AI-Powered Mental Health Diagnostics

A system to dynamically determine the risk level of patients with mental disorders using sensors, mobile devices, and machine learning. The system collects physiological data, activity, voice, and sleep from patients using sensors and devices. It extracts features and patterns using signal processing and machine learning algorithms. A risk level judgment model is trained based on labeled data to predict patient risk. The system monitors patients in real-time, identifies risk levels, and triggers alarms and interventions.

Source

Diagnosing attention deficit hyperactivity disorder (ADHD) using machine learning and natural language processing techniques. The method involves preprocessing text data using NLTK, then applying machine learning or deep learning algorithms to the preprocessed data to create classified output. When a patient's input text is compared to the classified output, if it meets the criteria, an ADHD diagnosis is made.

Source

A depression diagnosis system that uses multi-paradigm electroencephalography (EEG) and machine learning to distinguish between depressed and normal states. The system measures multiple types of EEG signals from a subject, extracts and preprocesses the necessary signals to minimize noise, generates a diagnostic model using machine learning, and diagnoses depression using the model. The extracted signals include steady-state EEG, P3 waves induced by sound stimulation, and LDAEP waves. This multi-paradigm EEG approach aims to provide more comprehensive and accurate depression diagnosis compared to using a single EEG signal type.

Source

Cloud-based, machine learned models for pairing mental health help seekers with care providers based on collaborative relationship patterns. The models use demographic, personality, and other variables to predict a strong therapeutic alliance between a client and provider. This aims to improve matchmaking and reduce dropouts by leveraging machine learning to optimize pairing based on factors beyond just client characteristics.

Source

A multi-level feature fusion classification method for mental illness text that uses deep learning techniques to accurately diagnose mental illnesses based on doctor-patient dialogue texts. The method involves extracting keywords from the texts using algorithms like TFTDF and CHI, then fusing the keyword features with lower-level features extracted using convolutional and recurrent neural networks. This multi-level feature fusion provides a comprehensive representation of the text for more accurate mental illness classification.

Source

Using generative adversarial networks (GANs) to predict changes in behavior, physiology, and well-being due to starting a new job or other stressful life event. The GANs are trained using wearable sensor data and self-reported assessments to generate predictions of feature changes. These predictions are then analyzed to identify indicators of resilience, i.e., maintaining mental health during stress. The GANs use multi-task learning with separate discriminators for features and clusters to improve performance on complex, multivariate datasets.

Source

Digital mental health assessment platform that provides comprehensive screening, testing, and reporting to help individuals accurately identify potential mental health conditions. The platform uses a tiered process with screening questions, confirmatory tests, and targeted rating scales to assess symptoms against diagnostic criteria. It generates reports with diagnostic indicators and suggestions for further evaluation by healthcare providers. The platform aims to educate and empower individuals to make informed decisions about their mental health and encourage seeking professional care.

Source

Automated automatic monitoring of mental health state of users based on data acquired through a mobile computing device. The automated monitoring includes obtaining data from a sensor or user-interface of a mobile computing device gathered during use of the mobile computing device by a user, inferring, from the data, with a trained machine learning and/or artificial intelligence (AI) model, a mental-health state of the user; and storing the mental health state in memory.

Source

Diagnosing mental disorders using encephalography (EEG) signals and deep learning. The method involves feeding EEG signals from a patient into a trained neural network to classify them as having a specific mental disorder like depression, bipolar disorder, or schizophrenia. The network learns discriminative features from the EEG signals to accurately diagnose the disorders. It can also identify subtypes of disorders like treatable depression. This allows objective, neuroimaging-based diagnosis of mental disorders using EEG signals and deep learning models.

Source

An AI-based screening tool for mental health that uses natural language processing and machine learning to assess emotional well-being and stress levels. The tool presents a questionnaire, processes the responses, assigns weights based on importance, generates additional questions, captures facial expressions, and calculates an emotional wellness index for each domain. It then generates a report describing distress levels and specific domains of concern. The tool aims to triage mental health issues and refer users to professional help as needed.

