Quantitative Computed Tomography Imaging Biomarkers in the Diagnosis and Management of Lung Cancer

Two-stage multitask U-Net construction for pulmonary nodule segmentation and malignancy risk prediction

BACKGROUND

Accurate segmentation of pulmonary nodules is important for image-driven nodule analysis and nodule malignancy risk prediction. However, due to interobserver variability caused by manual segmentation, an accurate and robust automatic segmentation method has become an essential task. Therefore, the aim of the present study was to construct an accurate segmentation and malignant risk prediction algorithm for pulmonary nodules.

METHODS

In the present study, we proposed a coarse-to-fine 2-stage framework consisting of the following 2 convolutional neural networks: a 3D multiscale U-Net used for localization and a 2.5D multiscale separable U-Net (MSU-Net) used for segmentation refinement. A multitask framework was proposed for nodules' malignancy risk prediction. Features from encoding and decoding paths of MSU-Net were integrated for pathology or morphology characteristic classification.

RESULTS

Experimental results showed that our method achieved state-of-art results on the Lung Image Database Consortium and Image Database Resource Initiative dataset. The proposed method achieved a Dice similarity coefficient (DSC) of 83.04% and an overlapping error of 27.47% on the dataset. Our method achieved accuracy of 77.8% and area under the receiver-operating characteristic curve of 84.3% for malignancy risk prediction. Moreover, we compared our method with the inter-radiologist agreement, and the average DSC difference was only 0.39%.

CONCLUSIONS

The results showed the effectiveness of the multitask end-to-end framework. The coarse-to-fine 2.5D strategy increased the accuracy and efficiency of pulmonary nodule segmentation and malignancy risk prediction of the computer-aided diagnosis system. In clinical practice, doctors can obtain accurate morphological characteristics and quantitative information of nodules by using the proposed method, so as to make future treatment plan.

Role of magnetic resonance spectroscopy in differential diagnosis of solitary pulmonary lesions

PURPOSE

The aim of our study was to evaluate the availability of magnetic resonance spectroscopy (MRS) for the differentiation of benign or malignant pulmonary nodules and masses.

METHODS

A total of 59 patients (45 male, 14 female) with pulmonary nodules and masses were included in this prospective study. MRS was applied to the pulmonary lesions of the patients and choline levels were determined. Afterwards CT-guided percutaneous needle biopsy was performed. According to the biopsy results, pulmonary lesions were benign in 25 patients and malignant in 34 patients.

RESULTS

Choline levels were significantly higher in malignant lesions compared with benign lesions (p < 0.001). When the other conditions were kept constant, the probability of malignancy significantly increased by 17.38-fold (95% CI, 3.78-79.93) in those with choline levels >1.65 µmol/g compared to those with choline levels ≤1.65 µmol/g (p < 0.001).

CONCLUSION

MRS is a noninvasive method that can be used in the differential diagnosis of pulmonary nodules and masses.

Effect of CT image acquisition parameters on diagnostic performance of radiomics in predicting malignancy of pulmonary nodules of different sizes

OBJECTIVES

To investigate the effect of CT image acquisition parameters on the performance of radiomics in classifying benign and malignant pulmonary nodules (PNs) with respect to nodule size.

METHODS

We retrospectively collected CT images of 696 patients with PNs from March 2015 to March 2018. PNs were grouped by nodule diameter: T1a (diameter ≤ 1.0 cm), T1b (1.0 cm < diameter ≤ 2.0 cm), and T1c (2.0 cm < diameter ≤ 3.0 cm). CT images were divided into four settings according to slice-thickness-convolution-kernels: setting 1 (slice thickness/reconstruction type: 1.25 mm sharp), setting 2 (5 mm sharp), setting 3 (5 mm smooth), and random setting. We created twelve groups from two interacting conditions. Each PN was segmented and had 1160 radiomics features extracted. Non-redundant features with high predictive ability in training were selected to build a distinct model under each of the twelve subsets.

RESULTS

The performance (AUCs) on predicting PN malignancy were as follows: T1a group: 0.84, 0.64, 0.68, and 0.68; T1b group: 0.68, 0.74, 0.76, and 0.70; T1c group: 0.66, 0.64, 0.63, and 0.70, for the setting 1, setting 2, setting 3, and random setting, respectively. In the T1a group, the AUC of radiomics model in setting 1 was statistically significantly higher than all others; In the T1b group, AUCs of radiomics models in setting 3 were statistically significantly higher than some; and in the T1c group, there were no statistically significant differences among models.

CONCLUSIONS

For PNs less than 1 cm, CT image acquisition parameters have a significant influence on diagnostic performance of radiomics in predicting malignancy, and a model created using images reconstructed with thin section and a sharp kernel algorithm achieved the best performance. For PNs larger than 1 cm, CT reconstruction parameters did not affect diagnostic performance substantially.

KEY POINTS

• CT image acquisition parameters have a significant influence on the diagnostic performance of radiomics in pulmonary nodules less than 1 cm. • In pulmonary nodules less than 1 cm, a radiomics model created by using images reconstructed with thin section and a sharp kernel algorithm achieved the best diagnostic performance. • For PNs larger than 1 cm, CT image acquisition parameters do not affect diagnostic performance substantially.

