Radiomics as a personalized medicine tool in lung cancer: Separating the hope from the hype

Abstract

Radiomics has become a popular image analysis method in the last few years. Its key hypothesis is that medical images harbor biological, prognostic and predictive information that is not revealed upon visual inspection. In contrast to previous work with a priori defined imaging biomarkers, radiomics instead calculates image features at scale and uses statistical methods to identify those most strongly associated to outcome. This builds on years of research into computer aided diagnosis and pattern recognition. While the potential of radiomics to aid personalized medicine is widely recognized, several technical limitations exist which hinder biomarker translation. Aspects of the radiomic workflow lack repeatability or reproducibility under particular circumstances, which is a key requirement for the translation of imaging biomarkers into clinical practice. One of the most commonly studied uses of radiomics is for personalized medicine applications in Non-Small Cell Lung Cancer (NSCLC). In this review, we summarize reported methodological limitations in CT based radiomic analyses together with suggested solutions. We then evaluate the current NSCLC radiomics literature to assess the risk associated with accepting the published conclusions with respect to these limitations. We review different complementary scoring systems and initiatives that can be used to critically appraise data from radiomics studies. Wider awareness should improve the quality of ongoing and future radiomics studies and advance their potential as clinically relevant biomarkers for personalized medicine in patients with NSCLC.

Understanding Sources of Variation to Improve the Reproducibility of Radiomics

“…Indeed, a fast-growing literature shows the great promise of radiomics signatures (radiomics features and models) as a "virtual biopsy" to assist in cancer diagnosis and prognosis, treatment plan, patient stratification, and assessment of tumor response to therapy. The current status of CT-based radiomics in lung cancer has been well summarized in a recent collection of review articles [e.g., (6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)].…”

Vulnerabilitäten bei Radiomics: Warum die populärste Radiomics-Signatur unbemerkt das Tumorvolumen gemessen hat

“…Fornacon-Wood et al. [ 9 ] hatten bei 70 % der identifizierten NSCLC-Radiomics-Arbeiten ≥ 6 methodische Einschränkungen identifiziert bei einem medianen Radiomics Quality Score (RQS) von 6 (mögliche Spannbreite von −8 bis 36). Die Tatsache, dass 36 % der untersuchten Arbeiten lediglich eine interne und vor allem 13 % gar keine Validierung durchgeführt haben, muss vermutlich als Indiz dafür gesehen werden, dass noch ein allgemein unzureichendes Bewusstsein für das „overfitting“ von Maschinenlernmodellen oder „validation data leakage“ auch bei Peer-Reviewern und Editoren bestand [ 9 , 10 ].…”

Positron Emission Tomography for Response Evaluation in Microenvironment-Targeted Anti-Cancer Therapy

“…On the basis of considering tumor heterogeneity, texture analysis has been explored, especially in the field of nuclear medicine [ 58 , 59 , 60 , 61 ]. The most exciting part of machine learning in medical imaging would be to extract patterns that are beyond human perception and classification due to the application of deep learning for diagnostic algorithms [ 62 , 63 ]. Radiologists should seek to work alongside AI in the future.…”