Type:
Educational Exhibit
Keywords:
Quality assurance, Cancer, Structured reporting, Screening, Outcomes analysis, RIS, MR, CT, Oncology, Computer applications
Authors:
D. J. Vining1, A. Pitici2, C. Popovici2, A. Prisacariu2, M. Kontak2; 1Houston, TX/US, 2Houston/US
DOI:
10.26044/ecr2019/C-3592
Background
The RADS structured reporting schema are designed to improve the content and clarity of radiology reports for the communication of screening examinations.
With the success of BI-RADS for mammography reporting,
there has been a proliferation of other schema for early cancer detection as well as non-cancer related applications including (1-9):
- Lung-RADS (Lung cancer)
- PI-RADS (Prostate cancer)
- C-RADS (Colon cancer)
- LI-RADS (Liver cancer)
- NI-RADS (Head/neck cancer)
- TI-RADS (Thyroid cancer)
- GI-RADS (Ovarian pathologies)
- CAD-RADS (Coronary stenosis)
However,
it is often difficult for radiologists to recall the details of each reporting system during a busy clinical practice unless the radiologist is well-versed in that particular specialty.
The different RADS schema vary in how each is determined and in the meaning of a particular RADS assessment score (Figure 1).
The RADS assessments are determined subjectively (e.g.,
BI-RADS),
by using lookup tables of disease features (e.g.,
PI-RADS),
or by summing values assigned to certain disease features (e.g.,
TI-RADS).
We developed a structured reporting solution that records key images and a radiologist’s verbal descriptions of findings and analyzes the descriptions using natural language processing (NLP) to extract common data elements that define disease features which are used to automatically calculate the appropriate RADS assessment for incorporation into a radiology report (Figure 2).