Effect of adaptive statistical iterative reconstruction-V (ASiR-V) levels on ultra-low-dose CT radiomics quantification in pulmonary nodules

Background

The weightings of iterative reconstruction algorithm can affect CT radiomic quantification. But, the effect of ASiR-V levels on the reproducibility of CT radiomic features between ultra-low-dose computed tomography (ULDCT) and low-dose computed tomography (LDCT) is still unknown. The purpose of study is to investigate whether adaptive statistical iterative reconstruction-V (ASiR-V) levels affect radiomic feature quantification using ULDCT and to assess the reproducibility of radiomic features between ULDCT and LDCT.

Methods

Sixty-three patients with pulmonary nodules underwent LDCT (0.70±0.16 mSv) and ULDCT (0.15±0.02 mSv). LDCT was reconstructed with ASiR-V 50%, and ULDCT with ASiR-V 50%, 70%, and 90%. Radiomics analysis was applied, and 107 features were extracted. The concordance correlation coefficient (CCC) was calculated to describe agreement among ULDCTs and between ULDCT and LDCT for each feature. The proportion of features with CCC >0.9 among ULDCTs and between ULDCT and LDCT, and the mean CCC for all features between ULDCT and LDCT were also compared.

Results

Sixty-three solid nodules (SNs) and 48 pure ground-glass nodules (pGGNs) were analyzed. There was no difference for the proportion of features in SNs among ULDCTs and between ULDCT and LDCT (P>0.05). The proportion of features in pGGNs were highest for ULDCT_{70% vs. 90%} (78.5%) and ULDCT_90% vs. LDCT_50% (50.5%). In SNs, the mean CCC for ULDCT_90% vs. LDCT_50% was 0.67±0.26, not different with that for ULDCT_50% vs. LDCT_50% (0.68±0.24) and ULDCT_70% vs. LDCT_50% (0.64±0.21) (P>0.05). In pGGNs, the mean CCC for ULDCT_90% vs. LDCT_50% was 0.79±0.19, higher than that for ULDCT_50% vs. LDCT_50% (0.61±0.28) and ULDCT_70% vs. LDCT_50% (0.76±0.24) (P<0.05).

Conclusions

ASiR-V levels significantly affected ULDCT radiomic feature quantification in pulmonary nodules, with stronger effects in pGGNs than in SNs. The reproducibility of radiomic features was highest between ULDCT_90% and LDCT_50%.

Radial endobronchial ultrasound-assisted transbronchial needle aspiration for pulmonary peripheral lesions in the segmental bronchi adjacent to the central airway

Background

Tissue samples from lesions located in the 3rd to 5th segmental bronchi are challenging to obtain. In this retrospective study, we aimed to evaluate the diagnostic rate of pulmonary peripheral lesions located in the 3rd to 5th segmental bronchi, near the inner field of lung on the computed tomography (CT) image and outside the bronchus, using radial endobronchial ultrasound (REBUS) followed by transbronchial needle aspiration (TBNA).

Methods

This retrospective study enrolled patients whose preoperative CT examinations showed a lesion located in the segmental bronchi (3rd to 5th), yet adjacent to the inner field of lung on the CT image. REBUS followed by TBNA was used to acquire tissue samples from these lesions. A bronchoscope was used to reach the bronchi surrounding the lesion, and an ultrasound probe was used to determine the lesion's location. Then, the ultrasound probe was withdrawn, and puncture was performed at the location that was determined by ultrasound. The tissue specimens obtained were subjected to pathological examination.

Results

Nineteen patients were enrolled in this study including 15 males and 4 females with an average age of 55 years old. Of the enrollees, 8 patients (42.1%) were successfully diagnosed with samples obtained through TBNA, including 6 cases of lung cancer, 1 case of non-specific inflammation, and 1 case of cryptococcal infection. The diagnostic rate was 42.1%. No post-procedural complications were observed among the patients. There was no significant difference in nodule diameter between patients with a diagnostic sample and those in whom TBNA failed to provide a diagnosis (2.99±0.96 vs. 2.26±1.27 cm, P=0.20).

Conclusions

With the assistance of REBUS, TBNA can acquire sufficient samples to achieve a reasonably diagnostic rate for parenchymal lung lesions located near the inner field of lung on the CT image without intrabronchial invasion.

Computer-aided diagnosis of ground glass pulmonary nodule by fusing deep learning and radiomics features

OBJECTIVES

This study aims to develop a computer-aided diagnosis (CADx) scheme to classify between benign and malignant ground glass nodules (GGNs), and fuse deep leaning and radiomics imaging features to improve the classification performance.

