Delta-radiomics features for the prediction of patient outcomes in non-small cell lung cancer.

Fave X, Zhang L, Yang J, Mackin D, Balter P, Gomez D, Followill D, Jones AK, Stingo F, Liao Z, Mohan R, Court L. · 2017

Read Delta-radiomics features for the prediction of patient outcomes in non-small cell lung cancer. by Fave X, Zhang L, Yang J, Mackin D, Balter P, Gomez D, Followill D, Jones AK, Stingo F, Liao Z, Mohan R, Court L. (2017) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc5428827.txt.

Introduction

Fave and co-authors test whether tumor image features change during radiotherapy and predict outcomes. Their study appeared in Scientific Reports. Radiomics extracts numbers from medical scans. Those numbers describe tumor shape and texture. Baseline features already predict survival in several cancers. Nobody had tracked their change on CT during lung treatment. The team followed 107 patients. All had stage III non-small cell lung cancer. Scans came before treatment and weekly during it.

Every radiomics feature changed significantly with dose. The team built models for three outcomes. Outcomes were survival, distant spread, and local recurrence. Baseline compactness improved survival prediction. It also improved prediction of distant metastases. Baseline features said nothing about local recurrence. Texture strength at treatment's end split recurrence risk instead. High and low risk groups separated cleanly. End-of-treatment values may flag tumor response. Such flags could guide added therapy.

Catalog record

CREATORFave X, Zhang L, Yang J, Mackin D, Balter P, Gomez D, Followill D, Jones AK, Stingo F, Liao Z, Mohan R, Court L.
YEAR2017
COLLECTIONPAPERS
FORMATTXT
FILE SIZE46 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC5428827

