Prediction of Tumor PD-L1 Expression in Resectable Non-Small Cell Lung Cancer by Machine Learning Models Based on Clinical and Radiological Features: Performance Comparison With Preoperative Biopsy.

OBJECTIVE: We investigated if PD-L1 expression can be predicted by machine learning using clinical and imaging features. METHODS: We included 117 patients with c-stage I/II non-small cell lung cancer who underwent radical resection. A total of 3951 radiomic features were extracted by defining the tumor (within tumor contour), rim (contour ±3 mm) and exterior (contour +10 mm) on preoperative contrast computed tomography. After feature selection by Boruta algorithm, prediction models of tumor PD-L1 expression (22C3: ≥1%, <1%) of resected specimens were constructed using Random Forest: radiomics, clinical, and combined models. Their performance was evaluated by 5-fold cross-validation, and AUCs were compared using Delong test. Next, study groups were categorized as patients without biopsy (training set), and those with biopsy (test set). Predictive ability of biopsy was compared to each prediction model. RESULTS: Of 117 patients (66 ± 10 years old, 48% male), 33 (28.2%) had PD-L1≥1%. Mean AUC of PD-L1≥1% for the validation set in radiomics, clinical, and combined models were 0.80, 0.80, and 0.83 (P = .32 vs. clinical model), respectively. The diagnosis of malignancy was made in 22 of 38 (58%) patients with attempted biopsies, and PD-L1 was measurable in 19 of 38 (50%) patients. Diagnostic accuracies of PD-L1≥1% from 19 determinable biopsies and 38 all attempted biopsies were 0.68 and 0.34, respectively. These were out performed by machine learning: 0.71, 0.71, and 0.74 for radiomics, clinical, and combined models, respectively. CONCLUSIONS: Our machine learning could be an adjunctive tool in estimating PD-L1 expression prior to neoadjuvant treatment, particularly when PD-L1 is indeterminable with biopsy.

Detection of EGFR mutations in early-stage lung adenocarcinoma by machine learning-based radiomics.

Background: We hypothesized that epidermal growth factor receptor (EGFR) mutations could be detected in early-stage lung adenocarcinoma using radiomics. Methods: This retrospective study included consecutive patients with clinical stage I/II lung adenocarcinoma who underwent curative-intent pulmonary resection from March-December 2016. Using preoperative enhanced chest computed tomography, 3,951 radiomic features were extracted in total from the tumor (area within the tumor boundary), tumor rim (area within ±3 mm of the tumor boundary), and tumor exterior (area between +10 mm outside the tumor and tumor boundary). A machine learning-based radiomics model was constructed to detect EGFR mutations. The combined model incorporated both radiomic and clinical features (gender and smoking history). The performance was validated with five-fold cross-validation and evaluated using the mean area under the curve (AUC). Results: Of 99 patients (mean age, 66±11 years; female, 66.6%; clinical stage I/II, 89.9%/10.1%), EGFR mutations in the surgical specimen were detected in 46 (46.5%). A median of 4 (range, 2 to 8) radiomic features was selected for each validation session. The mean AUCs in the radiomics and combined models were 0.75 and 0.83, respectively. The two top-ranked features in the combined model were the radiomic features extracted from the tumor exterior and the tumor, indicating a higher impact of radiomic features over relevant clinical features. Conclusions: Radiomic features, including those in the peri-tumoral area, may help detect EGFR mutations in lung adenocarcinomas in preoperative settings. This non-invasive image-based technology could help guide future precision neoadjuvant therapy.

Impact of postoperative complications on the long-term outcome in lung cancer surgery.

