Return To Abstract Listing

Scientific Session 07 — Efficacy/Administration/Informatics - Research and Safety

Tuesday, April 19, 2016

Abstracts 1118-3602

3452. Managing Tumor Measurement Data Within PACS

Machado L*,  Folio L. National Institutes of Health, Bethesda, MD

Address correspondence to L. Folio (

Objective: Metastatic tumor burden on cancer patients in therapeutic trials is assessed by selection and temporal measurement of target lesions on cross-sectional imaging; however, oncologists’ records often differ from radiologists’ reports. We aimed to improve the tumor assessment process by working closely with oncologists and exploiting new PACS capabilities. For example, our PACS (version 12.0, Carestream Health) can store measurements in tables and apply Response Evaluation Criteria in Solid Tumors (RECIST) automatically. Our radiologists routinely measure lesions previously measured on follow-up examinations, but not all lesions are targets according to oncologists’ records. In an attempt to minimize discrepancies, radiologists often meet with oncologists to select target lesions on baseline examinations. We believe this workflow improves agreement on follow-up examinations, maximizes the accuracy of measurements and calculations in PACS, and improves eventual direct exportation in cancer databases.

Materials and Methods: We compared measurements in radiologists’ reports on CT of the chest, abdomen, and pelvis to oncologists’ worksheets for a subset of patients with muscle-invasive urothelial cancer. We corrected discrepancies in PACS bookmark tables to eventually compare PACS-automated RECIST calculations to oncologists’ databases. We present the percentage of agreement for both target lesion selection and actual measurements between radiologists and oncologists.

Results: Our preliminary results included eight (of an eventual 40) patients with muscle-invasive urothelial cancer, with a total number of 34 (baseline and follow-up) CT examinations. Agreement of target lesion selection on baseline examinations was 78% (14/18) and measurement agreement was 79% (11/14) for baseline examinations. For follow-up examinations, oncologists selected a total of 49 target lesions and radiologists selected 44, representing 90% agreement. The percentage of agreement between radiologists’ and oncologists’ measurements (within 1 mm) was 84%.

Conclusion: We found better agreement in target lesion selection and measurements between radiologists and oncologists in follow-up examinations, when compared with baseline. We believe this result is due to our work flow, in which radiologists measured lesions that had been previously measured. Provided that oncologists and radiologists agree on target lesions on baseline examinations, measurements in PACS should be more accurate. The discrepancies were not errors or oversights but rather depended on work flow and variable choices of index or target lesions. Also, oncologists often measure lesions independently of the radiologist, especially if the patient is in the clinic and the report is not yet available. We believe managing measurements in PACS will be more efficient, potentially eliminating handwriting measurements on paper forms and eventually reducing errors with direct data exportation. This change will only be possible when bookmark tables of measurements and RECIST calculations in PACS match oncologists’ records.