About

Ana Abrantes is a Professor of Psychiatry and Human Behavior at Brown University. Her research focuses on developing and testing novel interventions to reduce relapse risk among individuals with alcohol and other drug use disorders, including nicotine dependence. Her current projects involve the use of text messaging, smartphone applications, neurostimulation techniques such as tDCS, and EMA/EMI approaches. Additionally, Dr. Abrantes conducts research on physical activity promotion for individuals with substance use and mental health disorders, testing interventions supported by devices like Fitbit, physical activity apps, and peer-facilitated approaches. Her work aims to improve treatment outcomes and health behaviors in these populations through innovative, technology-based strategies.

Research topics

• Medicine
• Psychology
• Psychiatry
• Clinical psychology
• Developmental psychology
• Physical therapy
• Social psychology

Selected publications

• Smartphones as Breathalyzers: AI-Based Detection of Alcohol Impairment from Smartphone Motion Sensors

  Journal of Media Literacy Education · 2026-03-27

  articleOpen access

  Alcohol impairment significantly degrades motor coordination, balance, and postural control, resulting in measurable alterations in human gait dynamics. These impairments contribute substantially to roadway accidents and injury-related fatalities worldwide with one person dying every 34 minutes in the US each year. Conventional detection approaches, such as breathalyzer tests, require active participation and are typically administered only after impairment is suspected. In contrast, modern smartphones are equipped with inertial measurement sensors capable of continuously capturing motion signals during routine activities. These sensors provide an opportunity to analyze gait patterns and investigate whether behavioral indicators of intoxication can be detected directly from smartphone motion data. This research investigates the feasibility of detecting alcohol impairment using machine learning models trained on motion signals collected from smartphone sensors. Specifically, the study analyzes multivariate time-series data recorded from tri-axial accelerometers and gyroscopes embedded in smartphones during controlled walking tasks. These sensors capture linear acceleration and angular velocity signals along three spatial axes, producing high-frequency measurements that characterize the dynamics of human locomotion. The raw sensor signals are segmented into fixed-length temporal windows representing short intervals of gait activity. Each segment encodes temporal patterns associated with stride regularity, stability, and movement variability that may be affected by alcohol-induced motor impairment. The primary objective of this work is to develop computational models capable of identifying intoxication-related signatures in these motion signals. Time-series representations derived from accelerometer and gyroscope measurements are analyzed to extract discriminative features that capture subtle variations in gait behavior. Machine learning models are then trained to differentiate between sober and alcohol-influenced walking patterns using these representations. Particular attention is given to addressing challenges common to wearable and mobile sensing applications, including sensor noise, inter-subject variability in gait patterns, and limited availability of labeled intoxication data. To improve model robustness and generalization, ongoing work explores representation learning strategies designed to leverage large volumes of unlabeled motion data collected during natural walking activities. These approaches aim to learn latent embeddings that capture fundamental characteristics of human gait dynamics while remaining sensitive to behavioral perturbations associated with intoxication. Such representations may enable improved performance when labeled training data is limited, while also facilitating transfer across subjects and sensing environments.

  Publisher

• Novel steroidal oximes as antiproliferative agents: Design, synthesis and biological activity evaluation

