The Study Of ¹⁸F-FDG PET Metabolic Spatial Heterogeneity Analysis In Lung Cancer

Lung cancer is one of the most common malignant tumors in the world,and its incidence is increasing year by year.The mortality rate of lung cancer is the highest in malignant tumor death.The main types of lung cancer include non-small cell lung cancer（NSCLC）and small cell lung cancer（SCLC）,of which NSCLC accounts for about 80%-85%of all,and the early diagnosis and the appropriate treatment can significantly improve the prognosis of NSCLC.Malignant tumors have significant heterogeneity,and their biological heterogeneity is often related to their cellular or molecular characteristics,such as cell proliferation,necrosis,fibrosis,blood flow and angiogenesis,cell metabolism,hypoxia and the expression of certain specific receptors.¹⁸F-fluorodeoxyglucose(¹⁸F-FDG)is an analog of glucose.As an agent for PET imaging,the heterogeneity of its uptake in the lesion is also related to cell growth patterns,blood supply,hypoxia,necrosis and other factors.It can provide important information for the diagnosis of occurrence,disease development,evaluation of treatment efficacy and prognosis of the disease.Positron emission tomography and x-ray computed tomography（PET/CT）imaging can provide information including tumor size,location,morphology,texture.PET/CT also can provide focal metabolic differences in terms of molecular level,and its diagnostic value in malignant tumors,especially in lung cancer,has been widely accepted clinically.Usually,the maximum standardizd uptake value（SUVmax）≥2.5 is used as a criterion for diagnosing malignant nodules.However,the SUV is not a specific diagnostic indicator for lung cancer,which might be affected by many factors.There are some benign nodules such as tuberculosis and other granulomatous nodules with higher glucose uptake.In these cases,the differential diagnosis of benign hypermetabolic diseases and malignant pulmonary nodules is difficult by using SUVmax≥2.5.Therefore,it is crucial to establish a new method to reduce the false positive rate.Delayed imaging based on time-to-peak differences in the uptake of FDG by benign and malignant lesions has become a routine procedure for PET image acquisition,however,its value reported for distinguishing the benign and malignant lesions is inconsistent.It has been recognized that standard medical image images contain far more information of lesions than routinely used by clinicians,but the ability to distinguish them with the naked eye is limited and sometimes cannot be interpreted.¹⁸F-FDG PET/CT imaging can demonstrate the temporal and spatial distribution of metabolic activity in tumor lesions,which is called heterogeneity analysis of PET/CT imaging.The difference of heterogeneous distribution of this metabolism is thought to be related to tumor phenotype and tumor microenvironmental changes.In recent years,many studies had shown that by the heterogeneity analysis of PET/CT imaging of lung cancer patients,it is possible to analyze the spatial distribution differences within the same tumor lesion and the metabolic differences of different tumor lesions.By analyzing the heterogeneity of PET/CT images,more post-treatment response and prognosis information of tumors can be obtained than conventional image analysis and thus to further achieve the goal of“precise medical treatment”.Our previous study found that the spatial distribution of the maximum metabolic threshold of 80%and 90%of tumors in benign and malignant solitary pulmonary nodules is significantly different.The diagnostic value of this difference for differentiating benign and malignant pulmonary nodules is significantly higher than that of conventional methods.Thus,we speculate that the metabolic distribution in lung malignancies has a specific spatial heterogeneity.However,previous studies have only been based on the local metabolic distribution of solitary pulmonary nodules,and lack of analysis of the metabolic spatial distribution of gross lung tumors and tumors as a whole.Therefore,in the first part of this paper,we intend to perform a three-dimensional spatial analysis of the metabolic heterogeneity distribution of NSCLC,to determine the the difference in the spatial distribution of metabolism in benign and malignant lesions of the lung,to seek its distribution characteristic,thus,to explore its diagnostic value for the lung malignant tumors.Furtherly,based on the results of the first