Radiology: New tool to PARSE reports for nuke med dose data

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon
radiation dose - 68.00 Kb

Researchers at Harvard Medical School have developed an open-source informatics toolkit capable of mining data from existing nuclear medicine report archives, allowing for calculation of organ doses for longitudinal patient-specific dose monitoring, according to a report published online May 24 in Radiology.

Ichiro Ikuta, MD, and colleagues from Brigham and Women’s Hospital (BWH) in Boston, pointed out that the creation of a tool that can give individualized dose information is essential because population averages fail to capture variations in protocols or administered activity over time.

“Meaningful patient-specific longitudinal radiation dose monitoring requires accurate collection of available radiation exposure data and robust methods to convert this data into relevant dose to the patient,” wrote the authors. “To perform internal dosimetry with reasonable accuracy, patient-specific information about administered activity is needed.”

The tool created by Ikuta and his co-authors is called PARSE—Perl Automation for Radiopharmaceutical Selection and Extraction. Text reports from BWH’s RIS are taken by PARSE and first formatted to standardize units of radioactivity. The toolkit then scans the report, searching for instances where units of radioactivity are described.

Five thousand nuclear medicine reports from Sept. 17, 1985, to Feb. 28, 2011, were randomly selected, of which 1,500 were used for an iterative code training process that tailored the search patterns of PARSE. After the toolkit was sufficiently trained, manual validation was performed on 2,359 reports to assess the effectiveness of PARSE. Validation checked for recall, which the authors defined as the proportion of complete reports from which all data fields were able to be extracted, and precision, which was the proportion of reports with extracted data that was correct.

Results of the validation process showed that PARSE was able to extract data with a recall rate of 97.6 percent and precision of 98.7 percent.

Ikuta and colleagues speculated that a tool such as PARSE could open up new possibilities to improve patient safety, quality assurance and longitudinal patient-specific radiation dose monitoring. “Linked notification schemes can trigger quality assurance review of improper administrations and can prompt examination of reporting practice…The data repository can be used to monitor protocol changes over the course of time and could be further used to benchmark against other practices, guidelines, or regulations.”

The authors also used PARSE to create organ dose heatmaps, which visualized organ dose accumulation. Unlike CT, where patient size causes significant variability to organ doses, nuclear medicine internal dosimetry is relatively insensitive to patient size, explained Ikuta et al.

“If one wishes to proceed to radiation risk estimation, these organ doses can then be mapped to associated risks of radiation-induced carcinogenesis by using the Biological Effects of Ionizing Radiation VII report or an alternative risk model of choice.”