Identifying EGFR mutations in lung adenocarcinoma by noninvasive imaging using radiomics features and random forest modeling.

OBJECTIVES: The tyrosine kinase inhibitor (TKI)-sensitive mutations of the epidermal growth factor receptor (EGFR) gene is essential in the treatment of lung adenocarcinoma. To overcome the difficulty of EGFR gene test in situations where surgery and biopsy samples are too risky to obtain, we tried a noninvasive imaging method using radiomics features and random forest models. METHODS: Five hundred three lung adenocarcinoma patients who received surgery-based treatment were included in this study. The diagnosis and EGFR gene test were based on resections. TKI-sensitive mutations were found in 60.8% of the patients. CT scans before any invasive operation were gathered and analyzed to extract quantitative radiomics features and build random forest classifiers to identify EGFR mutants from wild types. Clinical features (sex and smoking history) were added to the image-based model. The model was trained on a set of 345 patients and validated on an independent test group (n = 158) using the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. RESULTS: The performance of the random forest model with 94 radiomics features reached an AUC of 0.802. Its AUC was further improved to 0.828 by adding sex and smoking history. The sensitivity and specificity are 60.6% and 85.1% at the best diagnostic decision point. CONCLUSION: Our results showed that radiomics could not only reflect the genetic differences among tumors but also have diagnostic value and the potential to be a diagnostic tool. KEY POINTS: • Radiomics provides a potential noninvasive method for the prediction of EGFR mutation status. • In situations where surgeries and biopsy are not available, CT image-based radiomics models could help to make treatment decisions. • The accuracy, sensitivity, and specificity still need to be improved before the image-based EGFR identifier could be used in clinics.

Automatic Delineation of the Clinical Target Volume in Rectal Cancer for Radiation Therapy using Three-dimensional Fully Convolutional Neural Networks.

Accurate, robust, and fast delineation of the clinical target volume (CTV) for the use in radiotherapy of rectal cancer (RC) is highly sought-after. Convolutional neural networks (CNNs) have proven themselves very effective in various segmentation tasks on medical images. Despite this, their application in CTV delineation is not yet fully explored. This study uses the three-dimensional fully convolutional neural network architecture called V-net for CTV delineation. The West China Hospital (Chengdu, China) provided this study with 120 annotated CT scans. For improved performance and to battle data scarcity, the available scans were augmented. Trained on 100 CT-scans for 20 hours and tested on 20 previously unseen CT-scans the network achieved a mean dice similarity coefficient (DSC) of 0.90 and a mean delineation time per CTV of 0.60 seconds. The proposed method is compared with two other state-of-the-art CNNs and is shown to be superior.

Identifying epidermal growth factor receptor mutation status in patients with lung adenocarcinoma by three-dimensional convolutional neural networks.

OBJECTIVE:: Genetic phenotype plays a central role in making treatment decisions of lung adenocarcinoma, especially the tyrosine-kinase-inhibitors-sensitive mutations of the epidermal growth factor receptor (EGFR) gene. We constructed three-dimensional convolutional neural networks (CNN) to analyze underlying patterns in CT images that could indicate that EGFR gene mutation status but are invisible to human eyes. METHODS:: From 2012 to 2015, 503 Chinese patients with lung adenocarcinoma that had underwent surgery were included. Pathological types and EGFR mutation status were tested from surgical resections. EGFR mutations (exon 19 deletion or exon 21 L858R) were found in 215/345 (62.3%) and 91/158 (57.6%) patients in the training and independent validation set, respectively. CT images were taken before any invasive operation. The patients were randomly chosen to train the CNNs or validate the CNNs' performance. The performance was quantified using area under receiver operating characteristic curve (AUC), sensitivity, specificity, and accuracy. RESULTS:: The CNNs showed an AUC of 0.776 (range: 0.702-0.849, p< 0.0001) in the independent validation set and a fusion model of CNNs and clinical features (sex and smoking history) showed an AUC of 0.838 (range: 0.778-0.899, p< 0.0001), accuracy of 77.2%, sensitivity of 75.8% and specificity of 79.1% at the best diagnostic decision point. CONCLUSION:: The CNN exhibits potential ability to identify EGFR mutation status in patients with lung adenocarcinoma which might help make clinical decisions. ADVANCES IN KNOWLEDGE:: The CNN showed some diagnostic power and its performance could be further improved by increasing the training set, optimizing the network structure and training strategy. Medical image based CNN has the potential to reflect spatial heterogeneity.

Detection of epithelial growth factor receptor (EGFR) mutations on CT images of patients with lung adenocarcinoma using radiomics and/or multi-level residual convolutionary neural networks.

