"Machine Learning Analysis of Lingual Colorimetry and MADRS Anxiety-Depression Score in Acupuncture Patients"

Dr Benoit Bataille350 enrolled

Overview

This observational study aims to assess the potential relationship between tongue colorimetry (using standardized photographic techniques) and anxiety-depression scores measured by the Montgomery-Åsberg Depression Rating Scale (MADRS) in acupuncture patients. Data will be analyzed using machine learning methods to determine whether tongue color features correlate with MADRS scores, possibly contributing to a novel, non-invasive diagnostic tool for anxiety and depression assessment in clinical practice. Participation involves only tongue photography and completion of questionnaires, without any invasive procedures or treatment modifications.

This prospective observational study investigates the correlation between lingual colorimetry, captured using readily available and standardized modern photographic tools (iPhone cameras), and anxiety-depression scores evaluated by the Montgomery-Åsberg Depression Rating Scale (MADRS) among patients attending routine acupuncture consultations. Participants will undergo a simple and non-invasive photographic recording of their tongue using an iPhone, ensuring consistent lighting and standardized positioning to minimize variability. Simultaneously, participants will complete the MADRS questionnaire, a widely validated instrument for assessing anxiety and depression severity. No invasive procedures or therapeutic interventions beyond their usual acupuncture care will be performed. The acquired photographic data will be analyzed using machine learning algorithms to identify potential predictive relationships between distinct colorimetric characteristics of the tongue and the MADRS scores. The objective is to determine whether lingual imaging could serve as a reliable, non-invasive biomarker or complementary diagnostic tool for assessing psychological status in clinical practice. This approach leverages everyday technology (smartphones), promoting ease of replication and broader accessibility in clinical environments. Ultimately, findings from this study could facilitate early detection and monitoring of anxiety-depressive disorders, thus enhancing individualized patient care in complementary and integrative medicine.

Study Type

OBSERVATIONAL

Enrollment

350

Conditions

Eligibility

Sex: ALLMin age: 18 YearsHealthy volunteers:

Medical Language ↔ Plain English

Inclusion Criteria: * Patients aged 18 years or older * Native French speakers * No medical conditions or medications affecting tongue coloration * Consulting for acupuncture with symptoms involving an anxious component Exclusion Criteria: * Presence of psychosis or substance abuse * Eating or taking medication within one hour before the examination * Refusal to complete the Montgomery-Åsberg Depression Rating Scale (MADRS) questionnaire * Refusal to have their tongue photographed

Outcomes

Primary Outcomes

Correlation using Machine Learning between Tongue Colorimetric Features and MADRS Scores

This study aims to evaluate the correlation between standardized tongue colorimetric parameters (obtained through digital photographs) and depression/anxiety scores measured by the Montgomery-Åsberg Depression Rating Scale (MADRS). Machine learning methods will be used to analyze tongue images and identify potential associations between colorimetric features and MADRS scores. Participants will be categorized into two subgroups: MADRS \< 15 and MADRS ≥ 15, to assess the ability of tongue characteristics to differentiate depression severity.

Time frame: Baseline (single evaluation at enrollment)

Secondary Outcomes

Machine Learning Analysis of Tongue Features and MADRS Scores

* Density Diagram. * Partial Correlation Analysis (Train/Test Split): Relationship between tongue color and MADRS scores. * Logistic Regression Model (Train/Test) * Support Vector Machine (SVM) on PCA: Model performance evaluation with ROC analysis. * Deep Learning Model: CNN applied to tongue images, classification based on color and texture. * Shapley Values Interpretation: Feature importance analysis for AI models. * Correlation Between Individual MADRS Items and Tongue Zones: Zone-by-zone statistical analysis. * MANCOVA Analysis: Multivariate analysis of covariance to assess multiple dependent variables. * CNN-based Model on Images: Direct classification of tongue features using convolutional neural networks. * Subgroup Identification via PCA (3 Components): Exploring potential depression subtypes.