JMIR Publications

ABSTRACT

Background:

Vamana Karma (therapeutic emesis) is a classical purification therapy in Ayurveda, particularly indicated for Kapha dominant conditions. In traditional diagnostics, there is predominant subjectivity in the visual assessment of the character and content of vomitus leading to observer bias and interobserver variability. The integration of artificial intelligence (AI) and machine learning (ML) has facilitated the elimination of subjectivity and the preservation of consistency in evaluating treatment responses.

Objective:

The study intends to make and test an AI-assisted digital framework for the study of Vamana Karma. With the framework, vomitus events will be identified, their content classified, and facial signs and reactions in the patient studied to strengthen the reliability and repeatability of the procedure’s review.

Methods:

A prospective observational study is being conducted at a tertiary Ayurveda institute. Video data of Vamana procedures are recorded using standardized protocols. A dataset of annotated images and videos is being developed for training YOLOv8-based object detection models and ResNet-based content classification models. DeepFace will be used for facial expression analysis. The system's performance will be evaluated using standard machine learning metrics, and agreement with expert assessments will be measured using Fleiss' Kappa statistic.

Results:

Participant enrollment and data collection are ongoing, with final results expected by July 2025. Preliminary data analysis and model training are in progress.

Conclusions:

The proposed AI framework has the potential to modernize and standardize the evaluation of Vamana Karma, reducing observer dependency and improving clinical reproducibility. Successful implementation could serve as a model for the integration of AI into traditional healthcare practices.