We’ve heard about a number of biomarkers that can assess alzheimers risk before the onset of cognitive impairment or memory loss. These include such tests as MR scans for hippocampal volume loss, brain glucose utilization, and genetic testing for APOE4 alleles. Another possible biomarker that is even more easily examined than any of the above is simply the person’s speech/linguistic patterns.
Elif Eyigoz has an exceptional multi-disciplinary background with degrees in Philosophy (BA), Cognitive Science (MA), Linguistics (MA), and Computer Science (MS and PhD). After she completed her education, she joined IBM in 2014 as a natural language processing (NLP) algorithms developer at Watson Labs, and later joined IBM research in 2016 as a researcher.
#longevity #wellness #Ketones #lifestylemedicine #younger #ketosis #biohacking #acetone #RobertLufkinMD #rapamycin #metabolic #metabolism #alzheimers #elifeyigoz #linguistic
***Rapamycin Longevity Registry***
https://www.rapamycin.io/
*** CONNECT WITH ROBERT LUFKIN MD ON SOCIAL MEDIA ***
Web: https://robertlufkinmd.com/
Twitter:https://twitter.com/robertlufkinmd
Youtube: https://www.youtube.com/RobertLufkinMD
*** MEMBERSHIP ***
https://robert-lufkin.mykajabi.com/membership
*** SPONSORSHIPS & BRANDS ***
We do work with sponsors and brands. If you are interested in working with us for your health industry product or service, please contact us at: https://robertlufkinmd.com/contact
NOTE: This is not intended to be a substitute for professional medical advice, diagnosis, or treatment. Always seek the advice of your physician or other qualified health provider with any questions you may have. Never disregard professional medical advice or delay in seeking it because of something you have seen here.
Robert Lufkin MD may at any time and at its sole discretion change or replace the information available on this channel. To the extent permitted by mandatory law, Robert Lufkin MD shall not be liable for any direct, incidental, consequential, indirect or punitive damages arising out of access to or use of any content available on this channel, including viruses, regardless of the accuracy or completeness of any such content.
Disclaimer: We are ambassadors or affiliates for many of the brands we reference on the channel.
Robert Lufkin 0:00
Welcome back to the health longevity Secret show and I’m your host, Dr. Robert Lufkin. We’ve heard about a number of biomarkers that can assess Alzheimer’s disease risk before the onset of cognitive impairment or memory loss. These include such test says Magnetic Resonance scans for hippocampal volume loss, or brain glucose utilization, as well as genetic testing for a po e for alleles. Another possible biomarker that is even more easily examined than any of the above is simply the person’s speech, linguistic patterns, even if it goes as an exceptional multidisciplinary background with degrees in philosophy, bachelor’s degree, cognitive science, Master’s degree linguistics, master’s degree, computer science, a master’s degree and a PhD. After she completed her education, she joined IBM in 2014. And now works at Watson labs as a researcher. Before we start the episode, if you like what you hear, please consider supporting the work we do as well as joining us on your personal health longevity journey. You can do both by becoming a member of our community. The benefits include a private messaging area, live QA sessions, weekly premier videos, product discounts, free giveaways, and much more. You can join for as little as $1 per month and the first month is free. See the link in the show notes for more information. And now, please enjoy this presentation by Dr. Elif II agos.
Elif Eyigoz 1:52
What is very novel about our study is that this specific datasets made it possible for us to apply these mature techniques to this data set. And for us to discover these biomarkers for prediction of future onset of Alzheimer’s disease in speech. So if there will be an automated system and app or some other way of collecting data points on how they feel that way, and how are they their symptoms that way. And if their doctors could see these automatically, there will be a system to analyze these data points collected every day, the system can warn the patient saying hey, your situation is getting worse. Don’t be too much to go see your doctor go see your doctor to speech.
A key priority in Alzheimer’s Disease Research is the identification of early intervention strategies that will decrease the risk or delay the onset or slow the progression of Alzheimer’s disease. While many variables have been associated with risk of Alzheimer’s disease, there is still a great need for the development of cheap, reliable biomarkers of preclinical Alzheimer’s disease. aging related cognitive decline manifests itself in almost all aspects of language comprehension and production. The aim of this study is to test to what extent linguistic performance at a single time point can be utilized as a prognostic marker of future diagnosis of Alzheimer’s disease in cognitively normal subjects.
We use data from the Framingham Heart Study a large cohort longitudinal study spanning several decades. As a part of Framingham Heart Study, qualifying participants were administered a neuro psychological test battery and successive visits, which included the cookie fath picture description task from the Boston aphasia diagnostic examination. We applied computational techniques to extract linguistic variables from written responses to the cookie fast picture description task and compare their prognostic value with that of more traditional clinical variables that could easily be obtained in the screening period of a clinical trial including neuro psychological test scores, demographic and genetic information and medical history. Using the variables obtained when the participants were assessed to be cognitively normal, we develop models to predict whether or not a particular participant will develop mild cognitive impairment due to Alzheimer’s disease honor before 85 years old. We compare the predictive ability of linguistic variables with that of more traditional variables associated with high risk for Alzheimer’s disease.