METHODS

We first retrospectively collected 513 surgery histopathology confirmed GGNs from two centers. Among these GGNs, 100 were benign and 413 were malignant. All malignant tumors were stage I lung adenocarcinoma. To segment GGNs, we applied a deep convolutional neural network and residual architecture to train and build a 3D U-Net. Then, based on the pre-trained U-Net, we used a transfer learning approach to build a deep neural network (DNN) to classify between benign and malignant GGNs. With the GGN segmentation results generated by 3D U-Net, we also developed a CT radiomics model by adopting a series of image processing techniques, i.e. radiomics feature extraction, feature selection, synthetic minority over-sampling technique, and support vector machine classifier training/testing, etc. Finally, we applied an information fusion method to fuse the prediction scores generated by DNN based CADx model and CT-radiomics based model. To evaluate the proposed model performance, we conducted a comparison experiment by testing on an independent testing dataset.

RESULTS

Comparing with DNN model and radiomics model, our fusion model yielded a significant higher area under a receiver operating characteristic curve (AUC) value of 0.73 ± 0.06 (P < 0.01). The fusion model generated an accuracy of 75.6%, F1 score of 84.6%, weighted average F1 score of 70.3%, and Matthews correlation coefficient of 43.6%, which were higher than the DNN model and radiomics model individually.

CONCLUSIONS

Our experimental results demonstrated that (1) applying a CADx scheme was feasible to diagnosis of early-stage lung adenocarcinoma, (2) deep image features and radiomics features provided complementary information in classifying benign and malignant GGNs, and (3) it was an effective way to build DNN model with limited dataset by using transfer learning. Thus, to build a robust image analysis based CADx model, one can combine different types of image features to decode the imaging phenotypes of GGN.

The value of percentile base on computed tomography histogram in differentiating the invasiveness of adenocarcinoma appearing as pure ground-glass nodules

To investigate the value of percentile base on computed tomography (CT) histogram analysis for distinguishing invasive adenocarcinoma (IA) from adenocarcinoma in situ (AIS) or micro invasive adenocarcinoma (MIA) appearing as pure ground-glass nodules.A total of 42 cases of pure ground-glass nodules that were surgically resected and pathologically confirmed as lung adenocarcinoma between January 2015 and May 2019 were included. Cases were divided into IA and AIS/MIA in the study. The percentile on CT histogram was compared between the 2 groups. Univariate and multivariate logistic regression were used to determine which factors demonstrated a significant effect on invasiveness. The receiver operating characteristic (ROC) curve and the area under the curve (AUC) was used to evaluate the predictive ability of individual characteristics and the combined model.The 4 histogram parameters (25th percentile, 55th percentile, 95th percentile, 97.5th percentile) and the combined model all showed a certain diagnostic value. The combined model demonstrated the best diagnostic performance. The AUC values were as follows: 25th percentile = 0.693, 55th percentile = 0.706, 95th percentile = 0.713, 97.5th percentile = 0.710, and combined model = 0.837 (all P < .05).The percentile of histogram parameters help to improve the ability to radiologically determine the invasiveness of lung adenocarcinoma appearing as pure ground-glass nodules.

Exploratory Study of a CT Radiomics Model for the Classification of Small Cell Lung Cancer and Non-small-Cell Lung Cancer

Background: Radiomics can quantify tumor phenotypic characteristics non-invasively by applying feature algorithms to medical imaging data. In this study, we investigated the association between radiomics features and the tumor histological subtypes, and we aimed to establish a nomogram for the classification of small cell lung cancer (SCLC) and non-small-cell lung cancer (NSCLC). Methods: This was a retrospective single center study. In total, 468 cases including 202 patients with SCLC and 266 patients with NSCLC were enrolled in our study, and were randomly divided into a training set (n = 327) and a validation set (n = 141) in a 7:3 ratio. The clinical data of the patients, including age, sex, smoking history, tumor maximum diameter, clinical stage, and serum tumor markers, were collected. All patients underwent enhanced computed tomography (CT) scans, and all lesions were pathologically confirmed. A radiomics signature was generated from the training set using the least absolute shrinkage and selection operator algorithm. Independent risk factors were identified by multivariate logistic regression analysis, and a radiomics nomogram based on the radiomics signature and clinical features was constructed. The capability of the nomogram was evaluated in the training set and validated in the validation set. Results: Fourteen of 396 radiomics parameters were screened as important factors for establishing the radiomics model. The radiomics signature performed well in differentiating SCLC and NSCLC, with an area under the curve (AUC) of 0.86 (95% CI: 0.82-0.90) in the training set and 0.82 (95% CI: 0.75-0.89) in the validation set. The radiomics nomogram had better predictive performance [AUC = 0.94 (95% CI: 0.90-0.98) in the validation set] than the clinical model [AUC = 0.86 (95% CI: 0.80-0.93)] and the radiomics signature [AUC = 0.82 (95% CI: 0.75-0.89)], and the accuracy was 86.2% (95% CI: 0.79-0.92) in the validation set. Conclusion: The enhanced CT radiomics signature performed well in the classification of SCLC and NSCLC. The nomogram based on the radiomics signature and clinical factors has better diagnostic performance for the classification of SCLC and NSCLC than the simple application of the radiomics signature.

