JMIR Res Protoc. 2026 Feb 3;15:e79875. doi: 10.2196/79875.
ABSTRACT
BACKGROUND: Therapeutic emesis (TE), known as vamana karma, is a classical method of detoxification performed to eliminate vitiated kapha (bio-humor governing fluid regulation and structural cohesion of the body in normalcy) ailments from the body. The assessment of this complete process depends on physicians’ visual assessments of vomitus features and patient responses, introducing subjectivity and interobserver variability. Moreover, this method requires more than continuous monitoring; thus, it can sometimes lead to human error, resulting in missed expelled content or complications. We propose an artificial intelligence (AI) model to monitor TE to observe visual changes (ie, patient vomitus content and gestures) to provide better clinical outcomes. This approach is being explored for the first time in the traditional system of medicine.
OBJECTIVE: This study aims to develop and validate an AI-assisted digital framework for the objective evaluation of TE via (1) automatic vomitus detection, (2) content classification, (3) number of bouts expelled, (4) facial expressions and individual gestures, (5) determination of detoxification type, and (6) provision of a postpurificatory dietary regimen after completion.
METHODS: The study will be conducted in 3 phases. The first is the preparation of standard operating procedure for TE data collection. The second is data annotation of detected vomiting events. All analyses will be conducted using Python libraries, including scikit-learn (version 1.3.2; developed by the scikit-learn contributors, Python Software Foundation), TensorFlow (version 2.14.0; Google Brain Team, Google LLC), and tools supported under Google Summer of Code 2023 (Google LLC), along with SPSS Statistics (version 26.0; IBM Corp) for statistical analysis. In the third phase, model performance will be evaluated using standard machine learning metrics, and agreement with expert assessments will be measured using the Fleiss κ statistic. This study is exploratory in nature. Thus, 50 volunteers will be targeted.
RESULTS: This is the first study of its kind, so to create the dataset, we prepared a standard operating procedure for TE event data collection. Data collection was completed in December 2025. Data annotation and preliminary model preparation are ongoing, with final testing and validation expected to be completed by December 2025. External testing in the health care setting is expected to be completed by February 2026.
CONCLUSIONS: This work presents one of the first attempts to apply deep learning for objective analysis of the TE process in Ayurveda. By combining YOLOv9 for vomit detection and residual neural network for classification, the framework achieves promising accuracy in automated vomit identification. The results will demonstrate the potential of AI-assisted analysis in traditional medicine, although further clinical validation and expansion across multiple centers will be necessary before deployment in real-world settings.
PMID:41632954 | DOI:10.2196/79875
