Visuelle Beurteilung von Tumorgrößenveränderungen im CT – Eine Pilotstudie

Zielsetzung:

When assessing tumor response to therapy on CT, one frequently observes tumor changes that lie within RECIST boundaries of stable disease (-30%/+20%). We investigated the minimum lesion change that radiologists are able to detect without measuring.

Material und Methodik:

Computationally generated lesions (50 HU, range 9 to 33 mm) were displayed on a plain background (-750 HU) in a baseline and follow up fashion simulating lung CT examinations. White noise was added (bandwidth ± 100 HU). Diameter changes were experimentally manipulated to range from – 30% to + 20% at follow up. Images (63 cases, 124 lesions, range 1 to 5 lesions per case) were reviewed in a near-routine setting by 14 radiologists (mean years of training: 4.2, range 0.3 to 12.4). Response classifications by visual judgment (smaller/stable/larger) for all cases were recorded.

Ergebnisse:

Mean review time per case was 15 seconds (range: 6 to 170 seconds). All radiologists distinguished lesion changes within the range of lesion growth/shrinkage of -30%/+20%. Less than 3% (8/285) of lesions rated to increase were actually shrinking or exactly equal. For lesions rated to decrease, less than 1% (3/443) were actually growing or exactly equal. 95% of all stable ratings were located in the interval -7.5% to + 10.0% diameter change.

Schlussfolgerungen:

Radiologists might be able to detect size changes in the range of ± 10% in artificially generated lesions with the unaided eye.

E-Mail: [e-mail]