PatientsSPN sufferers have been collected retrospectively at the Department of Thoracic Surgery of Jiangsu Cancer Hospital from September 2021 to March 2023. The inclusion standards of this examine topics have been as follows: ① benign and malignant SPNs confirmed by histopathology after surgical procedure or biopsy and ② preoperative DLCT chest enhancement examination. The exclusion standards have been as follows: ① pure floor glass or subsolid nodules (containing floor glass elements) (n = 11); ② a historical past of malignant tumors (n = 5); ③ poor CT picture high quality as a consequence of the lack of steady thin-layer photographs or artifacts (n = 7); ④ the quantity of pulmonary nodules is a couple of or major nodules with a number of scattered lesions (n = 13); ⑤ a historical past of scientific antitumor remedy (n = 16); ⑥ lesions smaller than 10 mm (n = 8) and ⑦ nodules with poorly outlined boundaries resulting in poor segmentation (n = 4). After screening sufferers in response to the above standards, 250 sufferers with SPN have been finally enrolled in this examine for follow-up evaluation.This retrospective examine was accredited by the Ethics Committee of Jiangsu Cancer Hospital (ethics quantity: 2023-048), and the experiment was performed in strict accordance with the moral requirements set out in the 1964 Declaration of Helsinki and its subsequent amendments. The Ethics Committee of Jiangsu Cancer Hospital waived the written knowledgeable consent of the sufferers.DLCT picture acquisitionAll sufferers with SPN have been examined by DLCT (IQon, Philips Healthcare, Best, The Netherlands), and respiration coaching was carried out on every affected person earlier than scanning. With the sufferers mendacity supine on the scanning desk, scan from the thoracic inlet to the backside of the chest to cowl all lung tissue. The distinction agent (ioversol, iodine 350 mg/mL, Hengrui Medicine, Lianyungang, China) was injected into the proper elbow vein, after which the tube was flushed with 20 ml of regular saline. The injection fee was in the vary of 2.5–3.0 ml/s, and the picture acquisition throughout the enhancement interval was delayed by 50 s after the injection was accomplished. The scanning slice thickness of all photographs was 5 mm and the reconstructed slice thickness was 1 mm. Other parameters have been as follows: matrix, 512 × 512; collimator width, 64 × 0.625 mm; tube present automated modulation; rotation time, 0.50 s; tube voltage, 120 kVp; scanning subject of view, 372 mm; pitch, 0.900.DLCT picture quantitative featuresAll photographs have been processed and analyzed on the Philips workstation (IntelliSpace Portal, Philips Healthcare). A radiologist used the workstation’s built-in software program (Spectral CT Viewer, Philips Healthcare) to delineate the round areas of curiosity (ROIs) in the mediastinal window photographs and carry out quantitative evaluation, whereas one other senior radiologist supervised from the sidelines. Before the evaluation, neither physician was knowledgeable of the scientific information and the pathological diagnosis outcomes of benign and malignant SPN. Two examples (one SPN benign and the different SPN malignant) are proven in Fig. 3. ROIs ought to cowl the areas of lesions with uniform density on the enhanced photographs to the biggest attainable extent, avoiding calcification, blood vessels and necrotic areas. To guarantee the stability of the outcomes, the ROIs have been drawn on the largest layer of the lesion cross-section and the layers above and beneath it, and the common of the three measurements was taken as the remaining evaluation information. At the similar time, a round ROI was positioned on the aorta with the largest cross-section of the lesion for the standardization of quantitative parameters. Then, different quantitative parameters of the similar ROI have been obtained on the VNC photographs, Zeff photographs, IC photographs, ED photographs, CaS photographs, and 40 keV and 70 keV monoenergetic photographs.Figure 3A 41-year-old male affected person with inflammatory pseudotumor in the higher lobe of the left lung (a–d) and a 64-year-old male affected person with lung adenocarcinoma (e–h), each pathologically confirmed. a and e are enhanced part CT photographs, and the CT values of the lesions are 73.4 HU and 86.9 HU respectively; (b, f) are Zeff photographs with Zeff values of 8.11 and eight.45 respectively; (c, g) are iodine photographs, and the IC values are 1.42 mg/ml and a pair of.17 mg/ml respectively; (d, h) are postoperative pathological photos.The quantitative parameters obtained from the ROIs delineated on the lesions and aorta have been as follows: CTSPN_VNC, CTSPN_40 keV, CTSPN_70 keV, CT values of aorta on VNC photographs (CTaorta_VNC) and CT values of 70 keV after aortic enhancement (CTaorta). Derived parameters have been calculated primarily based on the above values, together with SARVNC, SAR40 keV, SAR70 keV, Δ40 keV, Δ70 keV, ΔSA_40 keV, ΔSA_70 keV, CER40 keV, CER70 keV, NEF40 keV, NEF70 keV and λHU. The calculation formulation of the above derived parameters have been as follows30,31,32,33,34:$${textual content{SAR}}_{{{textual content{VNC}}}} = {textual