Quantitative and Qualitative Impact of CT-Based Radiotherapy Dose Maps on Radiologists' Interpretation of Post-treatment Thoracic Surveillance Imaging

Abstract

For diagnostic radiologists, interpretation of surveillance imaging for oncology patients treated with radiation therapy (RT) can be challenging because (1) the imaging order may not adequately describe the radiation fields and (2) RT treatment effect and progression can appear similar. Volumetric dose visualization used for plan review is often inaccessible to radiologists. We hypothesize that displaying RT dose would improve radiologists' confidence and ability to correctly identify and distinguish irradiated targets and treatment effects. CT images were read by a board-certified cardiothoracic radiologist and a diagnostic radiology resident. The readers interpreted pre-RT, treatment planning, and 3-4 month post-RT CT images in anonymized software sessions first without, then-after a 1 month "washout" period-with access to RT dose overlay. Six color-coded isodose lines ranging from 25% to 110% represented in absolute cGy were displayed along with a brief clinical history. RT fractionation schedules ranged in BED10 from 39 to 112.5 Gy. Readers were asked to label the treated lesion(s) and treatment effect(s), and record their confidence using a Likert scale of 1-5 and agreement with statements using yes/no responses. Two readersindependently interpreted imaging for 32 patients who received thoracic RT to 1-5 lesion(s) for primary (24) or metastatic (8) cancer. Nineteen patients had 1 lesion and 13 patients had >1 lesion. Correct identification of all treated lesions significantly increased with the addition of dose visualization (61% to 81%; McNemar test, p = 0.00079), with the largest increase noted for cases with >1 lesion (15% to 54%; McNemar test, p = 0.0039). With the addition of dose information, the number of false negatives attributable to missed extranodal targets fell from 52% to 18%. Without dose information, 13% of labeled lesions and treatment effects fell outside of the 25% isodose lines, representing false positives. With the addition of dose information, false positives fell below 2% for both lesions and treatment effects. The readers' confidence that they had identified treated lesion(s) increased from a rating of 4.1 to 4.8 on a scale of 1-5 (Paired two-tail t test; p = 0.000005). Whendiagnostic radiologists have access to dose visualization, correct identification rate of irradiated lesions and treatment effects, as well as confidence in these identifications significantly increased. The decrease in false negatives could reduce potential missed identification of tumor progression while the decrease in false positives could reduce inaccurate identification of treatment failure in a new or stable lesion. Our results demonstrate that adding volumetric visualization of dose to imaging could improve quality of surveillance care for patients with irradiated thoracic malignancies.