Comprehensive Archive Of Imaging In Oncology(CHAVI),2022 – IMPRI Impact And Policy Research Institute

Policy Updates
Swanami Ghosh

Background

CHAVI (Comprehensive Archive of Imaging in Oncology) is India’s first imaging biobank solely focused on oncological imaging. It was founded as a research partnership by Tata Medical Center, Kolkata, and the Indian Institute of Technology (IIT), Kharagpur, as a part of the larger National Digital Library of India (NDLI) initiative. The establishment of CHAVI fills a crucial gap for high-quality, varied, and well-annotated medical image datasets. There is a paucity of data from the Indian population in existing global datasets, which prevents local implementation of AI models designed for diagnostics and prognostics to be effective. 

Goal:The main goal of CHAVI is collaborative cancer research by establishing a national bank of de-identified oncological images, along with detailed clinical, pathological, genomic, and treatment-related information. This facility will provide a basis for future collaborative studies among Indian cancer centers and allow scientists globally to use this novel dataset for numerous research investigations, such as the creation of sophisticated AI and machine learning models.

The study received ethics clearance from Tata Medical Center Institutional Review Board in August 2018 and was launched in September 2022. Financing for CHAVI has been availed through the Government of India Ministry of Education (previously Ministry of Human Resource Development).

Functioning

CHAVI is a multi-tier, web based medical imaging databank that functions as follows. CHAVI’s primary objective is to help create an organized collection of radiological and clinical datasets of cancer patients in a relational database, through collection, de-identification, annotation and secure storage. This application of healthy and diseased data spans several imaging types including CT, CBCT, PET-CT, X-RAY mammograms, MR, and ultrasound. 

The central functioning of CHAVI is listed below:

• De-Identification of Data: A custom developed solution which anonymizes DICOM files and de-identifies more complex Radiation Oncology (RT Structure Set, RT Dose, RT Plan, RT Registration) objects, while allowing researchers to maintain the longitudinal information with only the necessary data de-identified.
• Relational Data Model: An enhanced data model that allows the storage of clinical, pathological, genomic, and treatment data on cancer patients, and allows complex queries and holistic analysis.
• REDCap Integration: The use of a REDCap (Research Electronic Data Capture) based database template, which allows fast and standardization of data collection
• Controlled Data Access: Supports confined access to the CHAVI, which requires the researcher to comply with a data access policy and tailored for CHAVI data access, fill out a user profile, sign a user agreement, and this includes providing a statement of purpose the researcher will use CHAVI data to achieve in their work.
• FAIR Principles Compliance: CHAVI is compliant with FAIR (Findable, Accessible, Interoperable, and Reusable) data principles and is registered in the FAIRsharing database. 
• DICOM Viewer: An integrated DICOM viewer allows researchers to review previously uploaded de-identified datasets. 
• Role-Based Access Control: A strong role-based access control module limits the access of individual users and enhances data security. 

An integrated partnership between medical and technology organizations results in both the clinical relevance of the CHAVI database and strong technological infrastructure. 

source: CHAVI

Pilot Projects for CHAVI:

As a technology demonstrator, the CHAVI database was piloted with radiation oncology imaging data across a number of research projects undertaken at the Department of Radiation Oncology, Tata Medical Center. These research projects provided the basis for the initial development and validation of CHAVI’s custom image de-identification system and relational data model. The principal projects were:

• INTELHOPE: This is a randomized prospective trial on the effect of radiation dose escalation to the PET-defined gross disease in locally advanced oropharyngeal, laryngeal, and hypopharyngeal cancers. Patients receive PET-CT before treatment to establish the dose escalation volume. The aim of the study is to assess toxicity and effect on local control. 

• Lung IMPRINT: The project is on radiotherapy patients for lung cancer. It assesses a way of forecasting response to treatment based on alterations seen in cone beam CT images acquired during radiotherapy. The target is to enhance individualization of radiotherapy for improved local control and decreased morbidity. 

