Table 1 Variable List: Primary Experiment Data.
From: A Dataset for Understanding Radiologist-Artificial Intelligence Collaboration
Variable | Data Type | Range | Description |
|---|---|---|---|
uid_clean | str | Radiologist-session identifier | |
experiment_id | str | Radiologist identifier | |
experiment_session | int | 0, 1, 2, 3 | Session number |
design | dbl | 1, 2, 3 | Design arm |
patient_id | int | Patient identifier | |
pathology | str | Pathology | |
round | long | [1, 100] ∈ N | Patient-case read sequence within radiologist-session |
(minimum value is 1) | |||
treat | int | 0, 1 | Indicator if radiologist selected treat/follow-up. |
Missing for pathologies where it was deemed not relevant | |||
treatment | str | Information environment under which case was read | |
IN (NIN): Clinical history was (was not) provided, | |||
AI (NAI): AI assistance was (was not) provided, | |||
warmup: Practice image before the experiment started | |||
level | int | 0, 1, 2, 3 | Position in pathology hierarchy (e.g. level 0 is top-level) |
visible | int | 0, 1 | Indicator if pathology was visible when rad submitted case |
probability | dbl | [0, 1] | Probability radiologist reported on interface slider |
severity | str | Response to severity question, when relevant | |
size | str | Response to size question, when relevant | |
position | str | Response to position question, when relevant | |
active_time | dbl | Time spent actively working on case (in seconds) | |
raw_time | dbl | Total time spent on case (in seconds) | |
num_clicks | dbl | Number of clicks on a case | |
group_vietnam | int | 0, 1 | Indicator if the radiologist is from VinMac healthcare system, Vietnam |
group_teleradiology | int | 0, 1 | Indicator if the radiologist is from a tele radiology company |
group_pilot | int | 0, 1 | Indicator if the radiologist is from the experiment pilot |
group_ground_truth | int | 0, 1 | Indicator if the radiologist is a ground truth radiologist |
inventive_round | int | 0, 1 | Indicator if incentives for correct diagnosis were provided |
total_num_rounds | long | [65, 100] | The total number of reads by one radiologist |
chexbert_label | dbl | 0, 1 | ChexBert label for this case / pathology |
alg_pred | dbl | [0, 1] | AI prediction |
with_ai | byte | 0, 1 | Indicator if radiologist had access to AI |
with_ch | byte | 0, 1 | Indicator if radiologist had access to CH |
gt_average_XX | dbl | [0, 1] | Average of ground truth probabilities for group XX (us, vietnam, all, experiment) |
gt_binary_simple_XX | long | 0, 1 | Binary ground truth based on simple average for group XX |
gt_average_active_time_XX | dbl | (35, 344) | Average active time of the GT radiologists |