PURPOSE: Postoperative complications have a significant impact on perioperative outcomes; however, their association with the long-term prognosis remains unclear. We evaluated the impact of postoperative complications on the long-term outcomes after curative surgery in lung cancer patients. METHODS: This study included 1129 patients with primary lung cancer who underwent lobectomy between April 2011 and March 2017. Univariate and multivariate analyses were performed to assess the association of postoperative complications with the overall and recurrence-free survival. RESULTS: Postoperative complications were observed in 147 (13.0%) patients over a median follow-up period of 5-years. Compared to patients without complications, those with complications showed had worse long-term outcomes, including the 5-year overall survival (75.3% vs. 86.1%, p < 0.001) and 5-year recurrence-free survival (64.2% vs. 74.4%, p = 0.004). A multivariate analysis revealed that the incidence of postoperative complications was significantly associated with the overall survival (hazard ratio = 1.665, p = 0.006) and recurrence-free survival (hazard ratio = 1.416, p = 0.025) in all patients. The prognostic influence was greater in patients with pathological stages II and III cancer (overall survival: hazard ratio = 2.019, p = 0.005; recurrence-free survival: hazard ratio = 1.90, p = 0.001) than in those with pathological stage I cancer. CONCLUSION: Postoperative complications are independent predictors of the overall and recurrence-free survival in lung cancer patients, especially advanced-stage cancer patients.

Correlation Between Smoking Status and Short-term Outcome of Thoracoscopic Surgery for Lung Cancer.

BACKGROUND: Smoking has a major role in the risk of postoperative pulmonary complications. This study aimed to elucidate the correlation between smoking status and pulmonary complications after thoracoscopic surgery for lung cancer. METHODS: A total of 1751 patients who underwent thoracoscopic lobectomy or segmentectomy for lung cancer between April 2011 and March 2020 were assessed. The rate of pulmonary complications was evaluated according to smoking status and preoperative duration of smoking cessation. Univariate and multivariate logistic regression analyses were performed. RESULTS: Pulmonary complications were observed in 50 patients (2.9%), whereas 3 (0.2%) died within 90 days of surgery. The rate of pulmonary complications was higher in smokers than in nonsmokers (4.6% vs 0.9%; P < .001), and smoking history was an independent risk factor for pulmonary complications (odds ratio, 3.31; P = .007). The complication rate in patients with a cessation period of more than 2 months was significantly lower than that in patients who ceased smoking within 2 months (4.0% vs 8.5%; P = .043), but it was still higher than that in nonsmokers (4.0% vs 0.9%; P < .001). In the multivariable analysis for smokers, preoperative short-term smoking cessation within 2 months, male sex, histologic type, tumor size, and cardiopulmonary comorbidities were associated with pulmonary complications instead of pack-year smoking history. CONCLUSIONS: Smoking habits and preoperative smoking cessation were independently associated with pulmonary complications after thoracoscopic surgery for lung cancer. A preoperative smoking cessation period of 2 months or more is preferable for reducing the risk of such complications.

Three-dimensional image simulation for lung segmentectomy from unenhanced computed tomography data.

We developed a novel three-dimensional (3D) image simulation system, focused on pulmonary segmentectomy. The novel algorithms run by the software, which are independent of the differences in computed tomography (CT) values of vascular structures, enabled the creation of 3D images from unenhanced CT data with accuracy comparable to that from contrast-enhanced CT data. To evaluate the anatomical accuracy, we compared it between images created from unenhanced and contrast-enhanced CT in seven patients who underwent thoracoscopic segmentectomy. With regard to the automatic recognition of pulmonary vessels, the 3D image from unenhanced CT falsely recognized one or two points in two cases, whereas that from contrast-enhanced CT false recognitions in one case. Both 3D images had similar creation time and capability for identifying the intersegmental plain. The novel 3D image simulation for segmentectomy from unenhanced CT had sufficient anatomical accuracy for practical use but required attention due to inevitable minor false recognition.

Relationship between the three-dimensionally measured tumor doubling time of lung cancer and underlying interstitial lung disease: A retrospective case-control study.

