Detection of acute 3,4-methylenedioxymethamphetamine (MDMA) effects across protocols using automated natural language processing

Abstract

The detection of changes in mental states such as those caused by psychoactive drugs relies on clinical assessments that are inherently subjective. Automated speech analysis may represent a novel method to detect objective markers, which could help improve the characterization of these mental states. In this study, we employed computer-extracted speech features from multiple domains (acoustic, semantic, and psycholinguistic) to assess mental states after controlled administration of 3,4-methylenedioxymethamphetamine (MDMA) and intranasal oxytocin. The training/validation set comprised within-participants data from 31 healthy adults who, over four sessions, were administered MDMA (0.75, 1.5 mg/kg), oxytocin (20 IU), and placebo in randomized, double-blind fashion. Participants completed two 5-min speech tasks during peak drug effects. Analyses included group-level comparisons of drug conditions and estimation of classification at the individual level within this dataset and on two independent datasets. Promising classification results were obtained to detect drug conditions, achieving cross-validated accuracies of up to 87% in training/validation and 92% in the independent datasets, suggesting that the detected patterns of speech variability are associated with drug consumption. Specifically, we found that oxytocin seems to be mostly driven by changes in emotion and prosody, which are mainly captured by acoustic features. In contrast, mental states driven by MDMA consumption appear to manifest in multiple domains of speech. Furthermore, we find that the experimental task has an effect on the speech response within these mental states, which can be attributed to presence or absence of an interaction with another individual. These results represent a proof-of-concept application of the potential of speech to provide an objective measurement of mental states elicited during intoxication.

Fig. 1

a Partial correlations between the statistically significant features found in Table 2 identified as a function of drug condition and task (Monologue presented in the top row and Description in the bottom row). b Multidimensional scaling representation of the partial correlations in Fig. 1a. Observe the horizontal axis differentiating the monologue and description task for each drug condition, and the vertical axis differentiating low and high MDMA conditions for each task. Moreover, the dashed line contains exclusively all of the monologue tasks, stressing the consistency of the representation with the experimental conditions.

Fig. 2. Classification accuracy by task, feature type, and binary condition comparison.

The number of features obtained after feature selection is specified at the top of each bar. The symbols at the bottom of the bar indicate with which algorithm the maximum accuracy was achieved: o Linear SVM, * Random Forest, and + Nearest neighbors. The types of features are indicated as follows: A = Acoustic features only; B = Semantic features only; C = Psycholinguistic/syntactic features only; D = Combined features. PBO = placebo; MDMA 0.75 = 3,4-methylenedioxymethamphetamine 0.75 mg/kg; MDMA 1.5 = 3,4-methylenedioxymethamphetamine 1.5 mg/kg; OT = oxytocin 20 international units. Letters underlined in black indicate that at least one of the models achieved classification higher than chance at p < 0.05, underlined in red at p < 0.001.

Fig. 3. Weight representation of combined features found by optimal linear classification models (2 tasks x 4 conditions).

Weights are normalized to represent the relevant contribution of each feature as percentages. Two heatmaps are shown corresponding to both speech tasks analyzed in this study (left: monologue, right: description). Features that contributed less than 10% were not displayed here. First letter in the feature name indicates the type of feature: A = Acoustic, S = semantic, P = Psycholinguistic.

Fig. 4. Classification accuracy by feature type and binary condition comparison in the validation datasets.

PBO = placebo; MDMA 0.75 = 3,4-methylenedioxymethamphetamine 0.75 mg/kg; MDMA 1.5 = 3,4-methylenedioxymethamphetamine 1.5 mg/kg. Letters underlined in black indicate that at least one of the models achieved classification higher than chance at p < 0.05, underlined in red at p < 0.002 (note these are pessimistic significance estimates based on multiple differences between the training and validation datasets).