Comparison and Fusion of Deep Learning and Radiomics Features of Ground-Glass Nodules to Predict the Invasiveness Risk of Stage-I Lung Adenocarcinomas in CT Scan

For stage-I lung adenocarcinoma, the 5-years disease-free survival (DFS) rates of non-invasive adenocarcinoma (non-IA) is different with invasive adenocarcinoma (IA). This study aims to develop CT image based artificial intelligence (AI) schemes to classify between non-IA and IA nodules, and incorporate deep learning (DL) and radiomics features to improve the classification performance. We collect 373 surgical pathological confirmed ground-glass nodules (GGNs) from 323 patients in two centers. It involves 205 non-IA (including 107 adenocarcinoma in situ and 98 minimally invasive adenocarcinoma), and 168 IA. We first propose a recurrent residual convolutional neural network based on U-Net to segment the GGNs. Then, we build two schemes to classify between non-IA and IA namely, DL scheme and radiomics scheme, respectively. Third, to improve the classification performance, we fuse the prediction scores of two schemes by applying an information fusion method. Finally, we conduct an observer study to compare our scheme performance with two radiologists by testing on an independent dataset. Comparing with DL scheme and radiomics scheme (the area under a receiver operating characteristic curve (AUC): 0.83 ± 0.05, 0.87 ± 0.04), our new fusion scheme (AUC: 0.90 ± 0.03) significant improves the risk classification performance (p < 0.05). In a comparison with two radiologists, our new model yields higher accuracy of 80.3%. The kappa value for inter-radiologist agreement is 0.6. It demonstrates that applying AI method is an effective way to improve the invasiveness risk prediction performance of GGNs. In future, fusion of DL and radiomics features may have a potential to handle the classification task with limited dataset in medical imaging.

An Interpretable Deep Hierarchical Semantic Convolutional Neural Network for Lung Nodule Malignancy Classification

While deep learning methods have demonstrated performance comparable to human readers in tasks such as computer-aided diagnosis, these models are difficult to interpret, do not incorporate prior domain knowledge, and are often considered as a "black-box." The lack of model interpretability hinders them from being fully understood by end users such as radiologists. In this paper, we present a novel interpretable deep hierarchical semantic convolutional neural network (HSCNN) to predict whether a given pulmonary nodule observed on a computed tomography (CT) scan is malignant. Our network provides two levels of output: 1) low-level semantic features; and 2) a high-level prediction of nodule malignancy. The low-level outputs reflect diagnostic features often reported by radiologists and serve to explain how the model interprets the images in an expert-interpretable manner. The information from these low-level outputs, along with the representations learned by the convolutional layers, are then combined and used to infer the high-level output. This unified architecture is trained by optimizing a global loss function including both low- and high-level tasks, thereby learning all the parameters within a joint framework. Our experimental results using the Lung Image Database Consortium (LIDC) show that the proposed method not only produces interpretable lung cancer predictions but also achieves significantly better results compared to using a 3D CNN alone.

Computer-aided diagnosis of ground-glass opacity pulmonary nodules using radiomic features analysis

This study aims to develop a CT-based radiomic features analysis approach for diagnosis of ground-glass opacity (GGO) pulmonary nodules, and also assess whether computer-aided diagnosis (CADx) performance changes in classifying between benign and malignant nodules associated with histopathological subtypes namely, adenocarcinoma in situ (AIS), minimally invasive adenocarcinoma (MIA), and invasive adenocarcinoma (IAC), respectively. The study involves 182 histopathology-confirmed GGO nodules collected from two cancer centers. Among them, 59 are benign, 50 are AIS, 32 are MIA, and 41 are IAC nodules. Four training/testing data sets-(1) all nodules, (2) benign and AIS nodules, (3) benign and MIA nodules, (4) benign and IAC nodules-are assembled based on their histopathological subtypes. We first segment pulmonary nodules depicted in CT images by using a 3D region growing and geodesic active contour level set algorithm. Then, we computed and extracted 1117 quantitative imaging features based on the 3D segmented nodules. After conducting radiomic features normalization process, we apply a leave-one-out cross-validation (LOOCV) method to build models by embedding with a Relief feature selection, synthetic minority oversampling technique (SMOTE) and three machine-learning classifiers namely, support vector machine classifier, logistic regression classifier and Gaussian Naïve Bayes classifier. When separately using four data sets to train and test three classifiers, the average areas under receiver operating characteristic curves (AUC) are 0.75, 0.55, 0.77 and 0.93, respectively. When testing on an independent data set, our scheme yields higher accuracy than two radiologists (61.3% versus radiologist 1: 53.1% and radiologist 2: 56.3%). This study demonstrates that: (1) the feasibility of using CT-based radiomic features analysis approach to distinguish between benign and malignant GGO nodules, (2) higher performance of CADx scheme in diagnosing GGO nodules comparing with radiologist, and (3) a consistently positive trend between classification performance and invasive grade of GGO nodules. Thus, to improve the CADx performance in diagnosing of GGO nodules, one should assemble an optimal training data set dominated with more nodules associated with non-invasive lung adenocarcinoma (i.e. AIS and MIA).

Machine learning-based radiomics strategy for prediction of cell proliferation in non-small cell lung cancer

PURPOSE

To explore the feasibility and performance of machine learning-based radiomics classifier to predict the cell proliferation(Ki-67)in non-small cell lung cancer (NSCLC).