The Framingham Heart Study is a well documented community based cohort study initiated in 1948 with the purpose of longitudinal math mirroring of participants health participants undergone neuro psychological examinations every five or six years since 1999. Annual neurologic and neuro psychological examinations were performed when cognitive decline was reported by a family member of the participant upon referral by a physician or by the investigators of the Framingham Heart Study or after review of the participants medical records.
The test battery included the cookie fast picture description task from the Boston aphasia diagnostic examination, in which participants were asked to write down the description of the cookie theft picture. Picture description tasks are commonly used to assess discourse in subjects with disorders such as aphasia and dementia. Given a sensitivity to cognitive impairments cookie theft picture description task has become the most frequently used picture description task in clinical settings.
The neuro psychological test battery resulted in a dementia rating which represented the impression of the examiner who administered the test battery. Therefore, the dementia rating obtained from administration of a neuro psychological test battery was not a diagnosis. In addition, a dementia review panel of at least one neurologist and one neuro psychologist reviewed possible cognitive decline and dementia cases among the participants of the study. While the dementia ratings obtained from the neuro psychological testing were not diagnosis, the dementia review panel did determine the diagnosis dates from mild, moderate and severe Alzheimer’s disease as well as the date of cognitive impairment onset for the participants that reviewed diagnosis of dementia was based on criteria from DSM four and diagnosis of Alzheimer’s disease was based on criteria from ni N CDs, a D RDA.
To fit predictive models of future diagnosis of Alzheimer’s disease, we had to determine which participants to label as cases in which to label is controls. In this study, the onset of Alzheimer’s disease was defined as the onset of mild cognitive impairment and a participant who later received a diagnosis of Alzheimer’s disease. mild cognitive impairment is a heterogeneous condition. However, for mild cognitive impairment patients who eventually convert to Alzheimer’s disease, mild cognitive impairment is considered by many to represent early stage Alzheimer’s disease. Alzheimer’s disease patients who develop mild cognitive impairment honor before age 85 were defined as cases. We defined the normal aging group as the participants who were recorded to be dementia free honor after age 85. Alzheimer’s disease patients whose onset of mild cognitive impairment was after 85 years old were defined as the very late onset Alzheimer’s disease group.
The control group was defined as the combination of the normal aging group and the very late onset Alzheimer’s disease patients. The control group is shown as the blue box in the figure, and the cases are shown as the red box in the figure. According to this definition, all cases have already developed cognitive impairment due to Alzheimer’s disease at 85. And none of the controls have developed cognitive impairment due to Alzheimer’s disease at 85. As depicted in the figure. We identified a clinically defined test set by using the dementia reviews to label cases and controls. We selected one picture description task from each case and matched it to a picture description tasks collected in a control participant of approximately the same age, gender and level of education. In the test set, we included only samples collected prior to any cognitive impairment onset when the subjects were cognitively normal. For the test cases, the meantime to diagnosis with mild Alzheimer’s disease from cognitive normality was seven point 59 years was standard deviation of four point 91. And the meantime to cognitive impairment onset from cognitive normality was three point 93 years was standard deviation of three point 69
We automatically computed psycholinguistic variables using natural language processing techniques. These psycholinguistic variables have been identified in literature as discriminatory variables for already demented patients. syntactic complexity was assessed through analysis of parse trees. semantic content was assessed through analysis of participants mentioned of information content units. propositional idea density analysis was used to quantify syntactic and semantic complexity. misspellings use of punctuation and upper casing were analyzed to assess writing performance and style. language modeling analyses were performed to model the distributions of words sequences. For paucity lexical richness, and repetitiveness was assessed by using metrics such as number of words, number of unique words and frequencies of repetitions. We compared the predictive performance of the linguistic variables with more traditional clinical variables that could easily be obtained in the screening period of a clinical trial. The non linguistic variables were age, gender, education, number of Apo E alleles, to binary indicator variables capturing evidence of hypertension or diabetes and 32 variables resulting from the neuro psychological tests.
The Framingham Heart Study data available to us included a dementia review for only 39% of participants. As a result, only 80 of the participants qualified for inclusion in the test set. This left a very large number of participants on used. In order to obtain a larger data set, we resorted to utilizing a semi supervised learning approach. While most of the participants did not have dementia review data Lang for definitive labeling of cases, a dementia rating was available for the majority of the administration of the neuro psychological test battery. Dementia ratings were not clinical diagnosis. However, we created a training set using the dementia ratings, albeit with noisy labels. On the contrary, our test set included only clinically defined labels from dementia reviews based on clinical diagnosis. In the semi supervised learning terminology, our test dataset was ground truth label while our training dataset was weakly labeled. The semi supervised learning approach allowed us to include 703 samples from 270 participants in our study, where a data set consisting of a single sample from 80 participants was held out for testing. Further, we upsampled the weakly labeled training data to balance the number of samples in each class, allowing us to use more than 1000 samples for training. Before training predictive models, variable selection was performed strictly on the training data by using an unvarying test between the preclinical Alzheimer’s disease cases and the control groups for each variable and eliminating variables that were not statistically significant. We train models on the weekly labeled training data and then validated the models on the ground truth labeled test data. We experimented with logistic regression SVM and naive Bayes classifiers.