part of this study and based on the differences in the growth of benign and malignant nodules,in the second part of this paper,we intend to use the simple and easy-to-operate visual analysis to interpreting the spatial differences in the metabolic distribution of pulmonary nodules and masses with high ¹⁸F-FDG uptake.The value of this visual analysis applied for differentiating hypermetabolic nodules and masses was then evaluated.In the third part of this paper,the heterogeneity analysis of PET/CT imaging was performed retrospectively for preoperative lung cancer patients.The metabolic heterogeneity parameter was established and its correlation with lung cancer pathological parameters was analyzed.This study can provide important biological information for the diagnosis and treatment of lung cancer.Part 1.¹⁸F-FDG PET metabolic spatial heterogeneity quantitative analysis for the diagnosis of non-small cell lung cancerPurpose:This study aimed to:1)quantitatively analyze the metastatic spatial heterogeneity of benign and malignant lung nodules;divide the malignant nodules into two groups:metastatic and non-metastatic,and further analyze their heterogeneity;evaluate the diagnostic value of metabolic spatial heterogeneity quantitative analysis for NSCLC;2)calculate the diagnostic value of SUVmax≥2.5,RI>0 and PET/CT for NSCLC respectively;3)analyze and compare the diagnostic efficacy of these four different diagnostic methods for pulmonary nodules,plot the receiver operating characteristic curve（ROC）and compare its area under curve（AUC）,and confirm the diagnostic value of metabolic spatial heterogeneity quantitative analysis for pulmonary nodules.Method:From January 2012 to August 2016,367 patients with pulmonary nodules who underwent ¹⁸F-FDG PET/CT imaging were studied retrospectively.Among them,134patients were undergone delayed imaging for further confirmation.The nodules were diagnosed as malignant or benign ones by pathological or histological results and clinical follow-up.The malignant nodules were divided into non-metastaitc and metastatic groups.The spatial heterogeneities of ¹⁸F-FDG metabolism of all nodules were quantitatively analyzed.The age,nodule diameter,SUVmax,and RI_max of patients with benign or malignant nodules were compared.The mean metabolic volume of malignant and benign lesions were plotted and compared when 40%,50%,60%,70%,80%and 90%of SUVmax was set as the threshold,so did the malignant lesions with or without metastasis.The ipsilateral hilar angle was taken as the reference point;the maximum distance from the ROI boundary of each metabolic threshold to the ipsilateral hilar angle was measured and calculated.Linear regression analysis was performed on the distance values of different metabolic thresholds,and the corresponding linear slope k value was calculated.The absolute value of k represented the spatial heterogeneity of the lesion.The larger the k value,the higher the spatial heterogeneity,and vice versa.The diagnostic performance like sensitivity,specificity,accuracy,positive predictive value（PPV）and NPV（negative predictive value）of SUVmax,RI,PET/CT and the metabolic spatial heterogeneity quantitative analysis for NSCLC were compared.The ROC curves were plotted and the AUC were compared for these four diffenrent diagnosing methods.Result:Of the 367 pulmonary nodules,252 were malignant and 115 were benign ones.The maligant nodules were divided into metastatic（129 cases）and non-metastatic（123 cases）groups.The average lesion diameter was 28.06±14.84 mm;the mean SUVmax value was7.69±5.60.A total of 134 patients underwent delayed imaging in 367 patients,including 42patients with benign lesions and 92 patients with malignant lesions.The age,the lesion diameter and metabolic value of the nodules in malignant group were significantly higher than those of the benign group（P<0.001）,but there was no significant difference of RI between the two groups（P=0.113）.In the malignant lesions,the SUVmax value of the metastatic group was significantly higher than that in the non-metastatic group（P<0.001）,but there was no significant difference of RI between the two groups（P=0.274）.The mean metabolic volume of 40%,50%,60%and 70%threshold of SUVmax was statistically significant different between benign and malignant group,and the mean metabolic volume of 40%,50%,60%,70%and 80%threshold of SUVmax between non-metastatic and metastatic group was statistically significant different.Linear