A primer on the use of machine learning to distil knowledge from data in biological psychiatry.

Applications of machine learning in the biomedical sciences are growing rapidly. This growth has been spurred by diverse cross-institutional and interdisciplinary collaborations, public availability of large datasets, an increase in the accessibility of analytic routines, and the availability of powerful computing resources. With this increased access and exposure to machine learning comes a responsibility for education and a deeper understanding of its bases and bounds, borne equally by data scientists seeking to ply their analytic wares in medical research and by biomedical scientists seeking to harness such methods to glean knowledge from data. This article provides an accessible and critical review of machine learning for a biomedically informed audience, as well as its applications in psychiatry. The review covers definitions and expositions of commonly used machine learning methods, and historical trends of their use in psychiatry. We also provide a set of standards, namely Guidelines for REporting Machine Learning Investigations in Neuropsychiatry (GREMLIN), for designing and reporting studies that use machine learning as a primary data-analysis approach. Lastly, we propose the establishment of the Machine Learning in Psychiatry (MLPsych) Consortium, enumerate its objectives, and identify areas of opportunity for future applications of machine learning in biological psychiatry. This review serves as a cautiously optimistic primer on machine learning for those on the precipice as they prepare to dive into the field, either as methodological practitioners or well-informed consumers.

Genomic Machine Learning Meta-regression: Insights on Associations of Study Features With Reported Model Performance.

Many studies have been conducted with the goal of correctly predicting diagnostic status of a disorder using the combination of genomic data and machine learning. It is often hard to judge which components of a study led to better results and whether better reported results represent a true improvement or an uncorrected bias inflating performance. We extracted information about the methods used and other differentiating features in genomic machine learning models. We used these features in linear regressions predicting model performance. We tested for univariate and multivariate associations as well as interactions between features. Of the models reviewed, 46% used feature selection methods that can lead to data leakage. Across our models, the number of hyperparameter optimizations reported, data leakage due to feature selection, model type, and modeling an autoimmune disorder were significantly associated with an increase in reported model performance. We found a significant, negative interaction between data leakage and training size. Our results suggest that methods susceptible to data leakage are prevalent among genomic machine learning research, resulting in inflated reported performance. Best practice guidelines that promote the avoidance and recognition of data leakage may help the field avoid biased results.

Evaluating acoustic representations and normalization for rhoticity classification in children with speech sound disorders.

The effects of different acoustic representations and normalizations were compared for classifiers predicting perception of children's rhotic versus derhotic /ɹ/. Formant and Mel frequency cepstral coefficient (MFCC) representations for 350 speakers were z-standardized, either relative to values in the same utterance or age-and-sex data for typical /ɹ/. Statistical modeling indicated age-and-sex normalization significantly increased classifier performances. Clinically interpretable formants performed similarly to MFCCs and were endorsed for deep neural network engineering, achieving mean test-participant-specific F1-score = 0.81 after personalization and replication (σx = 0.10, med = 0.83, n = 48). Shapley additive explanations analysis indicated the third formant most influenced fully rhotic predictions.

Organoid intelligence: Integration of organoid technology and artificial intelligence in the new era of in vitro models.

Organoid Intelligence ushers in a new era by seamlessly integrating cutting-edge organoid technology with the power of artificial intelligence. Organoids, three-dimensional miniature organ-like structures cultivated from stem cells, offer an unparalleled opportunity to simulate complex human organ systems in vitro. Through the convergence of organoid technology and AI, researchers gain the means to accelerate discoveries and insights across various disciplines. Artificial intelligence algorithms enable the comprehensive analysis of intricate organoid behaviors, intricate cellular interactions, and dynamic responses to stimuli. This synergy empowers the development of predictive models, precise disease simulations, and personalized medicine approaches, revolutionizing our understanding of human development, disease mechanisms, and therapeutic interventions. Organoid Intelligence holds the promise of reshaping how we perceive in vitro modeling, propelling us toward a future where these advanced systems play a pivotal role in biomedical research and drug development.

Reproducible Speech Research With the Artificial Intelligence-Ready PERCEPT Corpora.

BACKGROUND: Publicly available speech corpora facilitate reproducible research by providing open-access data for participants who have consented/assented to data sharing among different research teams. Such corpora can also support clinical education, including perceptual training and training in the use of speech analysis tools. PURPOSE: In this research note, we introduce the PERCEPT (Perceptual Error Rating for the Clinical Evaluation of Phonetic Targets) corpora, PERCEPT-R (Rhotics) and PERCEPT-GFTA (Goldman-Fristoe Test of Articulation), which together contain over 36 hr of speech audio (> 125,000 syllable, word, and phrase utterances) from children, adolescents, and young adults aged 6-24 years with speech sound disorder (primarily residual speech sound disorders impacting /ɹ/) and age-matched peers. We highlight PhonBank as the repository for the corpora and demonstrate use of the associated speech analysis software, Phon, to query PERCEPT-R. A worked example of research with PERCEPT-R, suitable for clinical education and research training, is included as an appendix. Support for end users and information/descriptive statistics for future releases of the PERCEPT corpora can be found in a dedicated Slack channel. Finally, we discuss the potential for PERCEPT corpora to support the training of artificial intelligence clinical speech technology appropriate for use with children with speech sound disorders, the development of which has historically been constrained by the limited representation of either children or individuals with speech impairments in publicly available training corpora. CONCLUSIONS: We demonstrate the use of PERCEPT corpora, PhonBank, and Phon for clinical training and research questions appropriate to child citation speech. Increased use of these tools has the potential to enhance reproducibility in the study of speech development and disorders.

