Integrated Cognitive Assessment (ICA): Speed and Accuracy of Visual Processing as a Proxy to Cognitive Performance Abstract • Over 25 Years of research has not yielded a treatment for Alhzeimer’s disease • This is, in part, due to the lack of effective early detec- tion tools. • Prior research focused on macroscopic changes associ- ated with the disease, or on individual genes or molecular pathways. • Current neuropsychological tests fail to capture the sub- tleties of cognitive deficits at early stages of the disease. • We developed the ICA test that is more sensitive to less sever brain deteriorations. Seyed-Mahdi Khaligh-Razavi*, Sina Habibi, Elham Sadeghi, and Chris Kalafatis Acknolwedgement We thank the researchers and clinicians at Royan Institue who helped with some of the data collection: Seyed Masoud Nabavi, Haniye Marefat, Maryam Sadeghi, and Mahdiye Khanbagi. Results •ICA is correlated with a wide range of the standard-of-care cognitive tests. •ICA shows excellent test-retest reliability (r=0.96*). •ICA does not have a learning bias, and is largely independet of education. •The AI egnine can discrimate between cognitively imapired and non-impaired subjects with AUC of 95%. •As opposed to convential cognitive tests, becaues of the AI engine, the categoriation peforamcne of the test can be further improved with more data. *[email protected] ICA Outline Process Background science: predicting the reaction time of healthy human subjects in a rapid categorization task, using natural image statistics (Mirzaei et al., 2012) STIMULUS 100 ms ISI 20 ms MASK 250 ms ANIMAL ? CTA + The test takes advantage of millions of years of human evolution – the human brain’s strong reaction to animal stimuli Subjects are exposed to a succession of short exposure visual stimuli and asked to react to a simple question - whether or not they saw an animal. The images used vary in their properties, and the speed and accuracy of response to these vary according to the cognitive ability of the subject EDUCATION INDEPENDENT NO LEARNING BIAS LANGUAGE INDEPENDENT SELF ADMINISTERED SENSITIVITY THROUGH AI TIME EFFECTIVE (5 MIN) Conclusion ICA has advantages over the conventional standard-of-care cognitive tests. Because of its efficient administration, shorter duration, and automatic scoring, it allows for remote cognitive assessment and online home monitoring. Furthermore, the ICA test does not suffer from a learning bias, therefore it can be administered frequently to track changes in individuals’ cognitive performance over time. ICA Test Paerns of RT Feature Extracon Summary Features Aenon Accuracy Speed M2 Reaction Time (RT) Accuracy 2427 0 2488 1 1876 0 1160 1 Training and Tetsing the Artificial Intelligence (AI) engine AI Engine ICA Score ICA correlation with standard paper-based cognitive tests [448 Participatns in total] 212 parcipants [ICA Platform: RaspPi] 166 parcipants [ICA Platform: iPad] 166 parcipants 166 parcipants 58 parcipants 58 parcipants r=0.46, p<10 -12 r=0.80, p<10 -7 r=0.66, p<10 -7 r=0.54, p<10 -6 r=0.55, p<10 -4 r=0.60, p<10 -6 SDMT CVLT-II BVMT-R MoCA MoCA ACE-R ICA Score [ICA Platform: iPad] [ICA Platform: iPad] [ICA Platform: iPad] [ICA Platform: iPad]

ICA aai poster 30x40 - Cognetivity Neurosciences · 2019. 3. 16. · Integrated Cognitive Assessment (ICA): Speed and Accuracy of Visual Processing as a Proxy to Cognitive Performance

Download PDF Report

Upload
others
View
3
Download
0

Embed Size (px)

Citation preview

Page 1: ICA aai poster 30x40 - Cognetivity Neurosciences · 2019. 3. 16. · Integrated Cognitive Assessment (ICA): Speed and Accuracy of Visual Processing as a Proxy to Cognitive Performance

Integrated Cognitive Assessment (ICA): Speed and Accuracy of Visual Processing as a Proxy to Cognitive Performance

Abstract

• Over 25 Years of research has not yielded a treatment for Alhzeimer’s disease

• This is, in part, due to the lack of effective early detec-tion tools.

• Prior research focused on macroscopic changes associ-ated with the disease, or on individual genes or molecular pathways.

• Current neuropsychological tests fail to capture the sub-tleties of cognitive deficits at early stages of the disease.

• We developed the ICA test that is more sensitive to less sever brain deteriorations.

Seyed-Mahdi Khaligh-Razavi*, Sina Habibi, Elham Sadeghi, and Chris Kalafatis

AcknolwedgementWe thank the researchers and clinicians at Royan Institue who helped with some of the data collection: Seyed Masoud Nabavi, Haniye Marefat, Maryam Sadeghi, and Mahdiye Khanbagi.

Results•ICA is correlated with a wide range of the standard-of-care cognitive tests.•ICA shows excellent test-retest reliability (r=0.96*).•ICA does not have a learning bias, and is largely independet of education. •The AI egnine can discrimate between cognitively imapired and non-impaired subjects with AUC of 95%. •As opposed to convential cognitive tests, becaues of the AI engine, the categoriation peforamcne of the test can be further improved with more data.

*[email protected]

ICA Outline ProcessBackground science: predicting the reaction time of healthy human subjects in a rapid categorization task, using natural image statistics (Mirzaei et al., 2012)

STIMULUS100 ms

ISI20 ms

MASK250 ms

ANIMAL ?CTA

The test takes advantage of

millions of years of human

evolution – the human brain’s

strong reaction to animal stimuli

Subjects are exposed to a succession of short exposure visual stimuli and asked to react to a simple question - whether or not they saw an animal. The images used vary in their properties, and the speed and accuracy of response to these vary according to the cognitive ability of the subject

EDUCATION INDEPENDENT

NO LEARNING BIAS

LANGUAGE INDEPENDENT

SELFADMINISTERED

SENSITIVITY THROUGH AI

TIME EFFECTIVE (5 MIN)

Conclusion ICA has advantages over the conventional standard-of-care cognitive tests. Because of its efficient administration, shorter duration, and automatic scoring, it allows for remote cognitive assessment and online home monitoring. Furthermore, the ICA test does not suffer from a learning bias, therefore it can be administered frequently to track changes in individuals’ cognitive performance over time.

ICA Test Patterns of RT Feature Extraction Summary Features

AttentionAccuracy

Speed M2

Reaction Time (RT) Accuracy

2427 0

2488 1

1876 0

1160 1

Training and Tetsing the Artificial Intelligence (AI) engine

AI Engine

ICA Score

ICA correlation with standard paper-based cognitive tests [448 Participatns in total]

212 participants[ICA Platform: RaspPi]

166 participants[ICA Platform: iPad]

166 participants 166 participants 58 participants 58 participantsr=0.46, p<10-12r=0.80, p<10-7 r=0.66, p<10-7 r=0.54, p<10-6 r=0.55, p<10-4 r=0.60, p<10-6

SDMT CVLT-II BVMT-R MoCA MoCA ACE-R

ICA

Scor

[ICA Platform: iPad] [ICA Platform: iPad] [ICA Platform: iPad] [ICA Platform: iPad]