Article Text

P21 Pulmonary nodules: assessing the repeatability of imaging biomarkers of malignancy
  1. A Talwar1,
  2. JMY Willaime2,
  3. LC Pickup2,
  4. M Enescu2,
  5. D Boukerroui2,
  6. W Hickes1,
  7. MJ Gooding2,
  8. NM Rahman1,
  9. T Kadir2,
  10. FV Gleeson1
  1. 1Oxford University Hospitals NHS Foundation Trust, Oxford, England UK
  2. 2Mirada Medical Ltd, Oxford, UK


Introduction The BTS Pulmonary Nodule Guidelines recommend the use of nodule volumetry as a biomarker of malignancy in pulmonary nodules (PNs). Unfortunately, there is significant inter-scan volumetry measurement variability not representing true growth, of up to 25% (Gietema et al., 2007), with reliable growth detection requiring a scan time interval of up to 12 months. CT Texture analysis (CTTA) does not require volumetry detectable growth to detect change and may be a useful biomarker of malignancy.

Aims and objectives To assess the repeatability of texture features extracted from PNs and compare this to the inter-scan variability of volume measurements.

Materials and methods 40 patients, (20 with an indeterminate PN and 20 with pulmonary metastases) underwent two Low Dose Volumetric CT scans within a 60 minute period.

20 texture features previously used in combination to predict nodule probability of malignancy (BTS 2015) were extracted from each automatic contoured region surrounding the PN.

The variability of texture measurements within individual nodules was assessed by computing the relative differences between baseline and validation scans. Mean and standard deviation (sd) were estimated from the relative differences. Lower and upper limits of repeatability (LLR and ULR) were calculated as mean ± 1.96 × sd. The intra-class correlation coefficient (ICC) was also used to assess the repeatability of the image features for this group of patients.

Results Nodule volumes ranged from 76 to 8130 mm3, (mean 2D diameter 8.7 mm; sd 3.2) and were not statistically different between baseline and validation scans (p = 0.92, Wilcoxon rank sum test).

The mean difference in volume between the two scans was 37.4 mm3 (6.2%, sd 30.4).

18 out of 20 textural features displayed ULR and LLR below ± 26.2% (sd ≤ 12.3%). These were less variable than nodule volume (mean = 1.2%; sd = 14.4%; LLR = −27.0%; ULR = 29.5%). All features had high repeatability (0.87 ≤ ICC ≤ 0.99), see Figure 1.

Conclusion The repeatability of CTTA was comparable to automatic volumetric measurements that are currently recommended for use in clinical practice. To our knowledge this is the first study to assess CTTA repeatability, a promising biomarker of malignancy.

Statistics from

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.