Radiomics in precision medicine for lung cancer

Abstract

With the improvement of external radiotherapy delivery accuracy, such as intensity-modulated and stereotactic body radiation therapy, radiation oncology has recently entered in the era of precision medicine. Despite these precise irradiation modalities, lung cancers remain one of the most aggressive human cancers worldwide, possibly because of diverse genotypic alterations that drive and maintain lung tumorigenesis. It has been long recognized that imaging could aid in the diagnosis, tumor delineation, and monitoring of lung cancer. Moreover, accumulating evidence suggests that imaging information could be further used to tailor treatment type and intensity, as well as predict treatment outcomes in radiotherapy. However, these imaging tasks have been carried out either qualitatively or using simplistic metrics that doesn't take advantage of the full scale of imaging knowledge. Radiomics, which is a recent field of research that aims to provide a more quantitative representation of imaging information relating tumor phenotypes to clinical and genotypic endpoints by embedding extracted image features into predictive mathematical models. These predictive models can be a key component in the clinician decision making and treatment personalization. This review provides an overview of the radiomics application and its methodology for radiation oncology studies in lung cancer.

From Handcrafted to Deep-Learning-Based Cancer Radiomics: Challenges and Opportunities

“…While Reference [18] has briefly touched upon the deep learning pipeline as an emerging technology that can extract Radiomics features, it has not studied applicability of different deep architectures [20] and left the interpretability topic untouched. Furthermore, Reference [21] has mostly focused on the handcrafted Radiomics, while deep learning-based Radiomics is explained briefly without addressing different architectures, interpretability, and hybrid models. Although both types of Radiomics are covered in Reference [22], combination of hand-crafted and deep learning-based features are not considered.…”

Novel Quantitative PET Techniques for Clinical Decision Support in Oncology

“…Radiomics could be thought of as consisting of following two main procedures: (i) the extraction of quantitative imaging (static and dynamic features) from a previously defined tumor region(s) and (ii) the incorporation of the imaging features or traits into mathematical models for treatment outcome prediction that is aimed at providing added value for personalizing of treatment regimens in comparison with commonly used clinical predictors; this is illustrated in Figure 4. [69][70][71] Although the notion of radiomics traces its origin into quantitative imaging analysis in the areas of computer-aided detection or diagnosis in the 1980s, 72 its application for clinical and biological endpoints has started in the past decade only. This has been driven by recent advances in personalized/precision medicine.…”

“…Moreover, the combination of PET/CT was shown to predict local control in nonsmall cell lung cancer, 77 while the combination of PET/MRI was shown to improve prediction of metastasis to the lung in sarcoma. 77 Reviews of these applications are provided in the works of Constanzo et al and Avanzo et al 71,78 The features extracted from PET images (radiomics) could be derived via direct extraction (handcrafted) or indirectly using deep learning methodology. The latter is still in its infancy and we will focus in this review on the direct feature extraction methods, which has witnessed tremendous growth and success over the past decade.…”

Arcabouço para Classificação, Recuperação por Conteúdo e Radiômica de Imagens Médicas: uma investigação de biomarcadores quantitativos para o câncer de pulmão

“…Diversos algoritmos foram criados com o objetivo de reduzir a dimensionalidade espacial do vetor. Basicamente, estes métodos podem ser classificados em três tipos principais: filtro, wrapper e embarcado (CONSTANZO et al, 2017;HALL, 2011). Os filtros avaliam a relevância dos atributos por meio de heurísticas baseadas em características dos dados, como correlação e informação mútua (PENG; LONG; DING, 2005;HALL, 1999;KONONENKO, 1994).…”

“…Para isso, ele tenta maximizar a distância dos vetores de atributos das amostras para o hiperplano, ou seja, aumentando a margem de decisão entre as classes (Figura 25) (KAVZOGLU; COLKESEN, 2009). Figura 24: Mecanismo de criação de hiperplano do método SVM para classificação.Fonte:Constanzo et al (2017).…”