A novel representation of inter-site tumour heterogeneity from pre-treatment computed tomography textures classifies ovarian cancers by clinical outcome

Abstract

Purpose: To evaluate the associations between clinical outcomes and radiomics-derived inter-site spatial heterogeneity metrics across multiple metastatic lesions on CT in patients with high-grade serous ovarian cancer (HGSOC).

Methods: IRB-approved retrospective study of 38 HGSOC patients. All sites of suspected HGSOC involvement on preoperative CT were manually segmented. Gray-level correlation matrix-based textures were computed from each tumour site, and grouped into five clusters using a Gaussian Mixture Model. Pairwise inter-site similarities were computed, generating an inter-site similarity matrix (ISM). Inter-site texture heterogeneity metrics were computed from the ISM and compared to clinical outcomes.

Results: Of the 12 inter-site texture heterogeneity metrics evaluated, those capturing the differences in texture similarities across sites were associated with shorter overall survival (inter-site similarity entropy, similarity level cluster shade, and inter-site similarity level cluster prominence; p ≤ 0.05) and incomplete surgical resection (similarity level cluster shade, inter-site similarity level cluster prominence and inter-site cluster variance; p ≤ 0.05). Neither the total number of disease sites per patient nor the overall tumour volume per patient was associated with overall survival. Amplification of 19q12 involving cyclin E1 gene (CCNE1) predominantly occurred in patients with more heterogeneous inter-site textures.

Conclusion: Quantitative metrics non-invasively capturing spatial inter-site heterogeneity may predict outcomes in patients with HGSOC.

Key points: • Calculating inter-site texture-based heterogeneity metrics was feasible • Metrics capturing texture similarities across HGSOC sites were associated with overall survival • Heterogeneity metrics were also associated with incomplete surgical resection of HGSOC.

Keywords: Ovarian cancer; Radiogenomics; Radiomics; Survival; Texture.

Conflict of interest statement

Conflict of interest:

The authors of this manuscript declare no relationships with any companies, whose products or services may be related to the subject matter of the article.

Figures

Figure 1

(A) Patient with Classification of Ovarian Cancer (CLOVAR) mesenchymal subtype and an overall survival of 69 months. Texture-based results within each tumor site (i), the inter-site similarity matrix (ISM) (ii), the heterogeneity tree (iii) and the schematic of the dissimilarity of the various sites compared to the ovarian mass (iv). For instance, the diaphragmatic tumor implant (#8 in ii) has the largest dissimilarity compared to the ovarian mass. (B) Patient with CLOVAR mesenchymal subtype and an overall survival of 10 months. Texture-based results within each tumor site (i), the ISM (ii), the heterogeneity tree (iii) and the schematic of the dissimilarity of the various sites compared to the ovarian mass (iv). For instance, the left upper quadrant (LUQ) (#5 in b) has the largest dissimilarity compared to the ovarian mass, followed by the diaphragmatic and the omentum tumor implants. The numbers listed in the x axis of both figures indicates the numerical codes for lesion location (1=primary ovarian mass, 2=cul de sac, 4=omentum, 5=left upper quadrant, 7=gastro-hepatic ligament, 8=diaphragm).

Figure 1

(A) Patient with Classification of Ovarian Cancer (CLOVAR) mesenchymal subtype and an overall survival of 69 months. Texture-based results within each tumor site (i), the inter-site similarity matrix (ISM) (ii), the heterogeneity tree (iii) and the schematic of the dissimilarity of the various sites compared to the ovarian mass (iv). For instance, the diaphragmatic tumor implant (#8 in ii) has the largest dissimilarity compared to the ovarian mass. (B) Patient with CLOVAR mesenchymal subtype and an overall survival of 10 months. Texture-based results within each tumor site (i), the ISM (ii), the heterogeneity tree (iii) and the schematic of the dissimilarity of the various sites compared to the ovarian mass (iv). For instance, the left upper quadrant (LUQ) (#5 in b) has the largest dissimilarity compared to the ovarian mass, followed by the diaphragmatic and the omentum tumor implants. The numbers listed in the x axis of both figures indicates the numerical codes for lesion location (1=primary ovarian mass, 2=cul de sac, 4=omentum, 5=left upper quadrant, 7=gastro-hepatic ligament, 8=diaphragm).

Figure 2

Relationship between (A) the total number of disease sites per patient, inter-site texture heterogeneity metrics and 60 month survival and (B) the overall tumor volume per patient (calculated from CT as the sum of the volume from each segmented individual tumor site per patient), inter-site texture heterogeneity metrics and 60 month survival.

Figure 2

Relationship between (A) the total number of disease sites per patient, inter-site texture heterogeneity metrics and 60 month survival and (B) the overall tumor volume per patient (calculated from CT as the sum of the volume from each segmented individual tumor site per patient), inter-site texture heterogeneity metrics and 60 month survival.

Figure 3

Relation between Inter-site tumor Texture Heterogeneity Metrics and Genomic Alterations. (A) Heat map showing the relative similarities of the patients with respect to each other computed using the inter-site texture heterogeneity metrics. Similarity between two patients was computed using Euclidean distance measure after standardizing the features using z-score normalization. Two patients that are highly similar have a value close to 0 (red). (B) Patients were divided into two groups; the high risk group was defined as patients who underwent incomplete resection or had CCNE1 amplification and had an overall survival of ≤60 months. Conversely low risk group consisted of patients who underwent complete resection and had a survival of >60 months and no CCNE1 amplification. The SCP, SE and SCV values for each patient in the low and high risk categories according to the clinical variables and the model (classifier) are shown.

Figure 3

Relation between Inter-site tumor Texture Heterogeneity Metrics and Genomic Alterations. (A) Heat map showing the relative similarities of the patients with respect to each other computed using the inter-site texture heterogeneity metrics. Similarity between two patients was computed using Euclidean distance measure after standardizing the features using z-score normalization. Two patients that are highly similar have a value close to 0 (red). (B) Patients were divided into two groups; the high risk group was defined as patients who underwent incomplete resection or had CCNE1 amplification and had an overall survival of ≤60 months. Conversely low risk group consisted of patients who underwent complete resection and had a survival of >60 months and no CCNE1 amplification. The SCP, SE and SCV values for each patient in the low and high risk categories according to the clinical variables and the model (classifier) are shown.