Discovery Radiomics via Deep Multi-Column Radiomic Sequencers for Skin Cancer Detection

摘要

While skin cancer is the most diagnosed form of cancer in men and women, withmore cases diagnosed each year than all other cancers combined, sufficientlyearly diagnosis results in very good prognosis and as such makes earlydetection crucial. While radiomics have shown considerable promise as apowerful diagnostic tool for significantly improving oncological diagnosticaccuracy and efficiency, current radiomics-driven methods have largely rely onpre-defined, hand-crafted quantitative features, which can greatly limit theability to fully characterize unique cancer phenotype that distinguish it fromhealthy tissue. Recently, the notion of discovery radiomics was introduced,where a large amount of custom, quantitative radiomic features are directlydiscovered from the wealth of readily available medical imaging data. In thisstudy, we present a novel discovery radiomics framework for skin cancerdetection, where we leverage novel deep multi-column radiomic sequencers forhigh-throughput discovery and extraction of a large amount of custom radiomicfeatures tailored for characterizing unique skin cancer tissue phenotype. Thediscovered radiomic sequencer was tested against 9,152 biopsy-proven clinicalimages comprising of different skin cancers such as melanoma and basal cellcarcinoma, and demonstrated sensitivity and specificity of 91% and 75%,respectively, thus achieving dermatologist-level performance and reak hencecan be a powerful tool for assisting general practitioners and dermatologistsalike in improving the efficiency, consistency, and accuracy of skin cancerdiagnosis.