AIM: The comparison between chest x-ray (CXR) and computed tomography (CT) images is commonly required in clinical practice to assess the evolution of chest pathological manifestations. Intrinsic differences between the two techniques, however, limit reader confidence in such a comparison. CT average intensity projection (AIP) reconstruction allows obtaining "synthetic" CXR (s-CXR) images, which are thought to have the potential to increase the accuracy of comparison between CXR and CT imaging. We aim at assessing the diagnostic performance of s-CXR imaging in detecting common pleuro-parenchymal abnormalities. MATERIALS AND METHODS: 142 patients who underwent chest CT examination and CXR within 24 hours were enrolled. CT was the standard of reference. Both conventional CXR (cCXR) and s-CXR images were retrospectively reviewed for the presence of consolidation, nodule/mass, linear opacities, reticular opacities, and pleural effusion by 3 readers in two separate sessions. Sensitivity, specificity, accuracy and their 95% confidence interval were calculated for each reader and setting and tested by McNemar test. Inter-observer agreement was tested by Cohen's K test and its 95%CI. RESULTS: Overall, s-CXR sensitivity ranged 45-67% for consolidation, 12-28% for nodule/ mass, 17-33% for linear opacities, 2-61% for reticular opacities, and 33-58% for pleural effusion; specificity 65-83%, 83-94%, 94-98%, 93-100% and 79-86%; accuracy 66-68%, 74 -79%, 89-91%, 61-65% and 68-72%, respectively. K values ranged 0.38-0.50, 0.05-0.25, -0.05 -0.11, -0.01-0.15, and 0.40-0.66 for consolidation, nodule/mass, linear opacities, reticular opacities, and pleural effusion, respectively.

Coronary artery calcification (CAC) is a well-known cardiovascular risk factor and a reliable score to predict non-cancer survival. In the last year, the CAC score has been investigated in lung cancer (LC) screening, showing promising results in terms of mortality risk assessment. Nevertheless, its role in LC patients has still to be investigated. This study aims to evaluate the performance of a fully automated CAC scoring in predicting 5-year survival of patients who underwent surgical resection for stage I LC. This retrospective observational study included 536 consecutive patients with stage I LC who underwent preoperative chest CT with a 128+ slice CT scanner followed by surgical resection, between 2011 and 2022. The CAC score was measured by a commercially available, fully automated artificial intelligence (AI) software. CAC score was categorized into three validated risk categories: <100; 100-399; and ≥400. The primary outcome was the 5-year overall survival rate. A total of 110 (20.5%) patients had a CAC score ≥400, 149 (27.8%) of 100-399, and 277 (51.7%) <100. Male smokers had the highest CAC values: 32% (88/273) ≥400 and 36% (97/273) <100, while only 17% (5/29) of non-smoking males had CAC ≥400. Females had lower CAC values compared to males both in smokers and in non-smokers: only 10% (17/167) ≥400 in smoking females and 0% in non-smoking females. After a median follow-up of 3.7 years, the 5-year survival was 80.3% overall, 84.1% in CAC<100, 78.6% in CAC 100-399, and 73.3% in CAC >=400, with a statistically significant poorer outcome in patients with higher CAC (p=0.0072). We observed that CAC score was a risk factor associated with gender and smoking status and predicted the 5-year overall survival in patients with resected stage I LC. These results open new prospects for prevention of non-cancer mortality in early-stage LC patients.

Up to date, no predictive biomarkers can robustly identify patients with non-small cell lung cancer (NSCLC) who will benefit from immune checkpoint inhibitors (ICIs). Thus, we sought to non-invasively decode tumor-immune interactions implicated in ICI response by exploring the dynamic of blood immune-inflammatory markers and radiomic features in a cohort of advanced NSCLC treated with ICIs.