METHODS

245 histopathological confirmed NSCLC patients who underwent CT scans were retrospectively included. The Ki-67 proliferation index (Ki-67 PI) were measured within 2 weeks after CT scans. A lesion volume of interest (VOI) was manually delineated and radiomics features were extracted by MaZda software from CT images. A random forest feature selection algorithm (RFFS) was used to reduce features. Six kinds of machine learning methods were used to establish radiomics classifiers, subjective imaging feature classifiers and combined classifiers, respectively. The performance of these classifiers was evaluated by the receiver operating characteristic curve (ROC) and compared with Delong test.

RESULTS

103 radiomics features were extracted and 20 optimal features were selected using RFFS. Among the radiomics classifiers established by six machine learning methods, random forest-based radiomics classifier achieved the best performance (AUC = 0.776) in predicting the Ki-67 expression level with sensitivity and specificity of 0.726 and 0.661, which was better than that of subjective imaging classifiers (AUC = 0.625, P < 0.05). However, the combined classifiers did not improve the predictive performance (AUC = 0.780, P > 0.05), with sensitivity and specificity of 0.752 and 0.633.

CONCLUSIONS

The machine learning-based CT radiomics classifier in NSCLC can facilitate the prediction of the expression level of Ki-67 and provide a novel non-invasive strategy for assessing the cell proliferation.

Evaluation of response to conventional chemotherapy and radiotherapy by perfusion computed tomography in non-small cell lung cancer (NSCLC)

BACKGROUND

To evaluate changes in perfusion computed tomography (PCT) parameters induced by treatment with conventional chemotherapy (CCT) alone or with CCT and radiation therapy (RT) in patients with non-small cell lung cancer (NSCLC) and to determine whether these changes correlate with response as defined by the response evaluation criteria in solid tumours version 1.1 (RECIST-1.1).

METHODS

Fifty-three patients with a histological diagnosis of NSCLC prospectively underwent PCT of the whole tumour, before/after CCT or before/after CCT and RT. Blood flow (BF), blood volume (BV), permeability (PMB), and mean transit time (MTT) were compared before and after treatment and with the response as defined by RECIST-1.1. The relationship between changes in the perfusion parameters and in tumour size was also evaluated.

RESULTS

PCT parameters decreased after treatment, significantly for BV (p = 0.002) and MTT (p = 0.027). The 30 patients with partial response had a significant decrease of 21% for BV (p = 0.006) and 17% for MTT (p = 0.031). A non-significant decrease in all perfusion parameters was found in patients with stable disease (p > 0.137). In patients with progressive disease, MTT decreased by 10% (p = 0.465) and the other parameters did not significantly vary (p > 0.809). No significant correlation was found between changes in size and PCT parameters (p > 0.145).

CONCLUSIONS

Treatment of NSCLC with platinum derivatives, with or without RT, induces changes in PCT parameters. Partial response is associated with a significant decrease in BV and MTT, attributable to the effect of the treatment on tumour vascularisation.

Implication of total tumor size on the prognosis of patients with clinical stage IA lung adenocarcinomas appearing as part-solid nodules: Does only the solid portion size matter?

OBJECTIVES

The aim was to investigate the effect of clinico-radiologic variables, including total tumor (T_total) size and clinical T category, on the prognosis of patients with stage IA (T1N0M0) lung adenocarcinomas appearing as part-solid nodules (PSNs).

METHODS

This institutional review board-approved retrospective study included 506 patients (male:female = 200:306; median age, 62 years) with PSNs of the adenocarcinoma spectrum in clinical stage IA who underwent standard lobectomy at a single tertiary medical center. Prognostic stratification of the patients in terms of disease-free survival was analyzed with variables including age, sex, T_total size, solid portion size, clinical T category, and tumor location using univariate and subsequent multivariate Cox regression analysis. Subgroup analysis was performed to reveal the effect of the T_total size at each clinical T category.

RESULTS

Multivariate Cox regression analysis demonstrated that T_total size*cT1b [interaction term; hazard ratio (HR) = 1.091; 95% confidence interval (CI): 1.015, 1.173; p = 0.019] and cT1c (HR = 68.436; 95% CI: 2.797, 1674.415; p = 0.010) were independent risk factors for the tumor recurrence. When patients with cT1b were dichotomized based on a T_total size cutoff of 3.0 cm, PSNs with T_total > 3.0 cm showed a significantly worse outcome (HR = 3.796; 95% CI: 1.006, 14.317; p = 0.049). No significant difference was observed in the probability of recurrence between cT1b with T_total > 3.0 cm and cT1c (p = 0.915).

CONCLUSIONS

T_total size is a significant prognostic factor in adenocarcinoma patients in cT1b without lymph node or distant metastasis. PSNs in cT1b with T_total > 3.0 cm have a comparable risk of lung cancer recurrence to those in cT1c.

KEY POINTS

• Current T descriptor was a powerful prognostic factor in stage IA adenocarcinomas appearing as part-solid nodules. • Total tumor size further stratified risk of recurrence of adenocarcinomas in cT1b. • Upstaging of tumors in cT1b with total tumor size > 3.0 cm may be more appropriate.

