Lung Nodule Classification Using Biomarkers, Volumetric Radiomics, and 3D CNNs.

We present a hybrid algorithm to estimate lung nodule malignancy that combines imaging biomarkers from Radiologist's annotation with image classification of CT scans. Our algorithm employs a 3D Convolutional Neural Network (CNN) as well as a Random Forest in order to combine CT imagery with biomarker annotation and volumetric radiomic features. We analyze and compare the performance of the algorithm using only imagery, only biomarkers, combined imagery + biomarkers, combined imagery + volumetric radiomic features, and finally the combination of imagery + biomarkers + volumetric features in order to classify the suspicion level of nodule malignancy. The National Cancer Institute (NCI) Lung Image Database Consortium (LIDC) IDRI dataset is used to train and evaluate the classification task. We show that the incorporation of semi-supervised learning by means of K-Nearest-Neighbors (KNN) can increase the available training sample size of the LIDC-IDRI, thereby further improving the accuracy of malignancy estimation of most of the models tested although there is no significant improvement with the use of KNN semi-supervised learning if image classification with CNNs and volumetric features is combined with descriptive biomarkers. Unexpectedly, we also show that a model using image biomarkers alone is more accurate than one that combines biomarkers with volumetric radiomics, 3D CNNs, and semi-supervised learning. We discuss the possibility that this result may be influenced by cognitive bias in LIDC-IDRI because malignancy estimates were recorded by the same radiologist panel as biomarkers, as well as future work to incorporate pathology information over a subset of study participants.

Mid-gestational changes in cervicovaginal fluid cytokine levels in asymptomatic pregnant women are predictive markers of inflammation-associated spontaneous preterm birth.

OBJECTIVES: Perturbation of the choriodecidual space before the onset of spontaneous preterm birth (sPTB) could lead to a concomitant rise in both cervicovaginal fluid (CVF) cytokine and fetal fibronectin (FFN), and assessing the concentrations of both markers could improve the prediction of sPTB (delivery before 37 completed weeks of gestation). Therefore, we prospectively determined mid-trimester changes in CVF cytokine and FFN concentrations, and their predictive capacity for sPTB in asymptomatic pregnant women. STUDY DESIGN: CVF collected at 20+0-22+6 weeks (n = 47: Preterm-delivered = 22, Term-delivered = 25) and 26+0-28+6 weeks (n = 50: Preterm-delivered = 17, Term-delivered = 33) from 63 asymptomatic pregnant women at risk of sPTB were examined. Cytokine and FFN concentrations were determined by multiplexed bead-based immunoassay and 10Q Rapid analysis (Hologic, MA, USA) respectively. The 20+0-22+6/26+0-28+6 weeks ratios of cytokines and FFN concentrations were compared between preterm- and term-delivered women using Receiver Operating Characteristics curves to predict sPTB. Also, bacterial 16S rDNA from 64 samples (20+0-22+6 weeks n = 36, 26+0-28+6 weeks n = 28) was amplified by polymerase chain reaction to determine associations between vaginal microflora, cytokine and FFN concentrations. RESULTS: Changes in RANTES and IL-1ß concentrations between 20+0-22+6 and 26+0-28+6 weeks, expressed as a ratios, were predictive of sPTB, RANTES (AUC = 0.82, CI = 0.62-0.94) more so than IL-1ß (AUC = 0.71, CI = 0.53-0.85) and FFN (not predictive). Combining these markers (AUC = 0.83, CI = 0.63-0.95) showed similar predictive capacity as RANTES alone. FFN concentrations at 26+0-28+6 weeks correlated with IL-1ß (r = 0.4, P = 0.002) and RANTES (r = 0.3, P = 0.03). In addition, there was increased prevalence of vaginal anaerobes including Bacteroides, Fusobacterium and Mobiluncus between gestational time points in women who experienced sPTB compared to the term women (P = 0.0006). CONCLUSIONS: CVF RANTES and IL-1ß in mid-trimester of pregnancy correlate with quantitative FFN. The levels of CVF RANTES and IL-1ß decline significantly in women who deliver at term unlike women who deliver preterm. This observation suggests that sPTB may be characterised by sustained choriodecidual inflammation and may have clinical value in serial screening for sPTB if confirmed by larger studies.