Fusion-based spatio-temporal feature detection for robust classification of instantaneous changes in pupil response as a correlate of cognitive response

• US 7,938,785 B2
• Filed: 01/23/2009
• Issued: 05/10/2011
• Est. Priority Date: 12/27/2007
• Status: Active Grant

• Current Assignee: Teledyne Scientific & Imaging, LLC (Teledyne Technologies Incorporated)
• Original Assignee: Teledyne Scientific & Imaging, LLC (Teledyne Technologies Incorporated)
• Inventors: Qian, Ming, Privitera, Claudio, Zachery, Karen N, Aguilar, Mario
• Primary Examiner(s): Hoekstra; Jeffrey G
• Application Number:
  US12/358,495
• Publication Number:
  US 20090171240A1
• Time in Patent Office: 837 Days
• Field of Search: 600/300, 600/544, 600/545, 600/546, 600/558, 600/587, 600/595, 702/127, 702141-158
• US Class Current: 600/558
• CPC Class Codes: A61B 3/112 for measuring diameter of pupils
• CPC Class Codes: A61B 5/16 Devices for psychotechnics using teaching or educational appliances G09B1/00 - G09B7/00; Testing reaction times ; Devices for evaluating the psychological state
• CPC Class Codes: A61B 5/163 by tracking eye movement, gaze, or pupil change
• CPC Class Codes: A61B 5/7267 involving training the classification device
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients

Abstract

A computationally efficient and robust approach for monitoring the instantaneous pupil response as a correlate to significant cognitive response to relevant stimuli derives data samples d(n) of the pupil diameter (area), v(n) of pupil velocity and a(n) of pupil acceleration from the pupillary response and segments the data samples into a sequence of time-shifted windows. A feature extractor extracts a plurality of spatio-temporal pupil features from the data samples from the response and baseline periods in each window. A classifier trained to detect patterns of the extracted spatio-temporal pupil features for relevant stimuli generates an output indicative of the occurrence of absence of a significant cognitive response in the subject to a relevant stimulus.

Claims

1. A method for detecting relevant stimuli using pupillary response as a correlate to cognitive response, comprising:

providing a classifier, said classifier trained to detect patterns of spatio-temporal pupil features and generate an output indicative of the occurrence or absence of a relevant stimulus that evokes both a cognitive response and a pupillary response in a subject;

using a sensor to measure the pupillary response of a subject subjected to stimuli in an environment, wherein the occurrence of a relevant stimulus is initially unknown to the classifier;

deriving data samples d(n) of the pupil diameter from the pupillary response;

segmenting the data samples d(n) into a sequence of time-shifted windows, each window including a response period and a baseline period;

extracting a plurality of spatio-temporal pupil features from the data samples d(n) in said response and baseline periods in each said window; and

for each said window, presenting the extracted pupil features to the classifier to generate an output indicative of the occurrence or absence of a relevant stimulus.

2. The method of claim 1, wherein said plurality of pupil features includes at least one comparative pupil feature extracted from samples d(n) in both the response and baseline periods.

3. The method of claim 2, wherein said at least one comparative pupil feature is a difference between a first diameter, velocity or acceleration feature computed from samples d(n) in the response period and a respective second diameter, velocity or acceleration feature computed from samples d(n) in the baseline period.

4. The method of claim 2, wherein said plurality of pupil features includes at least one absolute pupil feature extracted from samples d(n) in only the response period.

5. The method of claim 4, wherein at least one of the absolute pupil features is a diameter, velocity or acceleration feature.

6. The method of claim 4, wherein at least one of the absolute or pupil features is a velocity or acceleration feature.

7. The method of claim 1, further comprising deriving data samples v(n) of the velocity and a(n) of the acceleration, said pupil features including a diameter, a velocity and an acceleration feature computed from the d(n), v(n) and a(n) samples in the response and baseline periods.

8. The method of claim 1, further comprising synthesizing the one or more temporal outputs indicative of the occurrence of a significant cognitive response to a particular relevant stimulus to reinforce or reject the output and, if reinforced, to resolve the time-stamp of the stimulus.

9. The method of claim 1, wherein the output is provided as a feature to the classifier for the next time-shifted window.

10. The method of claim 1, further comprising measuring light conditions, extracting a light feature indicative of illumination changes and inputting the light feature to the classifier to discount pupillary responses due to illumination changes.

11. The method of claim 1, wherein the output is paired with a sample of interest d(n) at a specified position within the window, said response period being the X leading samples d(n) in front of the specified position and said baseline period being the Y lagging samples behind the specified position.

12. The method of claim 1, wherein the output is paired with a sample of interest d(n) at the leading edge of the window, said response period being the immediately preceding X samples d(n) and said baseline period being the Y samples preceding the response period.

13. The method of claim 1, wherein the classifier is a linear classifier trained to detect patterns of linearly weighted combinations of said extracted spatio-temporal pupil features.

14. The method of claim 1, further comprising:

providing a training set of spatio-temporal pupil features including absolute diameter, velocity and acceleration features extracted from data samples d(n) in only the response period and comparative diameter, velocity and acceleration features from data samples d(n) in both the response and baseline periods;

selecting the plurality of spatio-temporal pupil features from the training set that separate relevant stimuli from non-relevant stimuli; and

training the classifier to detect patterns of the selected plurality of spatio-temporal pupil features.

15. A method for detecting relevant stimuli using pupillary response as a correlate to cognitive response, comprising:

providing a classifier, said classifier trained to detect patterns of pupil features and generate an output indicative of the occurrence or absence of a relevant stimulus that evokes both a cognitive response and a pupillary response in a subject;

using a sensor to measure the pupillary response of a subject subjected to stimuli in an environment, wherein the occurrence of a relevant stimulus is initially unknown to the classifier;

extracting a plurality of pupil features from the measured pupillary response; and

presenting the extracted pupil features to the classifier to generate an output indicative of the occurrence or absence of a relevant stimulus.

16. The method of claim 15, wherein the cognitive response is indicative of conscious intellectual activity.