Radiomic features analysis in computed tomography images of lung nodule classification

Purpose

Radiomics, which extract large amount of quantification image features from diagnostic medical images had been widely used for prognostication, treatment response prediction and cancer detection. The treatment options for lung nodules depend on their diagnosis, benign or malignant. Conventionally, lung nodule diagnosis is based on invasive biopsy. Recently, radiomics features, a non-invasive method based on clinical images, have shown high potential in lesion classification, treatment outcome prediction.

Methods

Lung nodule classification using radiomics based on Computed Tomography (CT) image data was investigated and a 4-feature signature was introduced for lung nodule classification. Retrospectively, 72 patients with 75 pulmonary nodules were collected. Radiomics feature extraction was performed on non-enhanced CT images with contours which were delineated by an experienced radiation oncologist.

Result

Among the 750 image features in each case, 76 features were found to have significant differences between benign and malignant lesions. A radiomics signature was composed of the best 4 features which included Laws_LSL_min, Laws_SLL_energy, Laws_SSL_skewness and Laws_EEL_uniformity. The accuracy using the signature in benign or malignant classification was 84% with the sensitivity of 92.85% and the specificity of 72.73%.

Conclusion

The classification signature based on radiomics features demonstrated very good accuracy and high potential in clinical application.

Pulmonary quantitative CT imaging in focal and diffuse disease: current research and clinical applications

The frenetic development of imaging technology-both hardware and software-provides exceptional potential for investigation of the lung. In the last two decades, CT was exploited for detailed characterization of pulmonary structures and description of respiratory disease. The introduction of volumetric acquisition allowed increasingly sophisticated analysis of CT data by means of computerized algorithm, namely quantitative CT (QCT). Hundreds of thousands of CTs have been analysed for characterization of focal and diffuse disease of the lung. Several QCT metrics were developed and tested against clinical, functional and prognostic descriptors. Computer-aided detection of nodules, textural analysis of focal lesions, densitometric analysis and airway segmentation in obstructive pulmonary disease and textural analysis in interstitial lung disease are the major chapters of this discipline. The validation of QCT metrics for specific clinical and investigational needs prompted the translation of such metrics from research field to patient care. The present review summarizes the state of the art of QCT in both focal and diffuse lung disease, including a dedicated discussion about application of QCT metrics as parameters for clinical care and outcomes in clinical trials.

Temporal Changes of Texture Features Extracted From Pulmonary Nodules on Dynamic Contrast-Enhanced Chest Computed Tomography: How Influential Is the Scan Delay?

OBJECTIVES

The aim of this study was to describe the temporal changes of various texture features extracted from pulmonary nodules on dynamic contrast-enhanced computed tomography (DCE-CT) and to compare the feature values among multiple scanning time points. We also aimed to analyze the variability of texture features across multiple scan delay times.

MATERIALS AND METHODS

This retrospective study was approved by the institutional review board of Seoul National University Hospital with waiver of patients' informed consent. Twenty patients (M:F, 6:14; mean age, 60.25 ± 11.97 years) with 20 lung nodules (mean size, 24.1 ± 12.3 mm) underwent DCE-CT with multiple scan delays (30, 60, 90, 120, 150, 180, 210, 240, 300, and 480 seconds) after precontrast scans. Lung nodule segmentation and texture feature extraction were performed at each time point using in-house software. Texture feature values were compared among the multiple time points using the Friedman test with post hoc pairwise Wilcoxon signed rank test. In addition, the dynamic range (DR) reflecting the variability between 2 time points to the interpatient range was calculated. Thereafter, we determined the stable time range that met both "DR greater than 0.90" and "no statistically significant difference" between all time point pairs for each feature. The degree of variability across all scan delay times was obtained using coefficients of variation.

RESULTS

Standard deviation, variance, entropy, sphericity, discrete compactness, gray-level cooccurrence matrix (GLCM) inverse difference moment (IDM), GLCM contrast, and GLCM entropy did not show significant differences between scan delays of 30 and 180 seconds with DR greater than 0.90 between all time point pairs. When the range was narrowed down to 60 to 150 seconds, an additional 2 values (mean and homogeneity) showed stability. Among the 13 texture features, entropy, sphericity, discrete compactness, and GLCM entropy exhibited the lowest variability (coefficient of variation ≤5%).

CONCLUSIONS

Most texture features exhibited stability with low variation between 60 and 150 seconds on DCE-CT. Thus, texture features extracted from contrast-enhanced CT with a scan delay range of 60 to 150 seconds can be used for tumor characterization despite the heterogeneity in delay time.

Intra-observer and inter-observer agreements for the measurement of dual-input whole tumor computed tomography perfusion in patients with lung cancer: Influences of the size and inner-air density of tumors

Background

This study was conducted to assess intra‐observer and inter‐observer agreements for the measurement of dual‐input whole tumor computed tomography perfusion (DCTP) in patients with lung cancer.