Source

Using AI to recommend personalized mental health treatment plans by ingesting medical assessment data and identifying factors of patients, applying a machine learning model, and generating treatment recommendations that consider factors like age, gender, income, etc. The AI learns from assessment results and can suggest providers, prescriptions, and care regimens. It also updates based on new assessments to improve recommendations.

Source

Artificial intelligence device to detect and manage psychological stress using sensors and machine learning. The device collects environmental data like light spectrum, along with physiological and behavioral data, to predict and manage stress levels. It uses recurrent neural networks (RNN) and support vector machines (SVM) to analyze the sensor data and extract personalized stress indicators. This allows quantifying and tracking stress levels over time. The device can provide personalized insights and recommendations to manage stress based on the analyzed data.

Source

Mental state classification system that provides accurate and reliable mental state classification to individuals and organizations in a non-face-to-face manner. The system uses a server with two platforms - a service platform for user interaction and a mental state classification platform for analysis. The service platform provides questionnaires and face image capture to users. The mental state classification platform extracts heart rate variability (HRV) from faces and uses algorithms to classify mental states based on answers and HRV. It generates reports indicating probabilities of multiple mental disorders. The system aims to provide objective and comprehensive mental state classification via remote interaction.

Source

Improving the accuracy of estimating a person's psychological state by analyzing their mood over time. The method involves calculating probabilities of mood transitions and average persistence times based on historical mood data. These metrics are then learned by a neural network to estimate psychological states. This allows identifying which specific mood transitions and durations have the strongest influence on psychological well-being, beyond just overall mood statistics.

Source

A gamified, AI-enabled mental health screening tool for children and adults that uses games, facial recognition, voice analysis, and machine learning to collect more comprehensive and accurate mental health data compared to traditional paper-pencil assessments. The tool engages users in age-appropriate gamified interviews that collect natural language responses, facial expressions, and voice shifts. AI analyzes the data to flag risk factors, generate personalized recommendations, and calculate risk-protective scores. The tool aims to improve early identification and treatment of mental health disorders by leveraging technology to overcome barriers like stigma, lack of training, and burden on providers.

Source

Diagnosing mental disorders with improved reliability using fuzzy techniques. The method involves obtaining a diagnostic sheet result, deriving symptoms from it, treating them with an artificial neural network, diagnosing if the user has a disorder based on the symptoms and a first disease code, and re-diagnosing using the network again if initial diagnosis is positive. Learning is done using fuzzy hierarchical and decision tree algorithms. This iterative, multi-step diagnosis process improves reliability for mental disorders where initial and late symptoms differ or not all symptoms appear.

Source

Using intracranial electrodes implanted during brain surgery to develop AI models that can accurately identify, classify, and predict psychological brain states based on the high-quality, continuous intracranial EEG data. The implanted electrodes capture high signal-to-noise EEG signals that are not feasible from scalp EEG. These long-duration, labeled iEEG datasets are used to train AI models like deep neural networks to recognize patterns associated with psychological states. The trained models can then be deployed to analyze real-time iEEG or scalp EEG for decision-making in applications like mental health monitoring.

Source

A system for more accurate and efficient diagnosis of autism spectrum disorder (ASD) using deep learning. The system involves a computer-based model that predicts the probability of ASD based on input features like behavioral observations, clinical assessments, and demographic information. The model analyzes videos, speech recordings, and standardized tests to identify patterns characteristic of ASD. It outputs a probability score for diagnosis.

Source

Automated system to provide personalized stepped mental health care without requiring frequent clinician consultations. The system continuously monitors physiological parameters using wearables and smartphones. It extracts features from the data, detects changes in mental state, and adapts the content presented to the user based on the detected changes. This allows tailored mental health resources to be delivered to individuals as their needs evolve. The system integrates physiological and conscious mental state data to improve accuracy.

