Large datasets, logistics, sharing and workflow in screening

Abstract

Cancer screening initiatives exist around the world for different malignancies, most frequently breast, colorectal, and cervical cancer. A number of cancer registries exist to collect relevant data, but while these data may include imaging findings, they rarely, if ever, include actual images. Additionally, the data submitted to the registry are usually correlated with eventual cancer diagnoses and patient outcomes, rather than used with the individual’s future screenings. Developing screening programs that allow for images to be submitted to a central location in addition to patient meta data and used for comparison to future screening exams would be very valuable in increasing access to care and ensuring that individuals are effectively screened at appropriate intervals. It would also change the way imaging results and additional patient data are correlated to eventual outcomes. However, it introduces logistical challenges surrounding secure storage and transmission of data to subsequent screening sites. In addition, in the absence of standardized protocols for screening, comparing current and prior imaging, especially from different equipment, can be challenging. Implementing a large-scale screening program with an image-enriched screening registry—effectively, an image-enriched electronic screening record—also requires that incentives exist for screening sites, physicians, and patients to participate; to maximize coverage, participation may have to be supported by government agencies. Workflows will also have to be adjusted to support registry participation for all screening patients in an effort to create a large, robust data set that can be used for future screening efforts as well as research initiatives.center

Commentary

Imagine a future where all screening exams are automatically entered into a national registry, along with relevant metadata about each individual screened. As a participant in the screening program, your data follow you wherever you go, throughout your life. If you cross national borders, arrangements exist to transfer your data to the other nation’s registry. At every subsequent screening visit, prior screening exams are available for comparison. Never would you have to undergo unnecessary invasive procedures because documentation of a previously completed work-up is unavailable. Screening is universally available; when a cancer diagnosis is made, relevant information about the diagnosis is uploaded to the registry so that it is available when the patient re-enters the screening pipeline. The resulting data set is robust, standardized, and available both to support both clinical care and answer research questions. Effectively, it becomes an image-enriched electronic screening record (IESR).

A number of cancer registries currently exist all over the world. In the UK, the National Health Service’s Breast Screening Programme collects mammography results and breast cancer diagnosis data.¹ A similar mammography registry exists in Germany.² The Finnish Cancer Registry’s Mass Screening Program aggregates data for breast, colorectal, and cervical cancer screening, and includes imaging results for mammography, as well as biopsy results and information about cancer diagnoses. Australia is actively developing a registry for bowel and cervical cancers as part of a larger digital health initiative. And in the USA, active screening programs exist for breast, colorectal, prostate, cervical, and now, lung, cancer. The latter—the American College of Radiology’s Lung Cancer Screening Registry (ACR LCSR)—is the first of its kind approved by the Centers for Medicare and Medicaid Services.³ The National Program of Cancer Registries and Surveillance and the Breast Cancer Screening Consortium, as well as a number of individual state-level cancer registries, collect demographic and diagnostic data on patients who undergo cancer screening or are ultimately diagnosed with cancer. However, across the world, these registries share a common theme: while imaging is a part of the screening process for cancers of the breast, bowel, and lung, only the interpretation of the image data, never the images themselves, is submitted to a screening registry. This is not surprising, as there are challenges in the form of large data sets, logistics, sharing, and workflow when considering an IESR.^{4, 5}

Large data sets

When considering incorporating image data into a large-scale screening record, it is important to remember that the image data will need to be accompanied by comprehensive metadata about the individual. Rather than solely being a screening registry, this proposed record functions as an electronic medical record for screening. This enables radiologists and other physicians to have access to relevant information about the patient at the time of interpretation of screening results or workup of detected findings. This combined requirement to support both image and non-image data requires an organized, standardized means of storing as well as transmitting the data between the screening sites and the IESR. In addition, it requires a solution that can easily transcend geographic boundaries, so that the data can accompany the patient seamlessly as he or she moves around through life.

While existing Integrating the Healthcare Enterprise (IHE) implementation profiles such as Patient Information Reconciliation, Patient Administration Management, and Patient Demographics Query provide some of the foundation for this new technical requirement, additional profiles would have to be developed.⁶ There is the added challenge of lack of standardization in the imaging protocols used at different screening sites. While mammography addresses this by specifying the number and type of views (craniocaudal, mediolateral oblique etc.), and the ACR’s lung cancer screening program specifies the CT scan parameters, this is not the case for other imaging-based screening programs, such as CT colonography.

