A Study on Dual-source CT Imagings Combined with CT Signs to Established a Nomographic Model for Calculate the Risk Classification of GIST
Objective The main aim of this study is to evaluate the effectiveness of using dual-source CT imaging features and CT signs to grade the risk of gastrointestinal stromal tumors(GIST)with a nomographic model,in order to improve precision in clinical treatment and prognosis assessment of GIST.Methods A total of 135 patients who were pathologically diagnosed with GIST in our hospital and had complete preoperative energy spectrum CT images were collected and included in the study,and 73 hospital-confirmed cases were collected as external validation.135 cases in our hospital were used as the training group,and 73 cases in other hospitals were used as the validation group.The independent risk factors were trained and validated,and the independent risk factors of the energy spectrum CT parameters,CT signs prediction model and radiomics prediction model were selected for multi-factor logic.Regression,retaining features with P＜0.05 for multivariate logistic regression modeling and obtaining nomograms.Results The values of single-energy CT iodine concentration,water concentration,and effective atomic number were computed using Dual-source CT results.Additionally,the normalized iodine concentration(NIC)and the gradient of the energy spectrum curves were also determined.Other imaging features were evaluated,such as tumor size,growth pattern,tumor necrosis/ulceration,tumor feeding or draining vascular proliferation(EVFDM),tumor outline and adjacent tissue infiltration.During the univariate analysis,it was observed that at 70keV,the single energy value fell in between 40keV and 140keV.This particular energy level resulted in lower image noise,a higher signal-to-noise ratio(SNR),and statistically significant differences in adjacent tissue invasion between the low-risk and high-risk groups(P＜0.05).Additionally,multivariate logistic regression analysis indicated that tumor size,along with other factors,played a significant role.tumor necrosis/ulcer,EVFDM,tumor contour,and adjacent tissue invasion(presence vs absence)were significantly associated with high malignant potential.After incorporating these independent factors into the nomogram model,the c-index for the training and validation groups were 0.872(95％CI:0.753-0.863)and 0.807(95％CI:0.697-0.893).respectively,The sensitivity was 0.865,the specificity was 0.915,and the diagnostic accuracy was 0.837.The AUC of energy spectrum CT imaging combined with CT feature map were 0.927(training group)and 0.905(validation group),showing good prediction.Conclusion The nomogram of gastrointestinal stromal tumors composed of different KeV imaging features of spectral CT combined with CT signs,including size,location,tumor necrosis/ulcer,EVFDM,tumor outline and adjacent tissue infiltration,etc.The malignant potential of gastrointestinal stromal tumors can be used to accurately forecast the primary tumor,thus providing useful aid in clinical treatment and assessing patient prognosis.
NETL
NSTL
万方数据
