NUMERO: #20260171

Título: Building standardized ecosystem for AI-ready medical imaging datasets of tobacco induced oral lesions: MIDAS Oral hub

Nome do Apresentador: Anubhuti SOOD

Categoria do Trabalho: Pôster - Investigação Original (IO)

Área Temática: Áreas afins no caráter multidisciplinar

Resumo: Objective
To develop standardized, interoperable framework for creating high quality artificial intelligence (AI) ready medical imaging datasets through the MIDAS-Oral Hub for oral healthcare.
Study design
A methodological development study was undertaken under the MIDAS-India initiative. Key bottlenecks in dataset creation were identified, including data collection, ontology design, annotation processes, and quality assurance. Standardized workflows were established for clinical, gross, and histopathological image acquisition. Interoperability was ensured through alignment with DICOM, FHIR, and ABDM standards. Structured ontologies and iterative annotation tools were developed, alongside objective and subjective image quality checklists.
Results
The MIDAS initiative boasts of 2.45 lakh imaging datasets with patient metadata, comprehensive manual detailing standardized protocols for data collection, storage, and sharing. Ontology includes approximately 2500 general concepts, with around 400 specific to oral lesions. A dedicated annotation platform supports iterative review, quality assurance, and controlled open access. Implementation of structured quality checklists and ontology-driven annotation significantly improved dataset consistency and reliability.
Conclusion
The MIDAS-Oral Hub establishes a scalable framework for generating high-quality, interoperable medical imaging datasets. It addresses key challenges in dataset standardization within India's healthcare system and provides a replicable model for other disease domains and low- and middle-income countries aiming to develop region-specific AI solutions.

Autor 1:  Anubhuti SOOD

E-mail 1:  [email protected]

Autor 2:   Deepika Mishra

E-mail 2:  [email protected]

Autor 3 :  Varun Surya

E-mail 3:  [email protected]

Autor 4:  Debnath Pal

E-mail 4:  [email protected]

Autor 5:  Harpreet Singh

E-mail 5:  [email protected]

Autor 6:  Rahul Yadav

E-mail 6:  [email protected]

Autor 7:  Hari Krishna Raju Sagiraju

E-mail 7:  [email protected]