Source

A machine learning system for processing multi-modal patient data to classify patients into categories. The system uses an encoder neural network to embed the multi-modal patient data into a lower-dimensional latent space. It then clusters the embeddings in the latent space to identify patient categories. This allows categorizing patients based on complex patterns in the multi-modal data beyond traditional manual criteria. The system jointly trains the encoder and decoder neural networks using a reconstruction loss and scaling factors based on relevance of feature subsets to medical conditions. This encourages confounding features to become entangled.

Source

Estimating a user's psychological state from their behavior data using a machine learning model that considers past reference points. The method involves extracting features from the behavior data for each time point, calculating the difference between the current feature and features from past times, and training a model using the features and their differences. This allows the model to estimate the user's psychological state based on how it compares to past states.

Source

Multimodal machine learning approach for mental health evaluation using data from multiple modalities like audio, video, and text. The approach involves processing the modalities separately to extract dynamic representations with features including time domain information. These representations are then fused using a transformer encoder to generate a combined representation. Mental health diagnosis is determined based on the relevance of this combined representation. Feature importance is determined using SHAP to validate the model and identify subsets of data for tracking symptom progression. This allows diagnosis using remotely collected data from devices like smartphones.

Source

A mental health classification system using multimodal neural networks to predict and monitor mental disorders like depression and PTSD using patient data like counseling notes, images, audio, and wearable sensor data. The system involves a neural network trained on consultation and activity data from diagnosed patients to classify mental health states of new patients. The network uses multimodal encoders to fuse and encode the diverse patient inputs. It also leverages techniques like dropout, ReLU activation, transfer learning, and reinforcement learning. The network is updated using federated learning to share parameters across devices.

Source

Multimodal mental illness diagnosis method using machine learning models to diagnose mental disorders based on user-provided audio-visual and text data. The method involves obtaining audio-visual data from user recordings and converting it to text using speech recognition. This is combined with text data from user answers to diagnostic questions. The audio-visual and text data are separately fed into pre-trained diagnostic models to generate probabilities of mental illness states. The probabilities are then combined to confirm the final diagnosis. This multimodal approach leverages AI models to objectively diagnose mental illness based on user-provided data.

Source

Diagnosing mental illness using natural language processing to provide automated screening and assessment of mental health conditions. The method involves extracting labels from user responses to questions about symptoms and history using pre-trained models. This allows identifying possible mental disorders based on the user's natural language input without manual intervention. The labels are then used to make a diagnosis. The method enables scalable, timely, and accessible mental health screening through natural language interaction.

Source

Real-time diagnosis of mental health disorders using multi-tier machine learning models based on user interactions. The system receives user actions during conversations and uses tiered machine learning models to determine emotional states and mental health disorders. The first tier identifies contexts based on user actions. The second tier selects emotional states from the contexts. By chaining these tiers, the system generates mental health disorder recommendations based on real-time user interactions.

Source

Using wearable medical sensors and neural networks to continuously monitor mental health conditions and provide early detection. The system involves training compact neural networks using raw sensor data augmented with synthetically generated data. Wearable medical sensors collect physiological data like heart rate, temperature, and motion. The neural networks are synthesized using a grow-and-prune paradigm to compress the models for efficient edge deployment. This allows continuous mental health monitoring using wearables with minimal battery and computational resources.

Source

Using facial expressions captured over time to predict mental indices like dementia or depression in individuals. The method involves analyzing emotion tags from images of the subject's facial expressions, calculating integrated emotion tags over time periods, extracting preferred features highly correlated with predicted results, and training a mental index prediction model based on those features. This allows predicting the subject's mental index by analyzing their long-term emotional expression patterns.

Source

A computer system and method for clinical diagnosis of attention deficit hyperactivity disorder (ADHD) in adults using a hybrid approach combining machine learning and knowledge modeling. The system receives clinical parameters from a user interface and applies them to both a machine learning model trained on past ADHD cases and a knowledge model capturing expert rules. The outputs from both models are combined to generate a decision on whether the parameters indicate ADHD, no ADHD, or require expert review.