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Delta-radiomics features for the prediction of patient outcomes in non–small cell lung cancer By Xenia Fave, Lifei Zhang, Jinzhong Yang, Dennis Mackin, Peter Balter, Daniel Gomez, David Followill, Aaron Kyle Jones, Francesco Stingo, Zhongxing Liao, Radhe Mohan, Laurence Court Published in: Scientific Reports DOI: 10.1038/s41598-017-00665-z ABSTRACT Abstract Radiomics is the use of quantitative imaging features extracted from medical images to characterize tumor pathology or heterogeneity. Features measured at pretreatment have successfully predicted patient outcomes in numerous cancer sites. This project was designed to determine whether radiomics features measured from non–small cell lung cancer (NSCLC) change during therapy and whether those features (delta-radiomics features) can improve prognostic models. Features were calculated from pretreatment and weekly intra-treatment computed tomography images for 107 patients with stage III NSCLC. Pretreatment images were used to determine feature-specific image preprocessing. Linear mixed-effects models were used to identify features that changed significantly with dose-fraction. Multivariate models were built for overall survival, distant metastases, and local recurrence using only clinical factors, clinical factors and pretreatment radiomics features, and clinical factors, pretreatment radiomics features, and delta-radiomics features. All of the radiomics features changed significantly during radiation therapy. For overall survival and distant metastases, pretreatment compactness improved the c-index. For local recurrence, pretreatment imaging features were not prognostic, while texture-strength measured at the end of treatment significantly stratified high- and low-risk patients. These results suggest radiomics features change due to radiation therapy and their values at the end of treatment may be indicators of tumor response. INTRODUCTION Lung cancer is responsible for the largest number of cancer deaths in both men and women in the United States1. Over 85% of lung cancer cases are non-small cell lung cancers (NSCLC)1, 2. Predicting a particular NSCLC patient’s response to treatment, even compared to patients in the same disease stage group, is extremely difficult. A model that could effectively identify patients whose tumors are not responding to treatment would be beneficial and could be used to recommend patients for adjuvant chemotherapy or a radiation boost. Radiomics is the extraction of quantitative imaging features from medical images. These quantitative values can be used to develop models for cancer diagnosis, patient prognosis, or relative tumor heterogeneity that can then guide clinical decisions3, 4. This process is similar to the current application of tumor stage or genetic information derived from tumor biopsy specimens for clinical decision making. Radiomics has the combined advantages of being highly patient-specific and non-invasive. Additionally, unlike biopsy specimens, radiomics allows for sampling the heterogeneity over the entire tumor. In recent years, numerous studies have examined the potential clinical utility of radiomics features calculated from computed tomography (CT) images of NSCLC. These studies have identified features that are linked to tumor histology5, 6, tumor stage7, patient overall survival8–15, and genetic mutations16–18. Changes in radiomics features, called delta-radiomics features, have also been studied for their prognostic potential in cancer. Delta-radiomics features have been successful in predicting the response of colorectal cancer liver metastases19 and metastatic renal cell cancer20 to chemotherapies. Delta-radiomics features have also been used to identify patients with esophageal cancer who would develop radiation pneumonitis during treatment21. Delta-radiomics features calculated from PET images were shown to be predictive of overall survival for NSCLC patients22. To our knowledge, however, no published reports have investigated the possibility of using delta-radiomics features calculated from CT images for determining prognosis in NSCLC patients. Currently, NSCLC tumor response is analyzed with the Response Evaluation Criteria in Solid Tumors (RECIST) guidelines23–25. These guidelines depend on changes in tumor size to evaluate tumor response. Tumor size is widely known to be correlated with survival and probability for distant metastases in NSCLC. However, it does not reflect changes in tumor heterogeneity or genetic profiles, both of which may be more indicative of individual tumor biology. By sampling the entire tumor and analyzing changes in the spatial variations in intensity, delta-radiomics features may fill this gap and provide better patient-specific outcome predictions. The main objective of this work was to determine whether therapy-induced changes in radiomics features, called delta-radiomics features, can improve models for predicting patient outcome when used in conjunction with clinical factors and radiomics features measured prior to treatment. METHODS Patient data For this study, we retrospectively reviewed the images and medical records for 137 NSCLC patients with a waiver of informed consent from the Institutional Review Board (IRB) at the University of Texas MD Anderson Cancer center. These patients were selected because they had been enrolled on an IRB approved clinical trial at the University of Texas MD Anderson Cancer Center where they were imaged weekly with a four-dimensional CT (4DCT) during their treatment26. Informed consent had been obtained from each study participant. All methods for the trial were performed in accordance with the University of Texas MD Anderson Cancer Center IRB guidelines and regulations and all experimental protocols were approved by the same IRB. Inclusion criteria for the trial were pathologically proven advanced NSCLC, Karnofsky performance status (KPS) ≥70 or ECOG score 0–1, forced expiratory volume ≥1 liter, and age between 18 and 85 years26. Exclusion criteria included small cell histology, prior radiotherapy to the planned radiation therapy fields, pregnancy, or oxygen dependence due to pre-existing lung disease26. The patients were treated with radiation therapy and concurrent chemotherapy. They had been randomized to receive treatment with either photons or protons to 66 or 74 Gy. Full trial details are available online26. In this retrospective analysis, treatment modality was not considered for classification purposes. For this retrospective analysis, the medical records of these patients were reviewed to determine their clinical factors: sex, age, smoking status, pack years, tumor histology, overall disease stage, T stage, N stage, KPS, and total prescribed radiation dose (Table 1). The primary endpoints for this retrospective analysis of the trial data were overall survival, freedom from distant metastases, and local-regional control. [Table: Clinical characteristics of the NSCLC patient population used for modeling in this analysis.] Landmark analysis Survival studies using measures of response that are calculated at multiple time points, such as the radiomics features measured at weekly intervals in this study, require a landmark time point to be used for calculating the time until the endpoint is reached27, 28. Otherwise a bias can be introduced by responders since they must have already survived to the time of treatment to be classified as responders27, 28. In this study, patients were classified as high or low risk using multivariate models that included clinical factors (recorded at the time of entrance to the study), pre-treatment radiomics features (measured from the treatment planning images), and delta-radiomics features (measured from the different weekly images through treatment). Because a variable number of days occurred for each patient between when they entered the trial and when their pre-treatment images were acquired and between their pre-treatment images and last weekly images, a landmark time point was required to measure survival. For this study, endpoints were defined from a landmark time point of 90 days from the day the patient was entered on the clinical trial until one of the endpoints of death, presence of distant metastases, or local-regional failure was met. Patients not reaching the endpoint were censored at their last follow-up date. The landmark point was calculated by determining the

Read the complete text