• Glioma Radiomics: In this retrospective analysis, the correlation between radiomics features of high-grade brain tumors (such as glioblastomas) and their prognosis is assessed. Prognosis in such tumors continues to be poor even with advances, and radiomics features may be of help. 

• CHAVI RO (Radiation Oncology): This refers to the general pilot phase in which radiation oncology imaging information was utilized to pilot-test and calibrate the CHAVI system, such as de-identification of particular DICOM RT objects (Structure Set, Dose, Plan, Registration).

These initiatives illustrate the ability of CHAVI to accommodate varied, complicated imaging and clinical information for large-scale data contributions.

source: CHAVI

Performance

Although precise performance statistics are not easily available in the public sphere for the past 2-3 years, a performance evaluation can be approximated based on its operational milestones and acknowledgment:

• Official Roll-Out and Exposure: In September 2022 CHAVI transitioned to an operational state, with major exposure as India’s first relational, fully annotated, and de-identified oncology image bank.

• Pilot Project Success: As a technology demonstrator, CHAVI successfully consolidated radiation oncology imaging data from pioneer research projects (INTELHOPE, Lung IMPRINT, Glioma Radiomics) and proved its capacity for data accumulation and management.

• Scalability and Integration: Its creation as a component of the NDLI indicates integration into a broader national digital environment, indicating potential for broader implementation and data contribution. The system is built to be scalable and promote contributions from many cancer centers.

• Research Output: The CHAVI website and scholarly databases include publications, e.g., “Design and Development of a Medical Image Databank for Assisting Studies in Radiomics” (2022) in the Journal of Digital Imaging. This reflects ongoing research using the platform and its technical architecture and initial applicability.

• User Engagement (Implied): The continued focus on engaging researchers, academicians, and industry experts in the use of and contributing to CHAVI suggests a growing user base and contribution of data, although particular numbers are not released to the public.

Quantitative numbers are scarce within easily available public reports, but the successful release of CHAVI, continued operational status, initial data consolidation, and acceptance with the scientific community suggest a positive trend for performance.

Impact

The impact of CHAVI is significant, mainly in expanding cancer research and taking advantage of digital health technologies in India.

• Catalyst for AI-Radiomics Research: CHAVI offers a vital, ethnically representative dataset necessary for training and validating radiomic AI models. This fills an important gap: the “poor performance of AI models developed and then used in India” because there are no representative Indian datasets. This will allow for more precise early detection, diagnosis, prognosis prediction, and personalized treatment planning.

• Increasing Data Diversity: With an emphasis on data from the Indian subcontinent, CHAVI addresses the challenge of ethnic bias prevalent in worldwide medical imaging datasets, making AI solutions built with CHAVI data more applicable and efficient for the Indian community.

• Promoting Collaboration: Agreement to FAIR principles and its open but controlled access policy promotes collaboration among national as well as international scientists. This cross-institutional data access is crucial to speed up scientific discovery and innovation in oncology.

• Indirect Enhancement of Patient Outcomes: Although CHAVI is not directly providing patient care, the research it will facilitate will culminate in the creation of better diagnostic tools and more efficacious, targeted treatment protocols, ultimately leading to better patient outcomes over the long term.

• Foundation for National Health Data Infrastructure: Being part of the NDLI, CHAVI plays an important role in establishing a strong national digital infrastructure for healthcare data, paving the way for similar programs in other medical specialties in India.

• Educational and Training Resource: The large and annotated datasets can be used as a goldmine by medical students, residents, and practitioners as a rich source of real-world cases to learn, train, and conduct research.

• Ethical Management of Data: The rigorous de-identification processes and restricted access policies of CHAVI safeguard the ethical management of sensitive patient data, ensuring responsible and credible research practices.

Emerging Issues

1. Volume of Data and Holistic Coverage:  Though a pioneering initiative, the existing volume of data from early pilot projects must be substantially increased to span the wide spectrum of cancer types, stages, and demographic variations in India. Limited comprehensiveness could limit its impact.Adopt a pan-India hub-and-spoke model (as conceived by MIDAS India) to onboard more and more cancer centers and medical institutions aggressively for constant data contribution. Offer incentives and technical support for smooth integration.