Source

A method for predicting mental health using EEG signals that involves processing time-domain and frequency-domain representations of EEG data using neural networks. The time-domain representation is created by averaging multiple EEG recordings for a user. The frequency-domain representation is created by decomposing the EEG into frequency bands. The neural networks are trained using transfer learning with initial weights from an image processing task, then further trained on EEG data to predict mental health. This allows leveraging existing image recognition models for EEG analysis. The method also involves determining optimal frequency ranges for processing the frequency-domain representations using optimization techniques.

Source

Using wearable devices to predict depressive states based on biological data like skin temperature, steps, sleep, heart rate, etc. A prediction model is generated from data of multiple subjects to learn features like quantiles, standard deviations, and correlations. This model is then used to predict depressive states in new subjects by inputting their biometric data.

Source

Using machine learning to screen for mental health disorders using self-reported questionnaires. The method involves training a diagnostic classifier using labeled training data from questionnaire responses, video, and audio. It extracts feature importance from the training and generates subset classifiers. The best subset is selected and stored as a screening tool. This allows concurrent diagnosis of multiple disorders using a single trans-diagnostic questionnaire. The classifier can use features from any modality like questions, video, and audio.

Source

Reducing complex psychopathological disorders into a set of latent factors common to a variety of different disorders using computational methods. The method involves analyzing responses from multiple psychological assessment batteries for a large number of users to identify underlying basis factors that are common across disorders. This involves applying latent factor analysis algorithms to a structured dataset of responses to reduce the number of disorder symptoms to a smaller set of underlying causal mechanisms. The resulting latent factor graph shows how these basis factors are related to multiple disorders.

Source

Tracking an individual patient's psychopathology through applied graph theory using wearables and prompted responses to collect objective and subjective measures over time. The system monitors physiological data like pulse rate and ecological momentary assessment (EMA) data like mood and decision making through wearables and prompts. It generates a symptom network graph connecting symptoms based on the data to reveal underlying variability and response to treatments. The graph provides clinicians a visualization of causal symptom relationships to target for treatment and monitor response.

Source

Detecting health risks like depression, suicide ideation, gestational diabetes, intimate partner violence, etc. using natural language processing (NLP) of unstructured text data like freeform user input. The NLP extracts features representing health risk factors, which are then classified to determine if they indicate the presence of the risk. Weights are assigned to the classifications and combined to predict the overall risk level. This allows detecting health risks from text data without direct disclosure.

Source

Using neural networks for predicting and early warning of mental disorders. The method involves collecting psychological test data from mental disorder patients, creating a database of psychological test results from the general population, selecting relevant features, training a neural network using the feature-selected database, and using the trained network to determine psychological characteristics and predict disorder classification and risk level for new patients. This allows early warning of mental disorders by leveraging machine learning models to analyze psychological test data.

Source

Disease diagnosis method using artificial neural networks that improves accuracy by capturing the semantic relationships between symptoms and diseases. The method involves processing disease data containing symptom labels. It segments the symptoms, constructs a symptom set, and inputs it into a neural network diagnosis model trained on historical data. The neural network learns the correlation dependency between symptom combinations and diseases, and uses semantic embedding to capture symptom meaning. This improves the network's ability to predict disease diagnosis from symptom inputs.

Source

Providing information on mental disorders using brain wave and brain activity data. The method involves generating EEG and brain activity data from a stimulus presented to an individual. This data is used to train a classification model that can determine if the person has a mental disorder like PTSD or major depressive disorder. The model learns to classify disorders based on brain activity in specific regions.