Psychophysiological Arousal in Young Children Who Stutter: An Interpretable AI Approach.

The presented first-of-its-kind study effectively identifies and visualizes the second-by-second pattern differences in the physiological arousal of preschool-age children who do stutter (CWS) and who do not stutter (CWNS) while speaking perceptually fluently in two challenging conditions: speaking in stressful situations and narration. The first condition may affect children's speech due to high arousal; the latter introduces linguistic, cognitive, and communicative demands on speakers. We collected physiological parameters data from 70 children in the two target conditions. First, we adopt a novel modality-wise multiple-instance-learning (MI-MIL) approach to classify CWS vs. CWNS in different conditions effectively. The evaluation of this classifier addresses four critical research questions that align with state-of-the-art speech science studies' interests. Later, we leverage SHAP classifier interpretations to visualize the salient, fine-grain, and temporal physiological parameters unique to CWS at the population/group-level and personalized-level. While group-level identification of distinct patterns would enhance our understanding of stuttering etiology and development, the personalized-level identification would enable remote, continuous, and real-time assessment of stuttering children's physiological arousal, which may lead to personalized, just-in-time interventions, resulting in an improvement in speech fluency. The presented MI-MIL approach is novel, generalizable to different domains, and real-time executable. Finally, comprehensive evaluations are done on multiple datasets, presented framework, and several baselines that identified notable insights on CWSs' physiological arousal during speech production.

A seq2seq model to forecast the COVID-19 cases, deaths and reproductive R numbers in US counties.

The global pandemic of coronavirus disease 2019 (COVID-19) has killed almost two million people worldwide and over 400 thousand in the United States (US). As the pandemic evolves, informed policy-making and strategic resource allocation relies on accurate forecasts. To predict the spread of the virus within US counties, we curated an array of county-level demographic and COVID-19-relevant health risk factors. In combination with the county-level case and death numbers curated by John Hopkins university, we developed a forecasting model using deep learning (DL). We implemented an autoencoder-based Seq2Seq model with gated recurrent units (GRUs) in the deep recurrent layers. We trained the model to predict future incident cases, deaths and the reproductive number, R. For most counties, it makes accurate predictions of new incident cases, deaths and R values, up to 30 days in the future. Our framework can also be used to predict other targets that are useful indices for policymaking, for example hospitalization or the occupancy of intensive care units. Our DL framework is publicly available on GitHub and can be adapted for other indices of the COVID-19 spread. We hope that our forecasts and model can help local governments in the continued fight against COVID-19.

A Weakly Supervised Learning Framework for Detecting Social Anxiety and Depression.

Although social anxiety and depression are common, they are often underdiagnosed and undertreated, in part due to difficulties identifying and accessing individuals in need of services. Current assessments rely on client self-report and clinician judgment, which are vulnerable to social desirability and other subjective biases. Identifying objective, nonburdensome markers of these mental health problems, such as features of speech, could help advance assessment, prevention, and treatment approaches. Prior research examining speech detection methods has focused on fully supervised learning approaches employing strongly labeled data. However, strong labeling of individuals high in symptoms or state affect in speech audio data is impractical, in part because it is not possible to identify with high confidence which regions of a long speech indicate the person's symptoms or affective state. We propose a weakly supervised learning framework for detecting social anxiety and depression from long audio clips. Specifically, we present a novel feature modeling technique named NN2Vec that identifies and exploits the inherent relationship between speakers' vocal states and symptoms/affective states. Detecting speakers high in social anxiety or depression symptoms using NN2Vec features achieves F-1 scores 17% and 13% higher than those of the best available baselines. In addition, we present a new multiple instance learning adaptation of a BLSTM classifier, named BLSTM-MIL. Our novel framework of using NN2Vec features with the BLSTM-MIL classifier achieves F-1 scores of 90.1% and 85.44% in detecting speakers high in social anxiety and depression symptoms.

DAVE: Detecting Agitated Vocal Events.

DAVE is a comprehensive set of event detection techniques to monitor and detect 5 important verbal agitations: asking for help, verbal sexual advances, questions, cursing, and talking with repetitive sentences. The novelty of DAVE includes combining acoustic signal processing with three different text mining paradigms to detect verbal events (asking for help, verbal sexual advances, and questions) which need both lexical content and acoustic variations to produce accurate results. To detect cursing and talking with repetitive sentences we extend word sense disambiguation and sequential pattern mining algorithms. The solutions have applicability to monitoring dementia patients, for online video sharing applications, human computer interaction (HCI) systems, home safety, and other health care applications. A comprehensive performance evaluation across multiple domains includes audio clips collected from 34 real dementia patients, audio data from controlled environments, movies and Youtube clips, online data repositories, and healthy residents in real homes. The results show significant improvement over baselines and high accuracy for all 5 vocal events.