  Bioorganic Chemistry · 2025-01-30 · 2 citations

  articleOpen access

  • A series of steroidal oximes were synthesised and evaluated as anticancer agents. • OX1 and EP2OX induced cell death by apoptosis or necrosis. • EP2OX causes DNA damage by inducing DSBs triggered by ROS production. • EP2OX was non-haemolytic. • EP2OX may be an excellent candidate for safer and more effective therapies. Oximes have been the subject of extensive research given their interesting anticancer activity. Steroids are also important scaffolds in drug discovery, not only due to their ability to penetrate cell membranes and bind to the nuclear and membrane receptors but also due to their suitability for structural modifications, allowing their use as cytotoxic and cytostatic anticancer agents. Combining the oxime group with the steroidal skeleton can be a suitable strategy to create novel anticancer agents. In this study, we designed and synthesised several novel steroidal oximes ( OX1 , OX2 , OX3 , OX3.1 , OX4 , EP2OX , FormOX and ExeOX ) and evaluated their anticancer activity in three of the most incident and deadliest types of cancer, prostate (PC3), lung (H1299) and triple-negative breast (HCC1806) cancers. Selectivity using a normal human cell line, MRC-5, and hemocompatibility were also assessed. EP2OX was the most active compound in the studied cancer cell lines (IC 50 values ranging from 1.13 to 3.70 µM) followed by OX1 (IC 50 values ranging from 18.69 to 29.95 µM). Further studies with EP2OX and OX1 showed that the first induced DNA damage by double-strand breaks triggered by ROS production, leading to apoptosis/necrosis (depending on the concentration), while the second induced cell death by apoptosis regardless of the concentration. Moreover, both compounds showed some selectivity towards cancer cells and proved to be non-haemolytic. Our results reinforce the importance of steroidal oximes in the oncology field, namely our novel compound EP2OX which might be the starting point for a potential drug candidate for treating these types of cancer.

  PublisherDOI

• Sex differences in the relationship between quit attempt and smoking cessation self-efficacy in adults with depressive symptoms

  Addictive Behaviors · 2025-03-12

  articleOpen accessSenior authorCorresponding
  PublisherOA PDFDOI

• A Confirmatory Factory Analysis of the Exercise Sensitivity Questionnaire (ESQ)

  Journal of Cardiopulmonary Rehabilitation and Prevention · 2025-02-27 · 1 citations

  articleOpen access

  PURPOSE: The Exercise Sensitivity Questionnaire (ESQ) is a self-report measure used to assess the extent to which different physical sensations of exercise elicit anxiety (ie, exercise sensitivity). The ESQ was developed for individuals with cardiovascular conditions and initially validated in a non-clinical sample. This study evaluates the factor structure and measurement invariance in a clinical sample of adults with various cardiovascular conditions. METHODS: This was a cross-sectional study with retrospective chart review. Patients (N = 265; 73% male, mean age 67.8 ± 10.5 years) were attending an orientation for outpatient medically supervised exercise-based cardiac rehabilitation. The factor structure was examined using Confirmatory Factor Analysis, and tests of measurement invariance were evaluated by sex and advanced age (<65 years, >65 years). Internal consistency, descriptive characteristics, and correlates of ESQ scores and its factors were evaluated. Concurrent validity was evaluated in a subset of patients (N = 57) with elevated exercise sensitivity. RESULTS: The Confirmatory Factor Analysis supported a 2-factor model, which was invariant, but not a 1-factor model, and reflected anxiety about (1) cardiopulmonary and (2) pain/weakness exercise sensations. Internal consistency of ESQ items was high. ESQ scores were associated with higher body mass index and shorter 6-Minute Walk Test distance, particularly the pain/weakness factor. ESQ scores evidenced preliminary concurrent validity with anxiety sensitivity and general anxiety but discriminant validity with depressive symptoms. CONCLUSIONS: There is support for the validity and reliability of ESQ scores as a 2-dimensional index of exercise sensitivity. The ESQ taps a psychological phenotype with relevance to exercise tolerance, and potentially cardiac rehabilitation participation, that warrants continued investigation.

  PublisherDOI

• Virtual Group-based Walking Intervention for Adults with Schizophrenia-Spectrum Disorders: Psychosocial Outcome Results from a Pilot Randomized Controlled Trial

  Community Mental Health Journal · 2025-07-11

  articleOpen access
  PublisherOA PDFDOI

• Resting State Functional Connectivity of the Default Mode Network During Opioid Use and Cessation in Treatment‐Seeking Persons

