Development of a LargeScale WholeBody CT Image Database
Development of a Large-Scale, Whole-Body CT Image Database 1 Health SHAMSI DANESHVARI 1 BERRY , HEATHER J. H. 2, 3 EDGAR Informatics and Information Management, University of Mississippi Medical Center, 2 Department of Anthropology, University of New Mexico, 3 Office of the Medical Investigator, New Mexico BACKGROUND RESULTS • Scientific research is often constrained by a lack of appropriate data, especially due to technological limitations. • One whole body CT dataset consists of ~12, 000 2 D images, which can be displayed as individual 2 D images or constructed into 3 D renderings. • The Office of the Medical Investigator (OMI) is a centralized medical examiner office for the state of New Mexico. • De-identification and data transfer to enable sharing of data with an outside collaborator is a multi-step process involving: • Any individual who dies of unknown or suspicious causes is routed to the OMI. This includes not only deaths from homicide and suicide, but a large proportion of natural and accidental deaths as well. In 2010, 35% of deaths in New Mexico were autopsied 1, 2. The racial and ethnic diversity of the state is reflected in the OMI sample. • In 2010, the Center for Forensic Imaging (CFI) at the OMI received a National Institute of Justice grant to determine if the traditional autopsy could be supplemented or supplanted by CT. • Every decedent that underwent an autopsy also received a full-body, high resolution CT scan. • Each individual’s record is ~12, 000 images that represent the whole body scans optimized for bone and soft tissue. • Each record includes 2 -6 scout images (see Figure 1). These scout images are comparable to anteroposterior and lateral whole body radiographs of the decedent. • After the grant ended, the OMI continued to scan nearly every decedent. This has resulted in 15, 000 individuals scanned, but with little associated data. 1. The download and conversion of the image data from Phillips i. Site PACS format to a universal format (DICOM) to enable viewing and quantitative analysis 2. De-identification of the DICOM data 3. Quality assurance of both the image data and de-identification process 4. Maintenance of the identification key 5. Transfer of image data to the outside entity, such as by shipping storage media via Fed. Ex, or by secure ftp transfer over the Internet • The de-identified metadata will be transferred to the Center for Advanced Research Computing (CARC). • CARC will host the website that provide the users with access to the metadata of the CT images. • Because of the large size of the decedent and skeleton CT images, end users will request the data through the website and be sent a portable hard drive. Processes for internet delivery are being investigated. 1. Scout images (Figure 1) will be available to the end user on the decedent metadata website. The full workflow for use of the free-access decedent database can be seen in Figure 2. Figure 1. Scout image example. Case number: 119. Age: 14. 92 years. Height: 116 cm. Weight: 51. 2 kg. BMI: 18. 6 OBJECTIVES • Collect lifestyle and health metadata from both the medical examiner’s database as well as interviews with the next of kin. • Create a Free access Decedent Database available to bona fide researchers as granted by the National Institute of Justice in 2016. RESEARCH DESIGN • Institutional Review Board (IRB) approval was granted from The University of New Mexico’s Human Subjects Research Review Committee to determine and validate a minimum data set (MDS) for the fullbody, 3 D, decedent CT images housed at the OMI. • To create the MDS, a Delphi method survey was conducted, which resulted in 59 metadata variables to associate with the CT images, making them useful in forensics, anthropology, medicine, and dentistry (Table 1)3. • Metadata for the Database will be collected from two sources: • The existing OMI database, VAST, which is used to store and organize information regarding the cause of death, lifestyle, and health information to conduct investigation into the cause of death. • Next of kin interviews to gather remaining unknown metadata fields in order to complement the data present in VAST. Figure 2. Workflow. Components in blue illustrate the end users’ experience of the database. 2. CONCLUSION 3. The metadata associated with the images will require application and confirmation that the data requestor is a bona fide researcher with reason to access the data. 4. Researchers will request access to the metadata and images through a simple process available on the website. Researchers must complete a “Request for Review of Research Involving OMI Resources” describing their education or research project, their data needs, and the members of their team who will require access to the data. 5. Members of the CFI will vet the researchers to ensure that they have adequate reasons to access the data. The CFI will grant or deny permission based on these requirements. 