We previously demonstrated that current radiology reporting practice is often insufficient for oncologists to apply RECIST 1.0 in the clinical trial setting[6]. We found that the information contained in the radiology reports and image markups were sufficient to apply RECIST 1.0 in only 26% of cases. We also found that radiologists often did not report the presence (33% of the study cases) or size (25% of cases) of a target lesion in follow-up studies, though they frequently created an image markup for the lesion (70% of cases). This suggests that the lesions oncologists are tracking are not being documented in the official medical record. We concluded that the current radiologist-oncologist workflow is insufficient with respect to applying quantitative methods to evaluate response to treatment. Informatics tools are needed for recording, tracking, reasoning, and communicating about the measurable cancer lesions.

We recently created an image Physician Annotation Device (iPad)[7], a plug-in to the OsiriX[8] image viewing application, to implement the Annotation and Image Markup (AIM) standard[9]. Our goal in this study was to extend iPAD to support lesion tracking, to adapt it to include lesion identification and specialized lesion annotations for response assessment, and to show that it will improve lesion identification and measurement required to apply response criteria.

We evaluated the enhanced iPAD tool by asking two users (a radiologists and an oncologist) to use iPAD in performing their response assessments in a central review for two cancer clinical trials. The first clinical trial was a phase II colorectal cancer trial in the metastatic setting with 10 patients, and the second was a phase II follicular lymphoma trial with 13 patients. The users viewed baseline studies and annotated lesions to indicate those who were the target lesions to be tracked on follow-up studies. On the follow-up studies, the users could view the annotations from the baseline study as a guide to those target lesions that need to be annotated. iPad was used to annotated 23 baseline and 69 follow-up imaging studies. The users also completed paper flowsheets to match their intention to annotate target lesions in each baseline and follow-up study. These flow sheets established the gold standard: the number of lesions that should be assessed on each follow-up study. We used the results from our prior study of sufficiency of lesion annotation with unassisted methods[6] as a historical control.

Figure 1: iPad user interface adapted for lesion tracking

Table 1 shows the results of our evaluation. Between the two studies, 86 target lesions were identified at baseline. To apply response criteria, these lesions would need to be annotated 254 times in follow-up studies. Of those expected, the users successfully annotated 237 target lesions in follow-up studies. Compared to the unassisted approach from our prior work[6] where 70% of the target lesions were assessed in follow-up, 93% of the target lesions in follow-up studies in the two trials were assessed by using our tool.

While there was a substantial improvement in the rate of target lesion annotation in follow-up studies with iPAD, when compared to unassisted methods, additional features are needed to support the 100% lesion tracking required to consistently apply quantitative response criteria. Specifically, a digital flow sheet could provide users feedback on complete data entry. Such views of the data are currently being developed.