BACKGROUND: We aim to analyze the ability to detect epithelial growth factor receptor (EGFR) mutations on chest CT images of patients with lung adenocarcinoma using radiomics and/or multi-level residual convolutionary neural networks (MCNNs). METHODS: We retrospectively collected 1,010 consecutive patients in Shanghai Chest Hospital from 2013 to 2017, among which 510 patients were EGFR-mutated and 500 patients were wild-type. The patients were randomly divided into a training set (810 patients) and a validation set (200 patients) according to a balanced distribution of clinical features. The CT images and the corresponding EGFR status measured by Amplification Refractory Mutation System (ARMS) method of the patients in the training set were utilized to construct both a radiomics-based model (MRadiomics) and MCNNs-based model (MMCNNs). The MRadiomics and MMCNNs were combined to build the ModelRadiomics+MCNNs (MRadiomics+MCNNs). Clinical data of gender and smoking history constructed the clinical features-based model (MClinical). MClinical was then added into MRadiomics, MMCNNs, and MRadiomics+MCNNs to establish the ModelRadiomics+Clinical (MRadiomics+Clinical), the ModelMCNNs+Clinical (MMCNNs+Clinical) and the ModelRadiomics+MCNNs+Clinical (MRadiomics+MCNNs+Clinical). All the seven models were tested in the validation set to ascertain whether they were competent to detect EGFR mutations. The detection efficiency of each model was also compared in terms of area under the curve (AUC), sensitivity and specificity. RESULTS: The AUC of the MRadiomics, MMCNNs and MRadiomics+MCNNs to predict EGFR mutations was 0.740, 0.810 and 0.811 respectively. The performance of MMCNNs was better than that of MRadiomics (P=0.0225). The addition of clinical features did not improve the AUC of the MRadiomics (P=0.623), the MMCNNs (P=0.114) and the MRadiomics+MCNNs (P=0.058). The MRadiomics+MCNNs+Clinical demonstrated the highest AUC value of 0.834. The MMCNNs did not demonstrate any inferiority when compared with the MRadiomics+MCNNs (P=0.742) and the MRadiomics+MCNNs+Clinical (P=0.056). CONCLUSIONS: Both of the MRadiomics and the MCNNs could predict EGFR mutations on CT images of patients with lung adenocarcinoma. The MMCNNs outperformed the MRadiomics in the detection of EGFR mutations. The combination of these two models, even added with clinical features, is not significantly more efficient than MMCNNs alone.

[Estimating Soil Iron Content Based on Reflectance Spectra].

In recent decades, the application of spectral technology in soil science is getting more and more attention. Soil information can be obtained quickly by using soil reflectance spectra to understand the physical and chemical properties of soil and to estimate soil iron content. In previous studies, the surface soil always is selected for the estimation of soil iron content by using spectroscopy. It needs to estimate total iron and, the different forms of soil iron is ignored, therefore, the estimation result is not ideal. In order to gets a different form of soil iron processing method of optimal model to evaluate the accuracy of models, as well as discuss the organic matter content and soil depth on the influence of different forms of soil iron estimation accuracy. A total of 160 soil samples were collected from 20 sites in Dongtai city, Jiangsu province. These samples were ground to 10 meshes and 100 meshes. In the use of 8 different methods for the pretreatment of the same time each method will be selected by a variety of parameters, using partial least squares regression method to model the total reflection band and the total iron, free iron, amorphous iron content in the soil respectively, then evaluation model precision. The results showed that: (1) the optimal model of three kinds of soil iron was all ground to 100 meshes and the best pretreatment method was MSC. The prediction accuracy of total iron was acceptable and R2 was less than 0.6. The results of free iron and amorphous iron inversion were better and the R2 was 0.77 and 0.69, respectively. The errors were small and the models were stable. (2) Because the ferric metasilicate in total iron is easily affected by external environment, the organic matter and soil depth are of great influence on the estimate precision of total iron the most. But the estimation accuracy of free iron is the least affected. Because of the low content of amorphous iron, the estimated model is also susceptible to the influence of organic matter and soil depth.

A Mechanism Study of Reflectance Spectroscopy for Predicting Soil Total Nitrogen.

Reflectance spectroscopy has been widely used for predicting soil properties due to its rapidity and convenience. In past decades, the application of soil spectroscopy on soil science studies has increased exponentially. The total nitrogen (TN) content in soil is an important index for soil fertility and the rapid prediction of TN content with spectroscopy serves an important function in precision agriculture. However, whether the TN content in soil is predicted through its relationship with soil organic carbon (SOC) or on its specific absorption is still debatable. The objective of this study was to explore the mechanism of reflectance spectroscopy for predicting TN in soils. Soils used for calibration were sampled from coastal soil in the north of Jiangsu province. Partial least squares regression (PLSR) analysis was used for the calibration datasets with different TN content when the sample number is the same in every dataset. In order to explore the mechanism of reflectance spectroscopy for predicting total nitrogen in soil, the changes of model accuracies and the correlation of TN and SOC were analyzed. The results indicated that the contents of TN and SOC in soil were relatively lower because the soil was derived from coastal sediments in the past 1 000 years and formed during cultivation. There was strong correlation between TN and SOC (R=0.98). The prediction accuracy of TN increased at first and then decreased slightly with the increase of mean, standard deviation of TN content. Meanwhile, the changes of prediction accuracy comply well with coefficients of variation. In conclusion, when the TN content is relatively low (mean TN<0.27 g·kg-1), the correlation coefficient between TN and SOC was moderately-high and TN was predicted on the basis of N absorbers. When the TN content is relatively high (mean TN>0.29 g·kg-1), strong correlation coefficients were obtained for TN and SOC and the model accuracy of SOC were better than TN. The effect of SOC to spectroscopy enhanced with the increase of SOC content, which masked the spectral features of N. Therefore, TN was predicted through the correlation with SOC when the TN content is high. This study revealed the mechanism of reflectance spectroscopy for predicting TN in soil and it could provide a theoretical basis for predicting soil TN content rapidly using reflectance spectroscopy.