Source

Detecting and managing cumulative distress levels in individuals by learning patterns of events that lead to elevated distress. The system identifies discrete events that contribute to distress, and over time learns sequences of events that escalate distress. It can then predict a user's distress level based on their current actions, and alert them if sequences similar to ones that led to critical distress are occurring. The system also provides recommendations to avoid events that contribute to distress, and suggests seeking events that reduce distress when levels are high.

Source

Automated detection and monitoring of mental health conditions using natural language processing and machine learning. The system analyzes structured and unstructured patient data like calls, journals, and medical records to extract clinical information. It correlates medical data with mental health conditions using NLP. A machine learning model determines the patient's mental health condition based on the extracted info. The patient is assigned a segment and recommended treatment. It continuously monitors patients during treatment to assess improvement. The real-time monitoring identifies at-risk patients, provides immediate assistance, and reduces undiagnosed cases.

Source

Reducing misdiagnosis rates in clinical decision support systems by training a diagnosis result recognition model. The method involves converting matching results between symptoms and diagnoses into sequence strings representing conditions. These strings are used to train a GAN model that generates diagnosis sequences. The GAN output is compared with the real diagnosis sequences to identify correctness. This trained recognition model is used to validate diagnoses output by a clinical decision support system and reduce misdiagnoses.

Source

Diagnosing mental health conditions using intelligent data analysis to predict medications based on user inputs like symptoms, age, gender, and medication history. The diagnosis involves training decision trees on mental health data and using techniques like random forest and k-nearest neighbors to classify user inputs and recommend medications accordingly. The goal is to provide personalized medication suggestions for mental health conditions based on user inputs.

Source

Artificial intelligence system for accurately diagnosing diseases from patient descriptions using machine learning. The system trains a learning model with two sub-models. One sub-model learns to extract symptom features from patient descriptions. The other sub-model learns to associate symptoms with diseases. The sub-models are jointly optimized to improve diagnosis accuracy from patient descriptions. This allows the system to diagnose diseases based on unclear and informal patient descriptions more accurately than traditional symptom mapping methods.

Source

Psychiatric diagnosis system that uses AI to efficiently and intelligently assess mental health conditions. It acquires patient symptom information, preprocesses it, classifies symptoms, determines severity, and suggests diagnoses. The system can also recognize patient emotions. This automated multi-dimensional assessment helps doctors make more efficient and accurate diagnoses.

Source

Identifying health risks from natural language input using machine learning to detect and predict health conditions like depression, anxiety, and suicidal ideation without direct disclosure. The system analyzes open-ended user text using NLP to generate feature vectors representing health risk factors. It classifies each feature as indicative or not using machine learning models. Prediction weights are assigned to the classifications to determine overall health risk likelihoods. This allows identifying health conditions from user input without requiring direct disclosure.

Source

Recommending diagnostic cases for patients based on their symptoms using artificial neural networks. The method involves analyzing unstructured symptom text from electronic medical records, extracting terms, and training a neural network to infer similarity between symptoms and diagnoses. This allows recommending diagnoses for new patients based on their symptom descriptions. The system uses a pipeline of steps: (1) extracting terms from symptom text, (2) training a neural network using historical diagnosis data, and (3) using the trained network to infer similarity between new symptoms and diagnoses.

Source

A method to screen and identify patients with mental illness using artificial intelligence and big data. The method involves analyzing emotional features of speech and facial expressions of mentally ill patients to recognize different emotional patterns associated with specific mental disorders. The method also involves collecting and labeling emotional feature data from mentally ill patients to train neural network models for fast screening and identification of mental illness based on emotional characteristics.

Source

A large-scale medical record based auxiliary diagnosis system using deep learning to aid in diagnosis decisions. The system has a recurrent neural network (RNN) model, a convolutional neural network (CNN) model, and a fusion computing unit. It converts medical records into fixed-length sequences using vectorization techniques. The RNN model processes the sequences from the past medical history to predict diagnoses, while the CNN model processes shorter sequences from symptoms to predict diagnoses. The fusion unit combines the outputs from both models for more accurate diagnosis predictions.

Source