Metallomic Classification\nof Pulmonary Nodules Using\nBlood by Deep-Learning-Boosted Synchrotron Radiation X‑ray\nFluorescence

Ambient air pollution is an important contributor to\nincreasing\ncases of lung cancer, which is a malignant cancer with the highest\nmortality among all cancers. It primarily manifests in the form of\npulmonary nodules, but not all will develop into lung cancer. Therefore,\nit is highly desired to distinguish between benign and malignant pulmonary\nnodules for the early prevention and treatment of lung cancer. Currently,\nhistopathological examination is the gold standard for classifying\npulmonary nodules, which is invasive, time-consuming, and labor-intensive.\nThis study proposes a metallomics approach through synchrotron radiation\nX-ray fluorescence (SRXRF) with a simplified one-dimensional convolutional\nneural network (1DCNN) to distinguish pulmonary nodules by using serum\nsamples. SRXRF spectra of serum samples were obtained and preliminarily\nanalyzed using principal component analysis (PCA). Subsequently, machine\nlearning algorithms (MLs) and 1DCNN were applied to develop classification\nmodels. Both MLs and 1DCNN based on full-channel spectra could distinguish\npatients with benign and malignant pulmonary nodules, but the highest\naccuracy rate of 96.7% was achieved when using 1DCNN. In addition,\nit was found that characteristic elements in serum from patients with\nmalignant nodules were different from those in benign nodules, which\ncan serve as the fingerprint metallome profile. The simplified model\nbased on characteristic elements resulted in good performance of sensitivity\nand F1-score > 91.30%, G-mean, MCC and Kappa > 85.59%, and accuracy\n= 94.34%. In summary, metallomic classification of benign and malignant\npulmonary nodules using serum samples can be achieved through 1DCNN-boosted\nSRXRF, which is easy to handle and much less invasive compared to\nhistopathological examination.