The National Institutes of Health (NIH) is turning to the private sector to help it collect, store, and disseminate medical images and data related to cancer screening.
The data management solution is needed to manage data stemming from the NIH’s National Lung Screening Trial (NLST). In a Feb. 12 sources sought notice on Beta.Sam.gov, NIH explained that the NLST has demonstrated a substantial reduction in lung cancer mortality in subjects screened with low-dose computerized tomography (LDCT) as compared to chest radiographs.
However, there was also a very high false-positive rate (FPR) with the LDCTscreens. In addition to the high FPR, NIH explained that there is a need for improvement in predicting risk among those with positive LDCT screens. The impact of a high FPR and the limited ability to predict risk levels is detrimental to both patient care and wellbeing, as well as putting a strain on the health system.
NIH is turning to AI, which it says is “poised to transform medical imaging” to help solve these issues. NIH believes that AI can substantially reduce the FPR of LDCT screening while minimally affecting test sensitivity, thereby reducing diagnostic uncertainty. For the AI to be useful, NIH says it needs to create high-quality image databases.
The image database must:
- Be sufficiently large, with an adequate number of images for training/deep learning data set and an independent validation data set;
- Be accessible to the research communities without burdensome technical and administrative hurdles but with adequate controls for data security;
- Have accurate and sufficient clinical and demographic data linked to the images;
- Have images produced on current, state-of-the-art imaging technology; and
- Have images performed on populations that are representative of the intended-use population of the AI tools.
NIH said that around 20,000 LDCT images will need to be collected, with roughly 10 percent associated with a subsequent lung cancer diagnosis. The images will also need to be collected from at least 75 different healthcare facilities or systems. For context on the needed size of databases, NIH noted that an LDCT image is about 0.1 gigabytes of storage.
In addition to providing secure storage of the images, demographic data, and clinical data, the solution needs to ensure authorized access and prevent unauthorized access to stored data and data storage resources (in-house, external data center, or cloud). The solution will also need to comply with all Federal data security regulations, including the Privacy Act, FISMA, and FEDRAMP.
Responses are due March 4, 2021.
