the model based on a ratio of the relative probability for any two models of the plurality of models exceeding a threshold of about 10. ### Claim 13 (depends on 1) A method as recited in claim 1, wherein the measured data is fluorescence correlation spectroscopy (FCS) data comprising one or more temporal autocorrelation function (TACF) curves based on measurements of fluorescence intensity time series at one or more voxels of a confocal microscope fluorescence measurement system. ### Claim 14 (depends on 13) A method as recited in claim 13, wherein determining the noise level and the noise correlation of the measured data further comprises determining a covariance matrix for a plurality of TACF curves. ### Claim 15 (depends on 14) A method as recited in claim 14, wherein, obtaining measured data further comprises determining a mean TACF curve from the plurality of TACF curves; and determining the noise level and the noise correlation of the measured data further comprises determining the noise level and the noise correlation for the mean TACF curve based on the covariance matrix for the plurality of TACF curves divided by a number of members in the plurality of TACF curves. ### Claim 16 (depends on 13) A method as recited in claim 13, wherein determining the noise level and the noise correlation of the measured data further comprises determining the noise level and the noise correlation of a single fluorescence intensity time series by determining a covariance matrix for deviations of individual time-lagged intensity products from a mean of time-lagged intensity products over a duration of the single fluorescence intensity time series. ### Claim 17 (depends on 16) A method as recited in claim 16, wherein determining the noise level and the noise correlation of a single fluorescence intensity time series further comprises performing a block transform on the individual time-lagged intensity products to yield independent samples of the individual time-lagged intensity products before determining the covariance matrix. ### Claim 18 (depends on 1) A method as recited in claim 1, wherein the measured data is mean square displacement (MSD) data based on one or more single particle trajectories (SPT) that each indicate voxel positions in the imaging system where an individual particle is detected as a function of time. ### Claim 19 (depends on 18) A method as recited in claim 18, wherein the measured data is a mean MSD curve based on a plurality of SPT from different particles undergoing the same physical process or from sub-trajectories of a single particle trajectory or both. ### Claim 20 (depends on 18) A method as recited in claim 18, wherein determining the noise correlation of the measured data further comprises determining a covariance matrix for the MSD data. ### Claim 21 (depends on 18) A method as recited in claim 18, wherein. obtaining measured data further comprises determining a mean MSD curve from a plurality of MSD curves; and determining the noise correlation of the measured data further comprises determining the noise correlation for the mean MSD curve based on the covariance matrix for the plurality of MSD curves divided by a number of members in the plurality of MSD curves. ### Claim 22 A non-transient computer-readable medium carrying one or more sequences of instructions, wherein execution of the one or more sequences of instructions by one or more processors causes an apparatus to perform at least the following: determine a plurality of models for motion of particles in a biological sample, each model including a corresponding set of one or more parameters; obtain measured data based on measurements at one or more voxels of an imaging system sensitive to motion of particles in the biological sample; determine noise correlation of the measured data; determine, based at least in part on the noise correlation, a marginal likelihood of the measured data for each model of the plurality of models; determine a relative probability for each model of the plurality of models based on the marginal likelihood; and determine, based at least in part on the relative probability for each model, a value for a parameter of the set of one or more parameters corresponding to a selected model of the plurality of models. ### Claim 23 An apparatus comprising: at least one processor; and at least one memory including one or more sequences of instructions, the at least one memory and the one or more sequences of instructions configured to, with the at least one processor, cause the apparatus to perform at least the following, determine a plurality of models for motion of particles in a biological sample, each model including a corresponding set of one or more parameters; obtain measured data based on measurements at one or more voxels of an imaging system sensitive to motion of particles in the biological sample; determine noise correlation of the measured data; determine, based at least in part on the noise correlation, a marginal likelihood of the measured data for each model of the plurality of models; determine a relative probability for each model of the plurality of models based on the marginal likelihood; and determine, based at least in part on the relative probability for each model, a value for a parameter of the set of one or more parameters corresponding to a selected model of the plurality of models.