6. After approval of the request, investigators must complete a Data Transfer Agreement outlining the terms and conditions under which the data may be used: • Use of the database and image data is non-exclusive. • Data may not be transferred to a 3 rd party without permission from the OMI Research Committee. 1. The University of New Mexico Center for Forensic Imaging and NIJ will be acknowledged in publications/presentations resulting from use of the data. 2. No attempt will be made to identify the decedents. § § § § § Birth date Death date Medical diagnoses Primary cause of death Contributing cause of death Smoking status Drinking history Current drug use Drug use history Surgical history Current medications History of broken bones Adult dental health Environmental conditions of the cadaver Method used for decedent ID Number of years in the US if born elsewhere Smoking history Current drinking status Presence of implanted devices Presence of dental caries • Current occupation • Length at current occupation • Current zip code • Sex/gender • Race • Country of origin • Parents’ country of origin • Number of pregnancies • Number of live births • Highest education level • Childhood SES status • Adult SES status • Repetitive or habitual activities • Dietary pattern • Birth weight • Presence of congenital abnormalities • Current height • Cadaver length • Current weight • Cadaver weight • Current bone density • Family history of cancer • History of radiation therapy • History of facial trauma • Presence of genetic disorders • Family history of genetic disorders • Presence of scoliosis • History of plastic surgery • Dental health as a child • Major life occupation • Occupation history • Exposure to carcinogens or lethal substances • Exposure to strenuous lifting at work • Length of military service • Manner of death • Time delay between death and CT scan • Location of death • CT scanner settings • Name of person entering information into database Table 1. Minimum data set variables for the Free Access Decedent Database 3. Figures prepared for publication/presentation that include identifying image details that can be rendered from the CT image data must be altered to insure that the decedent cannot be identified. • Research performed on the decedent information collected by the OMI is not covered under the Common Rule, which only applies to research on living humans (see 45 CFR 46. 101(a) & 46. 102(f)). With the Omnibus Final Rule of 2013, PHI cannot be disclosed about decedents for 50 years. However, information collected by the OMI is not covered under HIPAA but rather IPRA (New Mexico Inspection of Public Records). Even though HIPAA does not apply, the database will conform to HIPAA standards. FUTURE WORK 1. Currently, there are no research databases of this size or with full-body CT images of any size. This research will not only create this unique and vital Free Access Database, but also determine a blueprint for other institutions to create similar products. 2. The creation of such databases creates the opportunity to expand research samples and improve public health. This work will result in the first full-body CT database for researchers and educators. 3. To ensure the interoperability between this database and any future databases, Dr. Daneshvari Berry will standardize the fields present and apply terminology standards. 4. The database is currently under development and will be available to researchers by the end of 2018. 5. LITERATURE CITED 6. OMI. Office of the Medical Investigator 2010 Annual Report. University of New Mexico: 2010. 7. US Census Bureau. 2010 US Census. 2010 [cited 2011 January 19, 2011]; Available from: http: //2010. census. gov/2010 census/pdf/2010_Questionnaire_Info. pdf. 8. Berry, S. Metadata Determination for a Cadaveric Collection. Master [thesis]. Albuquerque: University of New Mexico; 2014. Available from: UNM Digital Repository. 9. ACKNOWLEDGEMENTS This project was made possible thanks to the National Institute of Justice grants NIJ 2010 -DN-BX-K 205 and NIJ 2016 -4305 which were respectively awarded to the CFI to evaluate the utility of postmortem CT scans to supplement autopsy and to the University of New Mexico to construct this Free Access Database. Special thanks to Emily Moes for all of her hard work on this project.
- Slides: 1