Methods

A total of 88 patients who had undergone DCTP, which had proved a diagnosis of primary lung cancer, were divided into two groups: (i) nodules (diameter ≤3 cm) and masses (diameter >3 cm) by size, and (ii) tumors with and without air density. Pulmonary flow, bronchial flow, and pulmonary index were measured in each group. Intra‐observer and inter‐observer agreements for measurement were assessed using intraclass correlation coefficient, within‐subject coefficient of variation, and Bland–Altman analysis.

Results

In all lung cancers, the reproducibility coefficient for intra‐observer agreement (range 26.1–38.3%) was superior to inter‐observer agreement (range 38.1–81.2%). Further analysis revealed lower agreements for nodules compared to masses. Additionally, inner‐air density reduced both agreements for lung cancer.

Conclusion

The intra‐observer agreement for measuring lung cancer DCTP was satisfied, while the inter‐observer agreement was limited. The effects of tumoral size and inner‐air density to agreements, especially between two observers, should be emphasized. In future, an automatic computer‐aided segment of perfusion value of the tumor should be developed.

Computerized margin and texture analyses for differentiating bacterial pneumonia and invasive mucinous adenocarcinoma presenting as consolidation

Radiologists have used margin characteristics based on routine visual analysis; however, the attenuation changes at the margin of the lesion on CT images have not been quantitatively assessed. We established a CT-based margin analysis method by comparing a target lesion with normal lung attenuation, drawing a slope to represent the attenuation changes. This approach was applied to patients with invasive mucinous adenocarcinoma (n = 40) or bacterial pneumonia (n = 30). Correlations among multiple regions of interest (ROIs) were obtained using intraclass correlation coefficient (ICC) values. CT visual assessment, margin and texture parameters were compared for differentiating the two disease entities. The attenuation and margin parameters in multiple ROIs showed excellent ICC values. Attenuation slopes obtained at the margins revealed a difference between invasive mucinous adenocarcinoma and pneumonia (P<0.001), and mucinous adenocarcinoma produced a sharply declining attenuation slope. On multivariable logistic regression analysis, pneumonia had an ill-defined margin (odds ratio (OR), 4.84; 95% confidence interval (CI), 1.26–18.52; P = 0.02), ground-glass opacity (OR, 8.55; 95% CI, 2.09–34.95; P = 0.003), and gradually declining attenuation at the margin (OR, 12.63; 95% CI, 2.77–57.51, P = 0.001). CT-based margin analysis method has a potential to act as an imaging parameter for differentiating invasive mucinous adenocarcinoma and bacterial pneumonia.

Lung nodule malignancy classification using only radiologist-quantified image features as inputs to statistical learning algorithms: probing the Lung Image Database Consortium dataset with two statistical learning methods

In the assessment of nodules in CT scans of the lungs, a number of image-derived features are diagnostically relevant. Currently, many of these features are defined only qualitatively, so they are difficult to quantify from first principles. Nevertheless, these features (through their qualitative definitions and interpretations thereof) are often quantified via a variety of mathematical methods for the purpose of computer-aided diagnosis (CAD). To determine the potential usefulness of quantified diagnostic image features as inputs to a CAD system, we investigate the predictive capability of statistical learning methods for classifying nodule malignancy. We utilize the Lung Image Database Consortium dataset and only employ the radiologist-assigned diagnostic feature values for the lung nodules therein, as well as our derived estimates of the diameter and volume of the nodules from the radiologists' annotations. We calculate theoretical upper bounds on the classification accuracy that are achievable by an ideal classifier that only uses the radiologist-assigned feature values, and we obtain an accuracy of 85.74 [Formula: see text], which is, on average, 4.43% below the theoretical maximum of 90.17%. The corresponding area-under-the-curve (AUC) score is 0.932 ([Formula: see text]), which increases to 0.949 ([Formula: see text]) when diameter and volume features are included and has an accuracy of 88.08 [Formula: see text]. Our results are comparable to those in the literature that use algorithmically derived image-based features, which supports our hypothesis that lung nodules can be classified as malignant or benign using only quantified, diagnostic image features, and indicates the competitiveness of this approach. We also analyze how the classification accuracy depends on specific features and feature subsets, and we rank the features according to their predictive power, statistically demonstrating the top four to be spiculation, lobulation, subtlety, and calcification.

Impact of Reconstruction Algorithms on CT Radiomic Features of Pulmonary Tumors: Analysis of Intra- and Inter-Reader Variability and Inter-Reconstruction Algorithm Variability

PURPOSE

To identify the impact of reconstruction algorithms on CT radiomic features of pulmonary tumors and to reveal and compare the intra- and inter-reader and inter-reconstruction algorithm variability of each feature.

METHODS

Forty-two patients (M:F = 19:23; mean age, 60.43±10.56 years) with 42 pulmonary tumors (22.56±8.51mm) underwent contrast-enhanced CT scans, which were reconstructed with filtered back projection and commercial iterative reconstruction algorithm (level 3 and 5). Two readers independently segmented the whole tumor volume. Fifteen radiomic features were extracted and compared among reconstruction algorithms. Intra- and inter-reader variability and inter-reconstruction algorithm variability were calculated using coefficients of variation (CVs) and then compared.