  European Journal of Neuroscience · 2025-01-01 · 3 citations

  articleOpen access

  Resting-state functional connectivity analyses have been used to examine synchrony in neural networks in substance use disorders (SUDs), with the default mode network (DMN) one of the most studied. Prior research has generally found less DMN synchrony during use and greater synchrony during cessation, although little research has utilized this method with opioid use. This study examined resting brain activity in treatment-seeking persons who use opioids at two points-when using opioids and when opioid-free-to determine whether the DMN exhibits different levels of connectivity during opioid use and cessation and whether differences in connectivity predict subsequent relapse. The sample included 11 participants who met DSM-5 criteria for opioid use disorder and initiated buprenorphine treatment following fMRI scans that were approximately 3 days apart. Results showed greater functional connectivity in the DMN and the rIFG of the salience network (SN) when participants were abstaining than when actively using opioids. These changes in connectivity predicted 76.2% of the variance in withdrawal symptom severity, with the DMN nodes accounting for an additional 30.9%. Findings warrant further longitudinal exploration of the role of DMN connectivity and its interactions with other networks in relation to abstinence and withdrawal status and examination of its utility as a prognostic marker of cessation or relapse.

  PublisherOA PDFDOI

• DUI Detection From Gait Using a Multichannel 1DCNN-Attention-BiLSTM Framework

  IEEE Access · 2025-01-01

  articleOpen access

  Alcohol intoxication increases Blood Alcohol Content (BAC) and impairs cognitive, motor, and psychomotor functions, and contributed to over 30% of motor vehicle fatalities in 2017. Traditional methods for detecting Driving Under Influence (DUI), such as breathalyzers and blood tests, are invasive, require external hardware, and are unsuitable for continuous monitoring. Passive approaches such as gait analysis from smartphone sensor data offer a non-invasive solution for detecting impairment and enhancing road safety. Prior work has explored traditional machine learning (ML) and some Deep Learning (DL) approaches but have limitations such as analyzing on handcrafted features and facing challenges such as class imbalance and gait variability. This paper proposes a novel DL framework for passive alcohol intoxication detection from smartphone accelerometer data. A subject-level pre-processing pipeline was employed to address inter- and intra-subject variability, including stratified splitting, low-pass filtering, sliding window segmentation, and random oversampling to mitigate class imbalance. We propose the Multichannel Hybrid 1D-CNN-Attention-BiLSTM (MC-Hybrid) model, which extracts short-term features via parallel 1D-CNNs, uses a self-attention mechanism to increase weights on predictive patterns, and utilizes a bidirectional LSTM to model temporal dependencies. Rigorous evaluation includes comparison to a comprehensive set of ML and DL baselines, investigation of multiple sensor data window sizes, attention types, and ablation studies. MC-Hybrid achieved 93% accuracy and an F1-score of 0.8653, outperforming the state-of-the-art by 9.5% and all baselines by 9.0%. Self-attention resulted in a 2% performance gain over other attention mechanisms, demonstrating its effectiveness in DUI detection. The proposed method could be a practical, non-invasive approach to detect alcohol impairment.

  PublisherDOI

• Evaluation of the efficacy of cardiac rehabilitation in older adults. A prospective cohort study

  REC CardioClinics · 2025-10-28 · 1 citations

  article1st authorCorresponding
  PublisherDOI

• Discriminating Between Marijuana and Alcohol Gait Impairments Using Tile CNN With TICA Pooling

  IEEE Open Journal of Engineering in Medicine and Biology · 2025-01-01

  articleOpen access

  <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Goal:</i> To investigate whether machine learning analyses of smartphone sensor data can discriminate whether a subject consumed alcohol or marijuana from their gait. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Methods:</i> Using first-of-a-kind impaired gait datasets, we propose <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MariaGait</i>, a novel deep learning approach to distinguish between marijuana and alcohol impairment. Subjects' time-series smartphone accelerometer and gyroscope sensor gait data are first encoded into Gramian Angular Field (GAF) images that are then classified using a tiled Convolutional Neural Network (CNN) with TICA pooling. To mitigate the insufficiency of positively labeled alcohol and marijuana instances, the tiled CNN was pre-trained on sober gait samples that were more abundant. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Results:</i> <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MariaGait</i> achieved an accuracy of 94.61%, F1 score of 88.61%, and 94.33% ROC AUC score in classifying whether the subject consumed alcohol or marijuana, outperforming baseline models including Multi-Layer Perceptron (MLP), Long Short Term Memory (LSTM), Multi-head CNN and Multi-head LSTM, Random Forest and Support Vector Machines (SVM)). <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Conclusions:</i> Our results demonstrate that <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MariaGait</i> could be a practical, non-invasive approach to determine which substance a subject is impaired by from their gait.

