DOD is Digitizing the World's Largest Pathological Sample Repository

DOD is Digitizing the World's Largest Pathological Sample Repository

The Joint Pathology Center's effort to upload 55 million samples of tissues could pave new opportunities in medical research and patient care.

Federal officials are digitizing the world's largest repository of pathological samples, paving new opportunities in computational pathology, diagnoses and biomedical research with a team of experts from across the field to help in the endeavor.

Since last year, the Defense Department’s Joint Pathology Center (JPC) has been undergoing the effort to upload the data of the 55 million slides and tissues it manages. Federal officials and academic partners behind the effort touched upon the opportunities that the project can bring to medical research and care, as well as challenge areas they’re overcoming during the SXSW conference last week.

“This project is extremely exciting to me and our team because it just represents an enormous opportunity to have a very positive impact on the military and the patients we serve in the federal government,” the center's director, Col. Joel Moncur, said. “[JPC] contains a whole host of both common and uncommon diseases that are, as mentioned, really a treasure trove of data. We do believe that in the modern era, that this is really an unrivaled source of high quality data that can be used for machine-learning algorithms.”

JPC is a collaboration with the Defense Digital Service (DDS), academic pathology experts and industry partners to digitize the samples. The Defense Digital Service, Moncur added, is especially helpful in helping the center establish an end-to-end process for sample digitization.

“We have this broad selection of experts in their field on this call because it was not just software implementation anymore,” the service's Principal Engineer and Cloud Architect Steven French said. “There are ethical concerns, there are legal concerns, there’s engagement with the public, there’s patient advocacy, there’s a lot of things wrapped up, and what would we encode the software to do or not to do. That transcends just the technological aspect, and DDS is kind of perfect for this because we bring a lot of people from private industry, and it’s not just programmers. It’s a broad swath covering pretty much every discipline you can think of.”

Defense Digital Service embarked on the initiative with a discovery sprint to examine the digitization challenge and then partnered with industry vendors to help with requirements around reverse-image searching, indexing slides and integrating metadata that the team would eventually use once they established ethics and stewardship rules around it. 

Johns Hopkins University helped with the slide scanning, which entailed using automated hardware to quickly scan the slides and send the information to a cloud environment where it undergoes a process of ingestion and indexing. 

Harvard Medical School Assistant Professor of Pathology Dr. Faisal Mahmood highlighted how computational pathology and the application of machine learning to the digitized data can help with a variety of current challenges in the pathology field, which has a reproducibility challenge. 

“It has been shown by a number of studies that there’s large inter-observer and intra-observer variability amongst pathologists, and making diagnoses and the cost of misdiagnosis can be very high, depending on the disease," Mahmood said. "Computational pathology really has the ability to standardize the diagnosis and assist in making a more reproducible diagnosis and, eventually, improving patient care.”

Mahmood added how the computational pathology and machine learning can help in identifying morphological markers that would be associated with patients’ response and resistance to treatment, especially if those patients cannot access genomic testing. This can make it easier for medical providers to determine more effective treatment options at an early stage of treatment for patients. 

If it is hazardous to conduct a pathology diagnosis or if there are no pathologists readily available for a patient, Mahmood said that machine learning, deep learning and modern artificial intelligence tools can use JPC's historical data to identify common information with patient data and help providers make better decisions for their patients. 

While the JPC is undergoing this effort, Vanderbilt Pediatrics and Law Professor of Health Policy Dr. Ellen Clayton is advising the team behind the digitization project in stewardship of the repository’s data. This has included the application of ethical use of AI, transparency in data collection and protecting the interests of people who provide pathological samples. 

Moncur added that the JPC is applying DOD’s five principles of ethical AI use, which the DOD adopted last year, in this project, and is currently developing a resource governance policy to follow best practices from Clayton. 

“[Resource governance policy] can help us address all the different aspects of what was described — data, the privacy of data, the security of data, operational security for the military, and all the things that go along with that,” Moncur said. 

The Defense Innovation Board recommended that the Joint Pathology Center digitize its repository just over a year ago, highlighting the benefits that such an effort could yield in helping researchers overcome accessibility barriers to large amounts of and meaningful data about diseases. 

The effort could take anywhere between five and 10 years to fully complete. However, Clayton emphasized how important the effort is for making the repository more accessible and useful.

“Digitization is going to make it much more valuable, so I’m excited about that,” Clayton said.

Standard