Natural speech algorithm applied to baseline interview data can predict which patients will respond to psilocybin for treatment-resistant depression

This machine-learning study (n=17) was able to predict the therapeutic effectiveness of psilocybin for treatment-resistant depression using an algorithm applied to natural speech data from the baseline interviews. The results were 85% accurate and 75% precise.

Abstract

“Background: Natural speech analytics has seen some improvements over recent years, and this has opened a window for objective and quantitative diagnosis in psychiatry. Here, we used a machine-learning algorithm applied to natural speech to ask whether language properties measured before psilocybin for treatment-resistant can predict for which patients it will be effective and for which it will not.

Methods: A baseline autobiographical memory interview was conducted and transcribed. Patients with treatment-resistant depression received 2 doses of psilocybin, 10 mg and 25 mg, 7 days apart. Psychological support was provided before, during and after all dosing sessions. Quantitative speech measures were applied to the interview data from 17 patients and 18 untreated age-matched healthy control subjects. A machine-learning algorithm was used to classify between controls and patients and predict treatment response.

Results: Speech analytics and machine learning successfully differentiated depressed patients from healthy controls and identified treatment responders from non-responders with a significant level of 85% of accuracy (75% precision).

Conclusions: Automatic natural language analysis was used to predict effective response to treatment with psilocybin, suggesting that these tools offer a highly cost-effective facility for screening individuals for treatment suitability and sensitivity. Limitations: The sample size was small and replication is required to strengthen inferences on these results.”

Authors: Facundo Carrillo, Mariano Sigman, Diego F. Slezak, Philip Ashton, Lily Fitzgerald, Jack Stroud, David J. Nutt & Robin L. Carhart-Harris

Summary

Natural speech analytics has seen some improvements over recent years.

1. Introduction

Quantitative analyses of natural speech have undergone significant advances in recent years and are being applied in psychiatry. For example, automatic analysis of speech incoherence has been used as a biomarker of schizophrenia and predicts conversion to psychosis in at-risk individuals with 100% accuracy.

In the present study, we sought to build on previous work by testing whether natural speech analytics combined with machine learning could predict clinical responses to psilocybin in patients with treatment-resistant depression.

2. Methods

This trial was an open-label design with two doses of psilocybin given one week apart to patients with TRD. The autobiographical memory test was performed by 17 patients and 18 age and sex matched controls.

The AMT is a structured interview in which participants are asked to provide specific autobiographical memories in response to specific cue words. Two different but balanced versions of the task were completed across the sample.

The sample comprised 17 patients and 18 healthy control subjects. The primary outcome measure was the Quick Inventory of Depressive Symptoms (QIDS-16), and 7 treatment responders and 10 non responders were found.

2.1. Analysis on subject speech

Emotional Analysis was used to quantify the emotional content of transcribed AMT interviews. The average positivity and average negativity of a text were calculated as decimal values between 0 and 1.

2.2. Machine learning

We used a Gaussian Naive Bayes classifier to classify patients versus controls and separately, responders versus non-responders.

3. Results

We asked whether our method can distinguish between controls and patients. We found that patients use significantly fewer positive words in their AMT interview responses, and that a machine learning classifier can identify patients with an accuracy of 82.85%.

We used machine learning to identify responders from non-responders and were able to predict treatment response with an above chance accuracy of 85%.

AVG P was the most sensitive variable for distinguishing patients from controls, and predicting responder versus non-responder. Responders used fewer emotional words at baseline, and fewer positive words especially.

4. Discussion

In a clinical trial of psilocybin for treatment-resistant depression, natural speech analytics combined with machine learning was able to differentiate depressed patients from healthy controls and predict responders versus non-responders with a significant level of precision.