Whether you’ve embraced voice recognition software or rely on a transcriptionist, there’s still the problem of large amounts of text that’s largely unusable. Enter natural language processing, which makes structured reporting and data mining possible by coding the text and extracting data.
Whether you’ve embraced voice recognition software or rely on a transcriptionist, there’s still the problem of large amounts of text that’s largely unusable.
“It’s nothing more than text. It’s not data,” said James Maisel, MD, chairman of ZyDoc, a medical knowledge management solutions company. “What the industry wants is data to do quality assurance, utilization review, research, analytics, case management.”
Enter natural language processing, a new technology taking dictated reports to a new, more usable level. Natural language processing (NLP) makes structured reporting and data mining possible by coding the text and extracting data. Instead of blocks of text, the technology organizes the reports, allowing for alerts and reminders - all in the interest of improved patient care, said Maisel, who will be speaking on the topic at RSNA 2011 this month.
NLP also promises time savings for the radiologist, who can dictate without having to click through menus to location, severity, size, and other elements of a structured report, Maisel explained. “When you dictate at 140 words per minute, clicking through a menu takes a long time,” he said.
Particularly with the new ICD-10 coding set that has far greater detail and the healthcare industry’s push for more structured, usable patient data, NLP can take that dictated description and code it properly, he said.
The coded terms are then deposited in a database for use in practice or department analytics. Data can be shared across systems, including to the billing system or electronic medical records. “Everyone would enter the data once and use it multiple times,” Maisel said.
This is a new technology, he said, and takes voice recognition applications to the next level. As an industry, radiology has been a generally successful adopter of voice recognition software, because the terminology is standardized and the language models provide for good accuracy, Maisel said. Voice recognition software can offer similar time-savings advantages with faster report turnaround, but does require effort on installation to learn the software and adapt the workflow.
Four Strategies to Address the Tipping Point in Radiology
January 17th 2025In order to flip the script on the impact of the radiology workforce shortage, radiology groups and practices need to make sound investments in technologies and leverage partnerships to mitigate gaps in coverage and maximize workflow efficiencies.
Can Generative AI Facilitate Simulated Contrast Enhancement for Prostate MRI?
January 14th 2025Deep learning synthesis of contrast-enhanced MRI from non-contrast prostate MRI sequences provided an average multiscale structural similarity index of 70 percent with actual contrast-enhanced prostate MRI in external validation testing from newly published research.
Can MRI-Based AI Enhance Risk Stratification in Prostate Cancer?
January 13th 2025Employing baseline MRI and clinical data, an emerging deep learning model was 32 percent more likely to predict the progression of low-risk prostate cancer (PCa) to clinically significant prostate cancer (csPCa), according to new research.