RSNA 2016

Abstract Archives of the RSNA, 2016


IN255-SD-THA6

Reproducibility of CT Texture Parameters by Leveraging Publically Available Patient Imaging Datasets

Thursday, Dec. 1 12:15PM - 12:45PM Room: IN Community, Learning Center Station #6



Shih-Hsin Chen, London, United Kingdom (Abstract Co-Author) Nothing to Disclose
Balaji Ganeshan, PhD, London, United Kingdom (Presenter) CEO, TexRAD Ltd; Director, Feedback plc; Director, Stone Checker Software Ltd; Director, Prostate Checker Ltd
Francesco Fraioli, MD, London, United Kingdom (Abstract Co-Author) Nothing to Disclose
PURPOSE

Heterogeneity is a key component of malignancy and computed tomography texture analysis (CTTA) is one possible way to quantify it non-invasively. This study aims to use a test-retest methodology to assess the reproducibility of the filtration-histogram technique of CTTA.

METHOD AND MATERIALS

The RIDER Lung CT dataset without the injection of contrast media (source The Cancer Imaging Archive – TCIA, sponsored by the Cancer Imaging Program, DCTD/NCI/NIH) was used for this purpose. For each patient, two readers drew the regions of interest (ROI) on the slice with the largest cross-section of the tumor. One operator analyzed the image data twice with a one-week interval between the two analyses. Filtration-histogram based CTTA was undertaken using TexRAD commercial research software (TexRAD Ltd, www.texrad.com - part of Feedback Plc, Cambridge, UK). The agreement of the texture parameters from CTTA analysis of repeated scans, for the same operator (intra-operator) and between the two operators (inter-operator) was assessed by intra-class correlation coefficients (ICC) and further visualized using Bland-Altman plots.

RESULTS

Our results showed that most of the filtration-histogram based CT texture parameters are generally reproducible with high ICC for the different permutations. The best ICC for the different texture quantifiers for intra-operator, inter-operator and repeated scan are as follows: Mean (0.968, 0.909, 0.787), SD (0.973, 0.960, 0.838), Entropy (0.998, 0.994, 0.993), Skewness (0.925, 0.862, 0.782), Kurtosis (0.942, 0.885, 0.689) and MPP (0.958, 0.942, 0.882). Bland-Altman plots further provided the average differences between the operators and limits of agreement for the different texture metrics, which are further useful in a clinical setting.

CONCLUSION

The filtration-histogram technique of CTTA is reproducible. This study using a test-retest methodology forms a vital part in the qualification process of CTTA as a potential imaging biomarker and its translation into clinic.

CLINICAL RELEVANCE/APPLICATION

The reproducibility of the filtration-histogram technique of CTTA suggests its potential application in routine clinical and multi-center setting.