Predicting adenocarcinoma recurrence using computational texture models of nodule components in lung CT

Abstract

Purpose: To investigate the importance of presurgical computed tomography (CT) intensity and texture information from ground-glass opacities (GGO) and solid nodule components for the prediction of adenocarcinoma recurrence. Methods: For this study, 101 patients with surgically resected stage I adenocarcinoma were selected. During the follow-up period, 17 patients had disease recurrence with six associated cancer-related deaths. GGO and solid tumor components were delineated on presurgical CT scans by a radiologist. Computational texture models of GGO and solid regions were built using linear combinations of steerable Riesz wavelets learned with linear support vector machines (SVMs). Unlike other traditional texture attributes, the proposed texture models are designed to encode local image scales and directions that are specific to GGO and solid tissue. The responses of the locally steered models were used as texture attributes and compared to the responses of unaligned Riesz wavelets. The texture attributes were combined with CT intensities to predict tumor recurrence and patient hazard according to disease-free survival (DFS) time. Two families of predictive models were compared: LASSO and SVMs, and their survival counterparts: Cox-LASSO and survival SVMs.Results: The best-performing predictive model of patient hazard was associated with a concordance index (C-index) of 0.81 ± 0.02 and was based on the combination of the steered models and CT intensities with survival SVMs. The same feature group and the LASSO model yielded the highest area under the receiver operating characteristic curve (AUC) of 0.8 ± 0.01 for predicting tumor recurrence, although no statistically significant difference was found when compared to using intensity features solely. For all models, the performance was found to be significantly higher when image attributes were based on the solid components solely versus using the entire tumors (p < 3.08 × 10 −5 ). Conclusions: This study constitutes a novel perspective on how to interpret imaging information from CT examinations by suggesting that most of the information related to adenocarcinoma aggressiveness is related to the intensity and morphological properties of solid components of the tumor. The prediction of adenocarcinoma relapse was found to have low specificity but very high sensitivity. Our results could be useful in clinical practice to identify patients for which no recurrence is expected with a very high confidence using a presurgical CT scan only. It also provided an accurate estimation of the risk of recurrence after a given duration t from surgical resection (i.e., C-index = 0.81 ± 0.02). C 2015 American Association of Physicists in Medicine. [http://dx

A 3-D Riesz-Covariance Texture Model for Prediction of Nodule Recurrence in Lung CT

“…This is consistent with other approaches in the literature using similar descriptor paradigms and validates the choice of the 3-D Riesz-covariance descriptors for particular medical imaging applications that can benefit from their 3-D, spatial and rotation invariance. The importance of preserving spatial co-variations between features is highlighted by large predictive performance improvements between covariance-versus average-based feature aggregation in Table I A, B versus C. The results obtained with the average (see Table I C) are consistent with our previous study where solid components of the tumor yielded best results [18]. However, thanks to the covariance-based aggregation, the use of larger ROIs including heterogeneous tissue architectures (e.g., GTV and GGO) improves the performance of the predictive models (see Table I A and B), which was not the case when using feature averages over the GTV or GGO (see Table I C).…”

“…(4), without the mapping to any tangent space. Finally, a linear SVM using the mean of the features obtained within each delineated region: although using the average of a feature sampling set as an aggregation function can filter out salient feature values we provide this experiment as a straightforward step from our work presented in [18], where the average of 2-D features on the CT slice of maximum diameter was used. We believe the better performance Table I.…”

“…It relies on the statistical notion of the covariance of feature observations for modeling texture distributions within regions of interest of arbitrary size. When compared to our previous study based on 2-D Riesz texture models of nodule recurrence (see [18]), the main novelty of the current approach is its natural extension to the 3-D domain via the spatial relation of 3-D Riesz-wavelet features through the covariancebased descriptors. The previous approach used the axial CT with the maximum nodule surface for classification, where the responses of the Riesz filters were aggregated based on the average.…”

“…Mattonen et al [17] separated the analysis of tumor components, although focusing on GGO regions only. In previous work, we specifically learned texture models of solid and GGO nodule components to predict nodule recurrence [18]. The models were based on linear combinations of 2-D steerable Riesz-wavelets learned using support vector machines (SVM).…”

Early-Stage Non–Small Cell Lung Cancer: Quantitative Imaging Characteristics of18F Fluorodeoxyglucose PET/CT Allow Prediction of Distant Metastasis

“…The rationale is that, by extracting a large number of putative imaging features, we could obtain a more comprehensive characterization of the underlying tumor phenotypes, which may ultimately correlate with clinical outcomes. This approach has been used to predict overall survival in patients with lung cancer with widely available imaging data such as CT (11,12) or FDG PET (13,14).…”