To assess whether linguistic variables associated with the time to diagnosis with mild Alzheimer’s disease, we use Cox Proportional Hazards models. Date of mild Alzheimer’s disease diagnosis was obtained from the dementia review. The participants who were recorded as dementia free in their dementia review were censored. Models for each single linguistic variable included as additional covariates of age, gender and education. For the Cox Proportional Hazards analysis, we use 143 participants of which 28 were censored with a total of 1159 person years were average was 8.1 years per person.
To identify possible longitudinal trends present in our multi dimensional assessment of cognitive status, we implemented a non negative matrix factorization analysis. For this analysis, we use the cases in the normal aging participants who have a record of cognitive impairment onset and their dementia review The time intervals between successive visits per participant varied significantly. In order to align the dataset around the date of cognitive impairment onset, we created synthetic samples by linear interpolation with a frequency of six months. This upsampling method allowed us to have a data point for each six month interval around the date of cognitive impairment onset for each participant. We then use non negative matrix factorization on the upsample dataset. For this analysis, we used all neuro psychological variables and linguistic variables that were statistically significant either on the test set on the training set or with the Cox Proportional Hazards model analysis.
We obtained a AUC of zero point 70 for an accuracy of zero point 70. Using linguistic variables alone. We observed a 10 point increase in accuracy obtained by adding linguistic variables to the non linguistic variables as non linguistic variables alone obtained accuracy of zero point 59. And the combination of linguistic and non linguistic variables obtained accuracy of zero point 69. The Z score indicated that AUC of zero point 74 corresponds to a four point 26 fold increase in predictability over chance. The results of the Cox Proportional Hazards analysis showed that using the preferentially generic terms instead of more specific words was associated with higher risk of developing Alzheimer’s disease. For example, referring to the boy in the picture as a son instead of a boy is more referentially specific. Similarly, referring to the girl in the picture as a daughter instead of a girl, and referring to the woman in the picture as a mother, instead of a woman is more referentially specific.
The figure shows the result of the non negative matrix factorization analysis. The black arrow is pointing to the date of cognitive impairment onset on the horizontal axis. The participants visits were aligned at their date of cognitive impairment onset for this analysis. The data was up sampled to have a data point for each six month interval before and after the date of cognitive impairment onset for each participant. Accordingly, the x axis in the figure shows each six month interval and Y axis shows the loading of each six month interval on the first component obtained by non negative matrix factorization. The blue line shows the controls progression and time. The red line shows the cases progression and time with a steeper decline for the cases, especially after the cognitive impairment onset. Similar comparative disease progression models were suggested in the literature for controls versus Alzheimer’s disease patients.
The linguistic variables that we identified as most relevant for predicting future onset of Alzheimer’s disease were prominently a graph via telegraphic speech and repetitiveness. These have been consistently identified in the literature as associated with cognitive decline and dementia. Another linguistic element that has been associated with Alzheimer’s disease was referential specificity. referential specificity was identified as having a strong weight in the Cox Proportional Hazards analysis. It has been hypothesized that using preferentially specific terms, as opposed to more general terms requires making inference therefore, it is more demanding on the cognitive resources. For example, referring the woman in the cookie half picture as a mother requires making an inference about the relationships between the subjects in the picture. This result is supported by a large number of studies showing that semantic impairments are among the earliest linguistic markers of dementia, verbosity and lexical richness metrics, which stand out as strong markers of cognitive impairment in already demented patients were not among the strong predictors of future diagnosis of Alzheimer’s disease in cognitively normal individuals in our study.
Our study differs from similar studies in various ways for First, our prediction is based on data collected while the participants were cognitively healthy. Second, we focus exclusively on variables readily attainable as part of the screening phase of an early intervention trial and assess predictive performance using only linguistic metrics derived from a single administration of a picture description task. Simple, naturalistic and inexpensive speech probes, as our results suggest, can provide an assistive tool for the early detection and progression monitoring of Alzheimer’s disease, particularly given that such probes can be easily adapted to remote digital platforms with low patient burden.
Unknown Speaker 20:43
No, this is not intended to be a substitute for professional medical advice, diagnosis or treatment. Always seek the advice of your physician or other qualified health provider with any questions you may have. Never disregard professional medical advice or delay in seeking of it because of something you’ve seen here. If you find this to be a value of you, please hit that like button and subscribe support the work we do on this channel. Also, we take your suggestions and advice very seriously. Please let us know what you’d like to see on this channel. Thanks for watching and hope to see you next time.
.