  PublisherDOI

• The Influence of Perimenstrual Daily Ovarian Hormones on Anxiety and Cigarette Craving

  Nicotine & Tobacco Research · 2025-02-27 · 2 citations

  articleOpen accessSenior author

  INTRODUCTION: Estradiol (E2) and progesterone (P) and their fluctuation during the female menstrual cycle have been independently linked to both nicotine reinforcement and anxiety. The fluctuation and withdrawal of E2 and P during perimenstrum (days before and during menses) is a vulnerability window for emotional distress, thus the hormonal influence on anxiety and craving may be amplified during perimenstrum. METHOD: Naturalistic daily data were collected from non-treatment-seeking females who endorsed daily cigarette smoking (N = 50). The daily protocol involved morning salivary index of E2 and P and ecological momentary assessments of anxiety and cigarette craving. Days of the menstrual cycle were coded as either occurring during perimenstrum (ie, seven days prior to and first 3 days after menstrual onset) or reference (ie, all other times during cycle). Using multilevel modeling, we tested the main and interactive effects of daily E2, P, and perimenstrum (yes/no) on same-day anxiety and cigarette craving. RESULTS: Results indicated significant three-way interactions between E2 and P both centered between and within perimenstrum for anxiety and craving. In perimenstrum, anxiety and craving were elevated regardless of hormonal balance. However, the association between P and anxiety varied in the context of E2, such that higher P and lower E2 dampened anxiety outside of perimenstrum. Similarly, higher P and lower E2 was associated with lower craving outside perimenstrum. DISCUSSION: These data provide high specificity for understanding hormonal influences on anxiety and craving during the menstrual cycle, which has implications for female-specific models and treatment of the anxiety-smoking comorbidity. Implications: This is the first study to document daily-level associations between salivary E2 and P, and their interaction, with anxiety and cigarette craving, in the context of the female menstrual cycle. Perimenstrum (ie, days before and during menses) appears to confer risk for anxiety and craving regardless of the hormonal balance. However, higher daily P dampened anxiety but only in the context of lower daily E2 on days outside of perimenstrum. For craving, higher P with lower E2 was associated with lower craving outside of perimenstrum. The ovarian hormonal milieu should be considered when understanding the etiology and subsequent treatment of anxiety-smoking comorbidity in females.

  PublisherOA PDFDOI

Recent grants

• A Tailored Physical Activity Smartphone App for Patients with Alcohol Dependence

  NIH · $742k · 2016–2022

• NIH Grant K23DA019950

  NIH · $925k · 2012

• tDCS for Increasing Exercise Adherence in Individuals with Elevated Depressive Symptoms

  NIH · $341k · 2017–2019

• Peer-Facilitated Physical Activity Intervention Delivered During Methadone Maintenance

  NIH · $433k · 2016–2018

• A Smartphone App to Facilitate Buprenorphine Discontinuation

  NIH · $414k · 2016–2020

Frequent coauthors

• Michael D. Stein

  Boston University

  296 shared

• Lisa A. Uebelacker

  Butler Hospital

  265 shared

• Richard A. Brown

  The University of Texas at Austin

  247 shared

• Claire E. Blevins

  Providence College

  211 shared

• Cynthia L. Battle

  Butler Hospital

  190 shared

• Samantha G. Farris

  Rutgers, The State University of New Jersey

  161 shared

• David R. Strong

  University of California, San Diego

  155 shared

• Tosca D. Braun

  Brown University

  120 shared

Education

• Ph.D., Clinical Psychology

  University of California, San Diego

  2001

• Ph.D., Clinical Psychology

  San Diego State University

  2001

• A.B., Psychology

  Harvard University

  1995