content{ CT}}_{{{textual content{SPN}}_{textual content{VNC}}}} /{textual content{CT}}_{{{textual content{aorta}}_{textual content{VNC}}}}$$$${textual content{SAR}}_{{{4}0;{textual content{keV}}}} = {textual content{ CT}}_{{{textual content{SPN}}_{4}0;{textual content{keV}}}} /{textual content{CT}}_{{{textual content{aorta}}}}$$$${textual content{SAR}}_{{{7}0;{textual content{keV}}}} = {textual content{CT}}_{{{textual content{SPN}}_{7}0;{textual content{keV}}}} /{textual content{CT}}_{{{textual content{aorta}}}}$$$$Delta_{{{4}0;{textual content{keV}}}} = {textual content{ CT}}_{{{textual content{SPN}}_{4}0;{textual content{keV}}}} {-}{textual content{CT}}_{{{textual content{SPN}}_{textual content{VNC}}}}$$$$Delta_{{{7}0;{textual content{keV}}}} = {textual content{CT}}_{{{textual content{SPN}}_{7}0;{textual content{keV}}}} {-}{textual content{CT}}_{{{textual content{SPN}}_{textual content{VNC}}}}$$$$Delta_{{{textual content{SA}}_{4}0;{textual content{keV}}}} = {textual content{CT}}_{{{textual content{SPN}}_{4}0;{textual content{keV}}}} {-}{textual content{CT}}_{{{textual content{aorta}}}}$$$$Delta_{{{textual content{SA}}_{7}0;{textual content{keV}}}} = {textual content{CT}}_{{{textual content{SPN}}_{7}0;{textual content{keV}}}} {-}{textual content{CT}}_{{{textual content{aorta}}}}$$$${textual content{CER}}_{{{4}0;{textual content{keV}}}} = Delta_{{{4}0;{textual content{keV}}}} /{textual content{CT}}_{{{textual content{SPN}}_{textual content{VNC}}}}$$$${textual content{CER}}_{{{7}0;{textual content{keV}}}} = Delta_{{{7}0;{textual content{keV}}}} /{textual content{CT}}_{{{textual content{SPN}}_{textual content{VNC}}}}$$$${textual content{NEF}}_{{{4}0;{textual content{keV}}}} = Delta_{{{4}0;{textual content{keV}}}} /left( {{textual content{CT}}_{{{textual content{aorta}}}} {-}{textual content{CT}}_{{{textual content{aorta}}_{textual content{VNC}}}} } proper)$$$${textual content{NEF}}_{{{7}0;{textual content{keV}}}} = Delta_{{{7}0;{textual content{keV}}}} /left( {{textual content{CT}}_{{{textual content{aorta}}}} {-}{textual content{CT}}_{{{textual content{aorta}}_{textual content{VNC}}}} } proper)$$$$uplambda _{{{textual content{HU}}}} = left( {{textual content{CT}}_{{{textual content{SPN}}_{4}0;{textual content{keV}}}} {-}{textual content{CT}}_{{{textual content{SPN}}_{7}0;{textual content{keV}}}} } proper)/left( {{7}0{-}{4}0} proper)$$Considering the variations in cardiac perform and hemodynamics between sufferers, the IC routinely measured on the iodine picture was normalized to the aorta, the NIC was calculated, and the NCaS, NED and NZeff have been calculated in the similar approach. The calculation formulation have been as follows:$${textual content{NIC}} = {textual content{IC}}_{{{textual content{SPN}}}} /{textual content{IC}}_{{{textual content{aorta}}}}$$$${textual content{NCaS}} = {textual content{CaS}}_{{{textual content{SPN}}}} /{textual content{CaS}}_{{{textual content{aorta}}}}$$$${textual content{NED}} = {textual content{ED}}_{{{textual content{SPN}}}} /{textual content{ED}}_{{{textual content{aorta}}}}$$$${textual content{NZeff}} = {textual content{Zeff}}_{{{textual content{SPN}}}} /{textual content{Zeff}}_{{{textual content{aorta}}}}$$Machine learning modelAll SPN sufferers have been randomly divided into the coaching and check units by 7꞉3 and the sufferers in the coaching set have been balanced at a ratio of 1:1 by utilizing the artificial minority oversampling method, in order that the quantity of benign SPN sufferers and malignant SPN sufferers was consistent35,36. The LASSO algorithm was used to display options from sufferers’ scientific information and DLCT parameters, and 6 classical ML models have been constructed in the coaching set primarily based on the chosen options, specifically AdaBoost, GNB, LR, RF, SVM and XGBoost37. The ROC curves of the 6 models have been plotted to acquire the AUC of the respective models, and the accuracy, sensitivity and specificity have been calculated concurrently to function the analysis indicators for the 6 models. The models have been evaluated by tenfold cross-validation, and the efficiency of the models was additional validated utilizing check set information. The above modeling course of was primarily based on R software program (model 4.2.3) and Python programming language (model 3.11.4).Statistical evaluationThe Shapiro–Wilk check was employed to evaluate the regular distribution of the information. The impartial pattern t-test was utilized for evaluating the options of steady information with a traditional distribution, whereas the Mann–Whitney U check was used for evaluating the options of steady information with a non-normal distribution. The chi-square check or Fisher’s precise check was utilized to check the traits of rely information. The statistical evaluation of sufferers’ scientific information and DLCT parameters was performed utilizing SPSS Statistics 26.0 (IBM Corp., Chicago, Illinois, United States of America) software program, with statistical significance indicated by P < 0.05.
https://www.nature.com/articles/s41598-024-55280-6
Dual-layer detector spectral CT-based machine learning models in the differential diagnosis of solitary pulmonary nodules