RESULTS

Among the 15 features, 5 first-order tumor intensity features and 4 gray level co-occurrence matrix (GLCM)-based features showed significant differences (p<0.05) among reconstruction algorithms. As for the variability, effective diameter, sphericity, entropy, and GLCM entropy were the most robust features (CV≤5%). Inter-reader variability was larger than intra-reader or inter-reconstruction algorithm variability in 9 features. However, for entropy, homogeneity, and 4 GLCM-based features, inter-reconstruction algorithm variability was significantly greater than inter-reader variability (p<0.013).

CONCLUSIONS

Most of the radiomic features were significantly affected by the reconstruction algorithms. Inter-reconstruction algorithm variability was greater than inter-reader variability for entropy, homogeneity, and GLCM-based features.

Measuring Computed Tomography Scanner Variability of Radiomics Features

OBJECTIVES

The purpose of this study was to determine the significance of interscanner variability in CT image radiomics studies.

MATERIALS AND METHODS

We compared the radiomics features calculated for non-small cell lung cancer (NSCLC) tumors from 20 patients with those calculated for 17 scans of a specially designed radiomics phantom. The phantom comprised 10 cartridges, each filled with different materials to produce a wide range of radiomics feature values. The scans were acquired using General Electric, Philips, Siemens, and Toshiba scanners from 4 medical centers using their routine thoracic imaging protocol. The radiomics feature studied included the mean and standard deviations of the CT numbers as well as textures derived from the neighborhood gray-tone difference matrix. To quantify the significance of the interscanner variability, we introduced the metric feature noise. To look for patterns in the scans, we performed hierarchical clustering for each cartridge.

RESULTS

The mean CT numbers for the 17 CT scans of the phantom cartridges spanned from -864 to 652 Hounsfield units compared with a span of -186 to 35 Hounsfield units for the CT scans of the NSCLC tumors, showing that the phantom's dynamic range includes that of the tumors. The interscanner variability of the feature values depended on both the cartridge material and the feature, and the variability was large relative to the interpatient variability in the NSCLC tumors for some features. The feature interscanner noise was greatest for busyness and least for texture strength. Hierarchical clustering produced different clusters of the phantom scans for each cartridge, although there was some consistent clustering by scanner manufacturer.

CONCLUSIONS

The variability in the values of radiomics features calculated on CT images from different CT scanners can be comparable to the variability in these features found in CT images of NSCLC tumors. These interscanner differences should be considered, and their effects should be minimized in future radiomics studies.

Pulmonary imaging after stereotactic radiotherapy-does RECIST still apply?

The use of stereotactic ablative radiotherapy (SABR) for the treatment of primary lung cancer and metastatic disease is rapidly increasing. However, the presence of benign fibrotic changes on CT imaging makes response assessment following SABR a challenge, as these changes develop with an appearance similar to tumour recurrence. Misclassification of benign fibrosis as local recurrence has resulted in unnecessary interventions, including biopsy and surgical resection. Response evaluation criteria in solid tumours (RECIST) are widely used as a universal set of guidelines to assess tumour response following treatment. However, in the context of non-spherical and irregular post-SABR fibrotic changes, the RECIST criteria can have several limitations. Positron emission tomography can also play a role in response assessment following SABR; however, false-positive results in regions of inflammatory lung post-SABR can be a major clinical issue and optimal standardized uptake values to distinguish fibrosis and recurrence have not been determined. Although validated CT high-risk features show a high sensitivity and specificity for predicting recurrence, most recurrences are not detected until more than 1-year post-treatment. Advanced quantitative radiomic analysis on CT imaging has demonstrated promise in distinguishing benign fibrotic changes from local recurrence at earlier time points, and more accurately, than physician assessment. Overall, the use of RECIST alone may prove inferior to novel metrics of assessing response.

CT-based radiomic analysis of stereotactic body radiation therapy patients with lung cancer

BACKGROUND

Radiomics uses a large number of quantitative imaging features that describe the tumor phenotype to develop imaging biomarkers for clinical outcomes. Radiomic analysis of pre-treatment computed-tomography (CT) scans was investigated to identify imaging predictors of clinical outcomes in early stage non-small cell lung cancer (NSCLC) patients treated with stereotactic body radiation therapy (SBRT).

MATERIALS AND METHODS

CT images of 113 stage I-II NSCLC patients treated with SBRT were analyzed. Twelve radiomic features were selected based on stability and variance. The association of features with clinical outcomes and their prognostic value (using the concordance index (CI)) was evaluated. Radiomic features were compared with conventional imaging metrics (tumor volume and diameter) and clinical parameters.

RESULTS

Overall survival was associated with two conventional features (volume and diameter) and two radiomic features (LoG 3D run low gray level short run emphasis and stats median). One radiomic feature (Wavelet LLH stats range) was significantly prognostic for distant metastasis (CI=0.67, q-value<0.1), while none of the conventional and clinical parameters were. Three conventional and four radiomic features were prognostic for overall survival.

CONCLUSION

This exploratory analysis demonstrates that radiomic features have potential to be prognostic for some outcomes that conventional imaging metrics cannot predict in SBRT patients.