regression analysis was performed on the distance between metabolic volumes with different threshold of the malignant lesions to the ipsilateral hilar angle（D40%-90%）.The mean slope k values of the two groups were 1.4002±0.6807 and 2.2848±1.5311,respectively.The difference between the two groups was statistically significant（P<0.001）.The mean k values of benign,non-metastatic and metastatic nodules were1.4002±0.6807,1.8636±0.95169 and 2.6864±1.8441,respectively,which indicated that the spatial heterogeneity of metabolism in the three groups increased sequentially,and the difference between the three groups was statistically significant（P<0.001）.It also indicated that compared with benign lung lesions,the high metabolic area of lung malignant tumors tended to be distributed to the hilum side.When SUVmax≥2.5 was set for diagnosing lung cancer,the sensitivity,specificity,accuracy,PPV and NPV were 88.89%,6.96%,63.22%,67.67%and 22.22%,respectively.When RI>0 was used as the positive index,the sensitivity,specificity,accuracy,PPV and NPV of the diagnosis of NSCLC were 86.96%,19.05%,65.67%,70.18%and 40%,respectively.The sensitivity,specificity,accuracy,PPV and NPV of PET/CT diagnosis of lung cancer were 93.25%,38.26%,76.02%,76.80%and 72.13%,respectively.There was a significant difference in the spatial heterogeneity of metabolism in benign and malignant lung lesions.The differential diagnosis of benign and malignant lung lesions was performed using this kind of difference.The diagnostic cutoff value was set as 1.3553 and the sensitivity,specificity,accuracy,PPV and NPV were 72.62%,64.35%,70.03%,81.70%and 51.75%,respectively.The AUCs under the ROC curves of each diagnostic method were:the AUC of SUVmax≥2.5 was 0.474（P=0.693）,the AUC of RI>0 was 0.597（P=0.073）,the AUC was0.623（P=0.022）for PET/CT diagnosis and the AUC of this metabolic spatial heterogeneity distribution parameter-the k value was 0.723（P<0.001）.Moreover,the AUC of the k value was significantly higher than the other three diagnostic methods,and the differences were significant（all P<0.05）.Conclusions:1.There was significant difference of metabolic heterogeneity between benign and malignant nodules,and between metastatic and non-metastatic malignant nodules;According to the results,the high¹⁸F-FDG metabolic area of lung malignant lesions tends to be distributed to the hilum side.2.Compared with the methods of SUVmax≥2.5,RI>0 and PET/CT,the metabolic spatial heterogeneity quantitative analysis had the largest area under the diagnostic curve of NSCLC,and would be very useful in diagnosing pulmonary nodules.Part 2.Diagnostic value of ¹⁸F-FDG PET metabolic spatial heterogeneity visual analysis for undetermined pulmonary nodules and masses with high ¹⁸F-FDG uptakePurpose:The aim of this study was to:1)to further clarify the characteristics and differences of spatial heterogeneity distribution of ¹⁸F-FDG metabolism between benign and malignant lesions;2)to evaluate the diagnostic value of ¹⁸F-FDG metabolic spatial heterogeneity distribution visual analysis-a simple and easy-to-operate method for evaluating high-metabolic and undetermined lung nodules and masses.Methods:From January 1,2015 to April 26,2017,301 patients（186 males and 115 females;mean age 61.6±11.2 years;age range 21-89 years）with undertemined pulmonary nodules or masses who undergone ¹⁸F-FDG PET/CT imaging were studied retrospectively.Visual analysis of spatial heterogeneity distribution of ¹⁸F-FDG metabolism was interpreted independently by two nuclear medicine physicians with more that 10-year PET/CT diagnostic experience.When the scores given by the two physicians were inconsistent,consensus will be reached through discussion.The proximal region of the lesion was defined as the region near the ipsilateral hilum,and the distal region was defined as the region remote from the ipsilateral hilum.The spatial distribution characteristics of¹⁸F-FDG metabolism were analyzed using a 5-point scoring system.The score of lesion≤2was judged to be negative,and the score≥3 was judged to be positive.The ROC curve was plotted to evaluate the diagnostic performance of SUVmax,RImax,PET/CT,and¹⁸F-FDG metabolic spatial heterogeneity visual analysis for high-metabolic undetermined pulmonary nodules and masses.The sensitivity,specificity,accuracy,PPV and NPV of each diagnostic method were calculated respectively.The Mc Nemar test was used to compare the sensitivity,specificity,and accuracy