Article Figures & Data
Figures
- FIG 1.
Patient-selection flow diagram.
- FIG 2.
Change in individual diagnostic performance of clinicians. Arrow lines depict the change in diagnostic performance for each clinician without (red dot) and with (blue dot) software augmentation. The radiology resident showed no change in specificity between the 2 occasions (black dot). FP indicates false-positive.
- FIG 3.
Source (A) and MIP image (B) from TOF-MRA of 75-year-old female patient. There was a basilar top aneurysm, <3 mm. Human readers detected this lesion, but the computer-assisted detection software did not.
- FIG 4.
Source (A) and MIP (B) images from TOF-MRA of a 35-year-old female patient. Some human readers missed this lesion at first, but computer-assisted detection software detected it. C, During the second interpretation, human readers accepted the diagnosis of the software.
Tables
Reader Human without CAD Assistance Human with CAD Assistance Human without vs with CAD Assistance (P Value) Sensitivity (%) (95% CI) Neurologist 56.3 (47.9–64.7) 84.4 (78.3–90.6) <.001 Neurosurgeon 74.8 (67.5–82.1) 85.2 (79.2–91.2) .032 Radiologist 86.7 (80.9–92.4) 83.7 (77.5–89.9) .493 Resident 76.3 (69.1–83.5) 92.6 (88.2–97) .001 Specificity (%) (95% CI) Neurologist 90.9 (86.8–94.9) 92.9 (89.3–96.5) .460 Neurosurgeon 92.9 (89.3–96.5) 93.4 (89.9–96.9) .842 Radiologist 95.9 (93.2–98.7) 94.9 (91.9–98) .629 Resident 99.5 (98.5–100) 99.5 (98.5–100) >.999 Sensitivity per lesion (%) (95% CI) Neurologist 60.4 (53–67.7) 91.7 (87.2–96.3) <.001 Neurosurgeon 82.2 (76–88.5) 96.4 (93.7–99.2) <.001 Radiologist 90.5 (85.7–95.3) 95.3 (92.2–98.3) .070 Resident 80.5 (74.6–86.4) 96.4 (93.7–99.2) <.001 No. of false-positives per case (95% CI) Neurologist 0.066 (0.044–0.099) 0.072 (0.048–0.108) .751 Neurosurgeon 0.069 (0.047–0.103) 0.09 (0.062–0.132) .200 Radiologist 0.039 (0.022–0.07) 0.075 (0.051–0.112) .028 Resident 0.006 (0.002–0.024) 0.018 (0.008–0.04) .179
