Radiomics in cervical cancer: Current applications and future potential

Highlights

• Early diagnosis will greatly improve the treatment outcome for patients with cervical cancer.

• Radiomics has the clinical value in predicting characteristics of patients with cervical cancer.

• Some issues had to be addressed before the application of radiomics in clinical practice.

• Radiomics can help improve the diagnosis and prognosis of patients with cervical cancer.

Abstract

Cervical cancer is the most commonly diagnosed cancer among women. Early diagnosis and prediction will greatly improve the treatment outcome. Many clinical parameters have been used as diagnostic and prognostic factors for cervical cancer patients, including tumor stage, histological type, lymph node status, but with limitations in prediction accuracy. The development of noninvasive biomarker with the potential to provide more specific tumor characterization before treatment begins or during therapy is urgent needed, which may permit clinicians to administer a more individualized anti-cancer treatment. Radiomics is a mathematical-statistical procedure extracting information from medial images, which has the potential for prediction of staging, histological type, node status, relapse and survival in patients with cervical cancer. In this manuscript, we reviewed recent clinical studies and future potential for the application of radiomics in the treatment of patients with cervical cancer, and discussed the current challenges and limitations of radiomics for oncology.

Introduction

Cervical cancer is the fourth most common causes of death from cancer in the developed countries and the second most common causes of cancer-related death in the developing countries (Torre et al., 2015). Although the incidence of cervical cancer continues to decline in the world, especially in the developed countries, the occurrence in young female is still on the rise (Lim et al., 2013). The treatment strategies for patients with cervical cancer are varied according to different tumor stages and nodal status. Patients with early stages and locally confined tumor (International Federation of Gynecology and Obstetrics classification [FIGO] stages IB1 and IIA1) may be treated by surgical resection alone, while patients with locally advanced or node-positive tumors may require combined radiochemotherapy (Koh et al., 2013). Although outcomes had been improved with multimodality treatment, cervical cancer remains a devastating disease with 5-year overall survival (OS) for FIGO stages II, III, and IVA is still only 65%, 40% and 15%, respectively (Green et al., 2001). About 30–40% of patients with cervical cancer will suffer from relapse during the treatment (Breuneval et al., 2015), especially for patients with advanced disease (Klopp and Eifel, 2012).

Early diagnosis will greatly improve the treatment outcome for patients with cervical cancer. Various biomarkers have been identified as prognostic factors in cervical cancer patients, including FIGO stage, histology, tumor volume, lymph node metastasis (LNM), and single-gene markers (Klopp and Eifel, 2012; Katanyoo et al., 2001; Kristensen et al., 1999; Rose et al., 2015; Xue et al., 2006; Prescott et al., 2010). There is a potential for clinicians to adjust therapeutic strategies in time to improve outcomes if we are able to predict the tumor stage, lymph node status (LNS), histological type, survival and response to treatment accurately. Currently, magnetic resonance imaging (MRI) and 18F-fluorodeoxyglucose (FDG) positron emission tomography/computed tomography (PET/CT) have been widely used for staging (Sala et al., 2010; Balleyguier et al., 2011), treatment decision making (Herrera, 2013), and prognosis evaluation in cervical cancer (Barwick et al., 2013; Bae et al., 2016; Heo et al., 2013). Nevertheless, both modalities have their limitations and inadequateness (Vincens et al., 2008; Meads et al., 2013).

Radiomics refers to a variety of mathematical methods that extracts a large number of features from medical images, including MRI (Sala et al., 2010; Balleyguier et al., 2011; Bonekamp et al., 2018; Li et al., 2018a), CT (Lubner et al., 2017; Krafft et al., 2018), ultrasound (US) (Hu et al., 2018; Yao et al., 2018), single-photon emission computed tomography (SPECT) (Rajkumar et al., 2015; Rahmim et al., 2017) and PET/CT (Oikonomou et al., 2018; Lv et al., 2018). Extracted texture features could be used as surrogate markers for underlying gene expression patterns as well as relevant biological tumor traits such as tumor morphology, and intratumor heterogeneity (Aerts et al., 2014; Lambin et al., 2012). Through investigating the relationship between texture features and biologic tumor characteristics, more valuable information could be extracted from medical images for staging, node status, and treatment response prediction for cancer patients (Ganeshan et al., 2012; Vallières et al., 2015; Huang et al., 2016; Tixier et al., 2011; Chicklore et al., 2013).Usually, a radiomic study design and workflow include (I) Image segmentation, (II) Radiomic feature extraction, (III) Dimension reduction and feature selection, (IV) Statistics analysis and model building (Fig. 1) (Gillies et al., 2016). Recently, a number of studies have been reported on application of radiomics in patients with cervical cancer (Jin et al., 2020; Liu et al., 2019; Lian et al., 2016; Bowen et al., 2018; Wu et al., 2018). In this manuscript, we mainly focus on the clinical value, current evidence, and future potential of radiomics in predicting tumor stage, histological type, LNM, survival, recurrence and distant metastasis for patients with cervical cancer.