BackgroundClinically suitable biomarkers to foresee the response to immune checkpoint inhibitors (ICIs) can be achieved by decoding tumor heterogeneity and its evolution during treatment. Thus, we determine whether the longitudinal assessment of radiomic features (RFs) and blood hallmarks of systemic inflammation (SI) may predict ICI efficacy in advanced NSCLC.MethodsOn 92 stage IV NSCLC patients undergoing ICIs, CT-derived RFs and peripheral blood SI parameters, including derived Neutrophil-to-Lymphocyte ratio (dNLR) and lactate dehydrogenase (LDH), were acquired at baseline (T0) and at first disease assessment (T1). Primary endpoint was ICI response per RECIST. CR/PR or SD ≥ 6 months defined clinical benefit (CB) while SD < 6 months or PD non-responders (NR). T1 - T0 delta variations of 852 RFs and 6 SI indices were challenged into machine learning-based predictive models. RFs pre-processing included redundant features elimination and Z-score standardization; L2 penalized logistic regression with Monte-Carlo cross-validation was implemented for wrapper-based feature selection and model training/test. Resulting delta- radiomic (ΔR), immune/inflammatory (ΔI) and integrated (ΔInt) models were compared based on performance metrics.ResultsMedian OS and PFS were 8.1 (95%CI, 4.1-12.2) and 2.6 months (95% CI, 1.1-4.4), respectively; 34 (37%) patients belonged to CB while 58 (63%) were NR. Applying a model validation calibrated at up to 10 parameters, 5 delta-RFs (first- and higher-order classes) and delta-LDH were selected according to ROC-AUC scores and adopted for respective ΔR, ΔI and ΔInt models. Testing the predictive ability of our designed classifiers, ROC-AUC and accuracy (± St.Dev) were 0.86 ± 0.08 and 0.78 ± 0.08 for ΔR, while 0.78 ± 0.09 and 0.67 ± 0.09 for ΔI. The performance of ΔInt model in discriminating ICI response reached ROC-AUC of 0.88 ± 0.07 and accuracy of 0.82 ± 0.08 (P<0.001), thus overtaking that of individual models.ConclusionsWe developed a dynamic blood-radiomic predictor of ICI efficacy in advanced NSCLC suggesting that non-invasive interception of systemic and local events implicated in cancer evolution may implement current predictive models.Legal entity responsible for the studyUniversity Hospital of Parma.FundingHas not received any funding.DisclosureAll authors have declared no conflicts of interest. BackgroundClinically suitable biomarkers to foresee the response to immune checkpoint inhibitors (ICIs) can be achieved by decoding tumor heterogeneity and its evolution during treatment. Thus, we determine whether the longitudinal assessment of radiomic features (RFs) and blood hallmarks of systemic inflammation (SI) may predict ICI efficacy in advanced NSCLC. Clinically suitable biomarkers to foresee the response to immune checkpoint inhibitors (ICIs) can be achieved by decoding tumor heterogeneity and its evolution during treatment. Thus, we determine whether the longitudinal assessment of radiomic features (RFs) and blood hallmarks of systemic inflammation (SI) may predict ICI efficacy in advanced NSCLC. MethodsOn 92 stage IV NSCLC patients undergoing ICIs, CT-derived RFs and peripheral blood SI parameters, including derived Neutrophil-to-Lymphocyte ratio (dNLR) and lactate dehydrogenase (LDH), were acquired at baseline (T0) and at first disease assessment (T1). Primary endpoint was ICI response per RECIST. CR/PR or SD ≥ 6 months defined clinical benefit (CB) while SD < 6 months or PD non-responders (NR). T1 - T0 delta variations of 852 RFs and 6 SI indices were challenged into machine learning-based predictive models. RFs pre-processing included redundant features elimination and Z-score standardization; L2 penalized logistic regression with Monte-Carlo cross-validation was implemented for wrapper-based feature selection and model training/test. Resulting delta- radiomic (ΔR), immune/inflammatory (ΔI) and integrated (ΔInt) models were compared based on performance metrics. On 92 stage IV NSCLC patients undergoing ICIs, CT-derived RFs and peripheral blood SI parameters, including derived Neutrophil-to-Lymphocyte ratio (dNLR) and lactate dehydrogenase (LDH), were acquired at baseline (T0) and at first disease assessment (T1). Primary endpoint was ICI response per RECIST. CR/PR or SD ≥ 6 months defined clinical benefit (CB) while SD < 6 months or PD non-responders (NR). T1 - T0 delta variations of 852 RFs and 6 SI indices were challenged into machine learning-based predictive models. RFs pre-processing included redundant features elimination and Z-score standardization; L2 penalized logistic regression with Monte-Carlo cross-validation was implemented for wrapper-based feature selection and model training/test. Resulting delta- radiomic (ΔR), immune/inflammatory (ΔI) and integrated (ΔInt) models were compared based on performance metrics. ResultsMedian OS and PFS were 8.1 (95%CI, 4.1-12.2) and 2.6 months (95% CI, 1.1-4.4), respectively; 34 (37%) patients belonged to CB while 58 (63%) were NR. Applying a model validation calibrated at up to 10 parameters, 5 delta-RFs (first- and higher-order classes) and delta-LDH were selected according to ROC-AUC scores and adopted for respective ΔR, ΔI and ΔInt models. Testing the predictive ability of our designed classifiers, ROC-AUC and accuracy (± St.Dev) were 0.86 ± 0.08 and 0.78 ± 0.08 for ΔR, while 0.78 ± 0.09 and 0.67 ± 0.09 for ΔI. The performance of ΔInt model in discriminating ICI response reached ROC-AUC of 0.88 ± 0.07 and accuracy of 0.82 ± 0.08 (P<0.001), thus overtaking that of individual models. Median OS and PFS were 8.1 (95%CI, 4.1-12.2) and 2.6 months (95% CI, 1.1-4.4), respectively; 34 (37%) patients belonged to CB while 58 (63%) were NR. Applying a model validation calibrated at up to 10 parameters, 5 delta-RFs (first- and higher-order classes) and delta-LDH were selected according to ROC-AUC scores and adopted for respective ΔR, ΔI and ΔInt models. Testing the predictive ability of our designed classifiers, ROC-AUC and accuracy (± St.Dev) were 0.86 ± 0.08 and 0.78 ± 0.08 for ΔR, while 0.78 ± 0.09 and 0.67 ± 0.09 for ΔI. The performance of ΔInt model in discriminating ICI response reached ROC-AUC of 0.88 ± 0.07 and accuracy of 0.82 ± 0.08 (P<0.001), thus overtaking that of individual models. ConclusionsWe developed a dynamic blood-radiomic predictor of ICI efficacy in advanced NSCLC suggesting that non-invasive interception of systemic and local events implicated in cancer evolution may implement current predictive models. We developed a dynamic blood-radiomic predictor of ICI efficacy in advanced NSCLC suggesting that non-invasive interception of systemic and local events implicated in cancer evolution may implement current predictive models.