2. Standardization and Interoperability: Sustaining and upgrading data standards to provide interoperability with new national and international medical imaging datasets and AI platforms with the progress of technology is an ongoing problem.Taking an active role in national and international standardization efforts on medical imaging and clinical data and updating the CHAVI platform regularly to integrate new standards are therefore very essential. 

3. Computational Resources and Accessibility:Large, sophisticated imaging datasets require a great deal of computational power, which may not be equally accessible to every researcher, especially those from smaller or less well-funded institutions.Investigate collaborations with national supercomputing centers and cloud computing providers to provide scalable computational capabilities. Create easy-to-use interfaces and APIs for researchers of different technical skills to access and analyze data easily.

4. Sustainable Funding and Strong Governance: As a research-oriented project, obtaining sustained, long-term funding is required to support maintenance, platform updating, and ongoing data acquisition . Well-defined governance and data ownership principles are also necessary.Diversifying funding sources through investigation of industry collaborations, philanthropic giving, and possibly a tiered access model (e.g., free for academic research, licensed for commercial use). Create an independent governing board with multi-disciplinary representation to manage data policy and operations.

5. Awareness and Researcher Adoption:Prospective uses among the broader research community, particularly beyond the original collaborating institutions, may remain limited. Thus organizing focused outreach programs, workshops, and training sessions among oncology, radiology, and computer science researchers is vital..

Way Forward 

CHAVI is a key foundation for India’s ambitions in digital health and advanced cancer research. It is important that CHAVI develops continually and is more widely adopted in order to meet the ambitions of “New India” in healthcare. A significant component of the existing portfolio of data should be formally expanded, to obtain more types of cancer, treatment procedures, and longitudinal follow up data. This work will require collaboration with more cancer centers, in different parts of India, within different demographic groups to achieve a national biobank of truly representative samples.

We should try to ensure  investment in technological infrastructure for including best practices for deidentification capabilities, data storage and management system and user interface. We should also explore the possibilities of actually using AI/ML tools within the platform for platform analysis and model development. This will require organizational expertise to plan interdisciplinary workshops, hackathons, and grand challenges involving various interested parties to stimulate new research using CHAVI. 

References : 

• Administrator, C. (n.d.). CHAVI – The first Oncology Image Bank of India. CHAVI. https://chavi.ai/
• CHAVI. (n.d.). Contribute to CHAVI. https://chavi.ai/contribute
• Desai, V. (2022, September 28). CHAVI – India’s first oncology image bank launched to accelerate collaborative cancer research. Healthcare Executive. https://www.healthcareexecutive.in/blog/chavi
• Chatterjee, S., Chakrabarty, S., & Mukhopadhyay, J. (2018). Patient information leaflet and consent form. https://chavi.ai/sites/default/files/2022-09/CHAVI%20RO%20Consent_%20English%20%281%29.pdf
• Press Information Bureau. (n.d.). IIT Kharagpur and Tata Medical Center set up comprehensive digital archive of cancer imaging. https://www.pib.gov.in/PressReleasePage.aspx?PRID=1580546
• Kundu, S., Chakraborty, S., Mukhopadhyay, J., Das, S., Chatterjee, S., Achari, R. B., Mallick, I., Das, P. P., Arunsingh, M., Bhattacharyya, T., & Ray, S. (2022). Design and development of a medical image databank for assisting studies in radiomics. Journal of Digital Imaging, 35(3), 408–423. https://doi.org/10.1007/s10278-021-00576-6

Acknowledgement: The author sincerely thanks Aasthaba Jadeja and IMPRI fellows for their valuable contribution.

About the author : Swanami Ghosh, economics undergraduate student in  Miranda House ,DU and a Research intern at Impact and Policy Research Institute (IMPRI)

Disclaimer: All views expressed in the article belong solely to the author and not
necessarily to the organisation.

Read more at IMPRI:
When Women Join the Boardroom, But Don’t Get a Seat at the Table
Jane Street’s India Signal: Redirect Risk Toward Venture Capital