Some degree of standardization is warranted to enable realistic comparisons between exams obtained from different screening sites. One goal of initiatives such as the Quantitative Imaging Biomarkers Alliance, sponsored by the Radiological Society of North America, and the Quantitative Imaging Network, sponsored by the National Cancer Institute is to promote this type of standardization in order to facilitate the use of imaging biomarkers in clinical trials. In addition, as machine learning techniques and algorithms become more heavily vetted and validated for use in patient care, they may accommodate some of this variation and still allow for meaningful comparison of examinations obtained at different sites.

Logistics

An IESR requires careful planning and maintenance of the data included in the registry. No registry is feasible without an organization or agency to curate the data. A standardized schema for the registry database has to be developed, and the necessary transactions to move data both from the screening site to the registry as well as back to the same or another screening site for comparison to subsequent studies must be established. At present, both of the ACR’s cancer registries require manual entry of patient data. This would be neither practical nor feasible in the IESR, and electronic transactions for automation of data transfer would be required. In many cases, this can reuse existing DICOM and HL7 standards (e.g. Clinical Document Architecture and Cross-Enterprise Document Sharing for Imaging—XDS-I). The FHIR (Fast Healthcare Interoperability Resources) standard, with its REST-based (Representational State Transfer) transactions, will likely also be instrumental, although its various components currently fall at all levels of the maturity spectrum. However, a specific combination of existing or future IHE implementation profiles may be necessary to address the needs of data sharing for screening.^{6, 7} Expecting patients to entrust the safety and security of their data to a third-party that is not the screening site requires implementation of the appropriate security measures to avoid potential data compromise.⁸ Educating the general public about the value of the IESR is a separate but necessary task, especially to provide reassurance and promote participation. This would be especially important in cases where participation in the IESR is not, in fact, universal. The appropriate incentives need to exist for patients, physicians, and screening sites. In cases where a patient requests that his or her data be removed from the registry, e.g. via the “right to erasure” provision offered by the European Union’s General Data Protection Regulation, policies need to be in place to support this request. This would include counseling the patient as to the potential risks to his or her health as a result of being removed from the registry.

Sharing

To develop an image-enriched screening registry, the necessary logistics overlap heavily with those of secure image sharing.⁹ While the technology of image sharing has evolved considerably over the past decade, it is not yet ubiquitous. In the current model, imaging centers and referring physicians’ offices pay for access to the technology, while patients are either not charged or pay a nominal fee. Concerns over cost, data security in the cloud, and perceived lack of control over the data may deter imaging sites from participating.¹⁰ Imaging centers may not be incentivized to share data, as doing so facilitates screening at one site but treatment elsewhere if a screening study should be positive. However, one potential benefit is the ability to access these exams for comparison to diagnostic exams, in addition to routine screening exams. Additionally, from the patient perspective, data sharing may lead to higher compliance with screening at a local site, with the assurance that data transfer to a tertiary center for treatment would be smoother than it currently is for many patients. As such, mandating participation of imaging sites in a screening registry requires incentivizing individual screening centers to share patient data and providing appropriate safeguards for the submitted data as well as for the data retained locally at the participating sites. The Mammography Quality Standards Act, enacted by the United States Congress in 1994, sets forth requirements for accreditation of centers that perform mammography, and requires that patients be able to request their images (in their original film or digital form), but not that they be transmitted or shared between sites automatically. By comparison, the ACR’s Lung Cancer Screening Registry is a voluntary program, but encourages centers to participate by meeting Centers for Medicare and Medicaid Services requirements for quality reporting reimbursements.

Workflow

Consider the screening experience that occurs as part of participation in an IESR. When the individual presents for the screening examination, relevant prior screening exams are downloaded to a local server to enable the radiologist to make comparisons to the new exam. They are viewable side-by-side with the new study, on the workstation typically used to interpret screening exams. The images, interpretation, and relevant metadata for the new exam are securely uploaded to the central repository in an expected, standardized format. The format for the data is clearly defined and made available to each screening location, as are tools to assist in the conversion to a standard format. Individuals who undergo screening are given access to their records for viewing or download. If the screening exam detects a finding that warrants further workup, the facility that performs the workup will be given a token with which to upload relevant data about the workup and eventual diagnosis.

Conclusions

Developing a robust electronic screening record has the potential to benefit patients on an individual basis as well as help scientists gain a better understanding of the imaging and non-imaging biomarkers that herald cancer development. From a technological standpoint, many of the components to achieve this currently exist: the electronic medical record, the vendor-neutral archive, DICOM- and HL7-based IHE implementation profiles for medical data and image transfer, and even the rapidly evolving FHIR standard for web-based access to medical records and medical images. As machine learning gains greater attention as a tool to automate repetitive tasks currently performed by radiologists and other physicians, development of robust, labeled data sets to effectively train and validate these algorithms will become even more important.