Abstract
PurposeThe purpose of my thesis was to assess the use of texture features derived from 2-[fluorine-18] fluoro-2-deoxy-d-glucose (18F-FDG) images in thoracic cancers. To achieve this, two studies were undertaken.
1. The aim of the first study was to determine retrospectively if texture features derived from 18F-FDG positron emission tomography/computed tomography (18F-FDG PET/CT) images of malignant pleural mesothelioma (MPM) were associated with overall survival in a cohort of patients scanned in our institution.
2. The aim of the second study was to correlate prospectively texture parameters from 18F-FDG PET/CT images of untreated non-small-cell lung cancer (NSCLC) with histological and immunohistochemical (IHC) parameters in order to obtain a better understanding of the biological factors that may be related to spatial heterogeneity of 18F-FDG PET images.
Methods
1. Fifty-eight consecutive patients (mean age 64.4 years, 51 male) with MPM, investigated between January 2006 and December 2011, were included in the first study. Patients with previous pleurodesis were excluded as this can cause significant benign inflammatory 18F-FDG uptake. 18F-FDG PET/CT scans were processed and analysed using a standard protocol. Calculation of the texture features was performed using in-house software implemented with MATLAB (MathWorks, Natick, Mass, US). Texture features, standardised uptake values (SUVs), metabolic tumour volume (MTV) and total lesion glycolysis (TLG) were derived from volume of interest (VOI) of the MPMs. Cox regression analysis was used to examine the effects of the PET parameters and other variables on survival outcomes.
2. Nineteen consecutive patients (mean age 70.5 years, 10 male) with histologically proven, ≥3cm NSCLC planned for surgery and undergoing 18F-FDG PET/CT imaging were recruited prospectively. Calculation of the texture features was performed using in-house software implemented with MATLAB (MathWorks, Natick, Mass, US). The lobectomy specimens were marked such that its orientation within the body was known. Histology included markers of angiogenesis, hypoxia, glycolysis and proliferation. As data were not normally distributed on Shapiro-Wilk testing, Spearman rank correlation was used to assess correlations between eighteen 18F-FDG PET derived texture parameters and 6 IHC stains.
Results
1. Univariable analysis indicated several variables including non-epithelioid histology (hazard ratio (HR) 2.13 (confidence interval (CI) 1.11-4.08)), log-TLG (HR 1.33 (CI 1.07-1.67)), first-order entropy (HR 1.61 (CI 1.02-2.56)) and first-order energy (HR 0.68 (CI 0.48-0.96)) were significantly associated with patient survival (p<0.05). Multivariable analysis showed that first-order entropy (HR 1.75 (CI 1.07-2.89)) was an independent predictor of patient survival.
2. All patients underwent imaging a median of 1 day before surgery (range: 1 day to 44 days). There were 10 adenocarcinomas (ADC), 8 squamous cell carcinomas (SCC) and 1 large cell neuroendocrine carcinoma (LCNC). Group 1 (all 19 patients): CD105 microvascular density (MVD), staining neovessel endothelial cells, correlated negatively with TLG and first-order energy; and positively with neighbourhood grey tone difference matrices (NGTDM) coarseness (r = -0.51, -0.47, 0.53, respectively: all p<0.05). CD34 MVD, staining vascular and lymphatic endothelial cells, correlated negatively with MTV and TLG; and positively with NGTDM coarseness (r = -0.59, -0.50, 0.62, respectively: all p<0.05). Ki67average (Ki67avg) and Ki67maximum (Ki67max) values correlated negatively with grey level co-occurrence matrix (GLCM) energy and first-order skewness (r = -0.47 and r = -0.48, respectively: all p<0.05).
Group 2 (18 patients with ADC and SCC): CD105 MVD correlated negatively with TLG, first-order energy and first-order entropy; and positively with NGTDM coarseness (r = -0.53, -0.52, -0.47, 0.53, respectively: all p<0.05). CD34 MVD correlated negatively with MTV and TLG; and positively with NGTDM coarseness (r = -0.61, -0.52, 0.64, respectively: all p<0.05).
Group 3 (10 ADC patients): CD105 MVD and CD34 MVD correlated with both MTV (r = -0.71, -0.76, respectively: all p<0.05) and NGTDM coarseness (r= 0.75, 0.77, respectively: all p<0.05).
Group 4 (8 SCC patients): CD105 MVD and CD34 MVD correlated negatively with first-order energy (r = -0.79, -0.74, respectively: all p<0.05).
Hypoxia-inducible factor-1 (HIF-1)-alpha correlated strongly with MTV, second-order textural parameters (GLCM energy, GLCM homogeneity and GLCM entropy) and high-order textural parameters (NGTDM coarseness and NGTDM contrast) (r = 0.73, 0.71, 0.83, -0.81, -0.73, -0.73, respectively: all p<0.05).
Hexokinase-II (HEX-II) correlated strongly with MTV, second-order textural parameter (GLCM contrast) and high-order textural parameter (NGTDM coarseness) (r = -0.72, 0.73, 0.72, respectively: all p<0.05).
If the correction for multiple correlation testing was to be applied, the only statistically significant correlation was between CD34 MVD and high-order coarseness (p = 0.004), in Groups 1 and 2.
Conclusion
1. Textural features have prognostic ability in predicting survival in MPM patients. This is superior to the currently used standard PET parameters such as SUVs. In particular, first-order entropy is significantly associated with overall survival in MPM.
2. Several standard and textural parameters extracted from 18F-FDG images of NSCLC correlate strongly with MVD, Ki67, HIF-1-alpha and HEX-II histological parameters suggesting relevant underlying biological mechanisms are associated with 18F-FDG distribution in tumours. My study has also uncovered interesting differences in PET texture correlations with histological subtypes in NSCLC; with only limited data in current literature my study would add value to it.
| Date of Award | 1 Nov 2020 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Gary Cook (Supervisor) & Vicky Goh (Supervisor) |