of ¹⁸F-FDG metabolic spatial heterogeneity visual analysis with other methods.The intra-observer variability of the visual score was determined by using Cohen’s kappa coefficient（κ）.P value<0.05 was considered to be statistically significant.Results:Of all 301 lung nodules and masses,194 were malignant（64.5%）and 107 were benign（35.5%）.The mean maximum diameter of the lesion was 27.7±13.2 mm（range,8-63 mm）.Among the malignant lesions,180（92.8%）patients’score was≥3,indicating that the ¹⁸F-FDG metabolic distribution in the proximal region of malignant lesions was significantly higher than that in the distal region.Among the benign lesions,78（72.9%）patients’score was≤2,indicating that the metabolic distribution of ¹⁸F-FDG in the proximal region of benign lesions was significantly lower than that in the distal region.Cohen’s kappa coefficient showed good agreement between the visual scores of the two observers with a kappa value of 0.635.The sensitivity,specificity,accuracy,PPV,NPV of SUVmax,RI,PET/CT and metabolic spatial heterogeneity distribution visual diagnosis for lung malignant tumors were 100%,0%,64.5%,64.5%,0%;95.3%26.2%,70.8%,71.2%,78.0%;95.9%,61.7%,83.7%,81.9%,83.7%and 92.8%,72.9%,85.7%,86.1%,89.2%respectively.In terms of diagnostic sensitivity,metabolic spatial heterogeneity visual analysis was not significantly different from other traditional diagnostic methods（p>0.05）,but its specificity（72.9%）was significantly higher than SUVmax（0%）and RI 26.2%（Both P<0.001）.ROCs for the diagnosis of benign and malignant lung undetermined lesions were drawn for these three diagnostic methods.The AUC for SUVmax,RI,PET/CT,and metabolic spatial heterogeneity visual analysis were 0.626,0.670,0.788,and 0.886,respectively.The AUC of metabolic spatial heterogeneity visual analysis was significantly greater than SUVmax,RI and PET/CT（all P<0.05）.Conclusions:1.The spatial heterogeneity distribution of ¹⁸F-FDG metabolism in benign and malignant lung lesions is different.The high metabolic region of lung malignant lesions tends to be distributed on the hilar side,while the high metabolic region of benign lung lesions is distributed to the hilar side.2.¹⁸F-FDG metabolic spatial heterogeneity visual analysis is simple and easy-to-operate,and it can significantly improve the diagnostic specificity for high metabolic lung nodules and mass diagnosis.It can be used as a new auxiliary diagnostic analysis method clinically.Part 3.The association between ¹⁸F-FDG PET intratumoral metabolic heterogeneity and pathological parameters in non-small cell lung cancerPurpose:The aim of this study was to evaluate the association between intratumoral metabolic heterogeneity on ¹⁸F-FDG PET imaging and the pathological type,differentiation and T stage of NSCLC.Methods:234 consecutive patients who were performed preoperative ¹⁸F-FDG PET/CT and pathologically diagnosed as NSCLC were assessed retrospectively.Metabolic heterogeneity factor（MHF）derived from the volume-threshold function from 40%to 90%was calculated for primary lesions.The association between MHF and the pathological type,differentiation and T stage of NSCLC were analysed,respectively.Results:In 236 primary lung cancers,there were 69 of squamous carcinoma（SCC）,133 of adenocarcinoma（ADC）,26 of adenosquamous carcinoma（ASC）and 8 of large cell lung carcinoma（LCC）.The difference of MHF value among these four pathological types was statistically significant（F=17.611,P=0.000）,and the MHF value of SCC was significantly higher than that of other pathological types（P<0.01）.Undifferentiated,poor,middle and high differentiated groups had statistically significant difference in MHF value（F=23.932,P=0.000）,and with the reduction of the grades of tumor,MHF value increased.The difference of MHF value among T1,T2,T3 groups was also statistically significant（F=11.082,P=0.000）,and with the increase of T stages,MHF value increased.Linear-by-linear association analysis revealed strong correlations between MHF and pathological grades and T stages（P=0.000,0.032）.Moreover,multivariate ordinal logistic regression analysis showed that MHF was independently associated with tumor T stage and differentiation after adjusting for clinical factors.Conclusion:MHF showed high association with pathological parameters,which can be used to accurately predict the biological behaviors of tumors from the molecular level thus to provide important biological information for tumor diagnosis and treatment.