OBJECTIVE: The coexistence of interstitial lung disease (ILD) is associated with poor prognosis in patients with lung cancer. The tumor doubling time (TDT) of lung cancer reflects cancer aggressiveness and is related to its prognosis. However, the relationship between the TDT of lung cancer and underlying ILD has not been fully evaluated. This study aimed to identify this crucial relationship. MATERIALS AND METHODS: Patients with lung cancer who underwent surgery between 2007 and 2020 were reviewed retrospectively. The propensity score matching method was used to balance the characteristics of patients with ILD (n = 100) and those without ILD (n = 100). TDT was calculated based on the difference of three-dimensional volumes defined from the two-time CT scans before surgery. We compared the TDT of lung cancer and other characteristics between the two groups. RESULTS: The median TDT of all patients was 149 days. The TDT was significantly shorter in patients with ILD (134 days) than in those without (204 days). The rate of short-term tumor enlargement (TDT < 90 days) was significantly higher in patients with ILD than in those without ILD, and ILD was an independent factor related to short-term tumor enlargement (odds ratio, 2.30; p = 0.015). We focused on 25 patients with usual interstitial pneumonitis (UIP) findings of patients with ILD. However, the presence of the UIP pattern was not related to the TDT among patients with ILD. CONCLUSION: ILD was an independent predictor of short-term tumor enlargement in lung cancer patients, regardless of the presence of the UIP pattern.

Novel three-dimensional image simulation for lung segmentectomy developed with surgeons' perspective.

We developed a novel three-dimensional (3D) image simulation system that is especially focused on pulmonary segmentectomy using new 3D computed tomography (CT) software. Based on contrast-enhanced high-resolution computed tomography (HRCT) images, the new software can quickly construct 3D pulmonary and bronchovascular images and generate a proposal for the appropriate segments to be resected. We performed the 3D image simulation and evaluated its accuracy in 20 patients for whom thoracoscopic segmentectomy was planned. We evaluated the anatomical validity comparing with HRCT findings and anatomical consistency with the operative findings on a three-point scale, respectively. The 3D image was evaluated as "good" for anatomical validity in 19 cases (95%) and for anatomical consistency with operative findings in 18 cases (90%). The novel 3D image simulation appeared to be easy to prepare, was anatomically reliable, and, therefore, was determined to be potentially useful.

Chest wall fistula in a patient with pyothorax-associated lymphoma.

Fistula formation is an uncommon but serious therapeutic complication of pyothorax-associated lymphoma (PAL) because it decreases the quality of life in patients. Furthermore, a collapsed lung may predispose to pneumonia. In PAL, the lesions might invade the skin and optimal irradiation dose, region, and timing should be carefully determined.

A rapidly growing mature mediastinal teratoma with a testicular epidermoid cyst and familial Mediterranean fever.

Anterior mediastinal teratomas are common and are generally characterized as slow growing tumors. Very few reports documenting rapidly growing tumors exist. Here, we describe a case of a mature teratoma showing rapid growth in 1 year treated with complete surgical resection.

The utility of three-dimensional computed tomography for prediction of tumor invasiveness in clinical stage IA lung adenocarcinoma.

BACKGROUND: It is critical to have an accurate measurement of solid tumor size in order to predict the invasiveness of small lung adenocarcinomas. Some lesions cannot be measured accurately via High-resolution computed tomography (HRCT) due to their irregular shape and unclear borders. For this reason, we evaluated the relative efficacy of three-dimensional (3D) CT for predicting invasive adenocarcinoma. METHODS: We evaluated 195 patients with clinical stage IA adenocarcinomas, including 109 with lesions documented as invasive that were surgically resected at our institute during 2017. All lesions were categorized as either (I) lesions that were difficult to evaluate (i.e., hazy lesions; HL) or (II) more typical lesions (TL). The relationships between solid tumor size as determined by HRCT, solid tumor volume as determined by 3D CT and pathologic diagnosis were evaluated. RESULTS: Fifty-seven patients (29%) were diagnosed with HL. We set the cut-off value for the solid volume at 225 mm3 as predictive for invasive adenocarcinoma. When evaluating all 195 patients as a group, the accuracy, sensitivity, and specificity based on the solid tumor volume were similar to those based on the solid tumor size. When we limit our analysis to the HL group, the specificity based on solid tumor volume (65.5%) was higher than that based on solid tumor size (44.8%) with a difference that approached statistical significance (P=0.070). CONCLUSIONS: 3D CT was equivalent to HRCT for predicting invasive adenocarcinoma and may be particularly useful for diagnosing lesions that are difficult to evaluate on HRCT.