The method and efficacy of support vector machine classifiers based on texture features and multi-resolution histogram from 18F-FDG PET-CT images for the evaluation of mediastinal lymph nodes in patients with lung cancer

doi:10.1016/j.ejrad.2014.11.006

European Journal of Radiology

Volume 84, Issue 2, February 2015, Pages 312-317

https://doi.org/10.1016/j.ejrad.2014.11.006 Get rights and content

Highlights

•
Three support vector machine classifiers were constructed from PET-CT images.
•
The areas under the ROC curve for SVM1, SVM2, and SVM3 were 0.689, 0.579, and 0.685, respectively.
•
The areas under curves for maximum short diameter and SUV_max were 0.684 and 0.652, respectively.
•
The algorithm based on SVM was potential in the diagnosis of mediastinal lymph nodes.

Abstract

Objectives

In clinical practice, image analysis is dependent on simply visual perception and the diagnostic efficacy of this analysis pattern is limited for mediastinal lymph nodes in patients with lung cancer. In order to improve diagnostic efficacy, we developed a new computer-based algorithm and tested its diagnostic efficacy.

Methods

132 consecutive patients with lung cancer underwent ¹⁸F-FDG PET/CT examination before treatment. After all data were imported into the database of an on-line medical image analysis platform, the diagnostic efficacy of visual analysis was first evaluated without knowing pathological results, and the maximum short diameter and maximum standardized uptake value (SUV_max) were measured. Then lymph nodes were segmented manually. Three classifiers based on support vector machine (SVM) were constructed from CT, PET, and combined PET-CT images, respectively. The diagnostic efficacy of SVM classifiers was obtained and evaluated.

Results

According to ROC curves, the areas under curves for maximum short diameter and SUV_max were 0.684 and 0.652, respectively. The areas under the ROC curve for SVM1, SVM2, and SVM3 were 0.689, 0.579, and 0.685, respectively.

Conclusion

The algorithm based on SVM was potential in the diagnosis of mediastinal lymph nodes.

Introduction

For patients newly diagnosed with lung cancer, the choice of appreciate treatment was largely dependent on the exact evaluation of pathological status of mediastinal lymph nodes. Currently, in clinical practice, various kinds of imaging methods, such as CT, MR, and ¹⁸F-FDG PET-CT, are widely used to evaluate lymph nodes. In the usual radiologically diagnostic procedure, information from images, such as location, size and shape of lymph nodes, was visually perceived, quantitatively measured, and subjectively integrated with clinical information. Finally, with some confidence, radiologists come to a conclusion about the pathological condition of mediastinal lymph nodes. This analysis pattern was natural to our mental procedure and its diagnostic efficacy for mediastinal lymph nodes has been well established [1], [2], [3], [4]. It is obvious that in this routine analysis pattern, relatively superficial and easily accessible information was fully taken advantage. Although this kind of straightforward information analysis pattern has greatly changed our clinical practice, it is still reasonable to ask whether we had made full use of the radiological information obtained at risk from radiation. In fact, computers can quantitatively extract much information that cannot be visually perceived by people, and in order to increase the information obtainable from medical images, many attempts have been made in this field belonging to the computer-aided diagnosis (CAD) technology. CAD technology has made a great progress in the fields such as lesion detection, lesion 3D volumetry, and lesion qualitative diagnosis. The enormous potential of CAD technology has been widely proved by continually increasing clinical application. However, in the field of qualitative diagnosis for lesions, ideal diagnostic accuracy is still hard to reach. In the field of qualitative lesion analysis, most algorithms focused on the traditional radiological signs that came from the straightforward visual perception. Although much effort has been made on such as demarcation and quantization of various radiological signs, these methods did not significantly improve the diagnostic efficacy [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19]. It is obvious that the biological essence of various radiological findings cannot be simply reflected by the artifact “radiological signs”, which were commonly established from human visual perception. However, from a different perspective, according to the principle of CT, we know that CT attenuation values and their change can reflect the density distribution of tissues in the human body. However, due to the common window display technique simplifies CT attenuation value distribution it is hard to disclose the associated distribution information only by virtue of visual analysis. Meanwhile, for ¹⁸F-FDG PET-CT, general clinical analysis can only visually recognize the radioactive distribution and cannot proceed further. In order to mine more image information beyond visual perception, in this study, we developed a new algorithm, which combined image texture parameters, and support vector machine technique, to further improve diagnosis efficacy of metastases of mediastinal lymph node.

Section snippets

Patients

In this study, we evaluated 132 consecutive patients with lung cancer between June 2009 and July 2013. All patients underwent ¹⁸F-FDG PET/CT examination as a part of workup before surgery and did not receive any therapy for tumor. The interval between ¹⁸F-FDG PET/CT examination and operation was within one week. All patients underwent lobectomy combined with systematic hilar and mediastinal lymph node dissection in our institute. The pathological results were retrieved from the medical

The diagnostic results from radiologists

In these 132 patients, a total of 768 lymph nodes were visually identified. In these nodes, the mean maximum short diameter was 6.31 ± 2.65 mm, ranging from 4 mm to 27 mm, and mean SUV_max was 2.13 ± 1.67, ranging from 0.4 to 18.1. The detailed distribution of maximum short diameter and SUV_max was listed in Table 2.

The diagnostic efficiency of maximum short diameter and SUV_max was evaluated by receiver operating characteristic (ROC) curve, which was listed in Fig. 3. The area under curve for maximum

The diagnostic results from SVMS

In order to construct the corresponding SVMs, we randomly chose a train sample set of 30 lymph nodes with even distribution of benign and malignant lymph nodes. Based on the results of train set, we got the corresponding best diagnostic efficacy of SVMs (SVM1, SVM2, and SVM3, respectively) derived from PET images, CT images, and combined PET + CT images, respectively, which were showed in Table 3, Table 4.

Discussion

From CT images, what we could obtain directly about mediastinal lymph nodes were the morphology, size and attenuation. For the attenuation, in routine clinical practice of radiological diagnosis, the common information we got by qualitative means was the mean attenuation value around some region on the image; we seldom qualitatively exploited information about the spatial variation of attenuation in human body for diagnosis purpose. However, current preliminary attempts on the characterization

Conflict of interest

None declared.

Acknowledgements

This research was supported by the general program of National Natural Science Fund of China (Grant No. 81171405) and the research program of Science.

References (25)

Y. Wu et al.
Diagnostic value of fluorine 18 fluorodeoxyglucose positron emission tomography/computed tomography for the detection of metastases in non-small-cell lung cancer patients
Int J Cancer
(2013)
R. Maeda et al.
The maximum standardized ¹⁸F-fluorodeoxyglucose uptake on positron emission tomography predicts lymph node metastasis and invasiveness in clinical stage IA non-small cell lung cancer
Interact Cardiovasc Thorac Surg
(2009)
T. Takenaka et al.
Prediction of true-negative lymph node metastasis in clinical IA non-small cell lung cancer by measuring standardized uptake values on positron emission tomography
Surg Today
(2012)
S. Sachs et al.
Increased standardized uptake value in the primary lesion predicts nodal or distant metastases at presentation in lung cancer
Clin Lung Cancer
(2005)
Y. Kawata et al.
Computer aided differential diagnosis of pulmonary nodules using curvature based analysis
J. Shiraishi et al.
Computer-aided diagnosis to distinguish benign from malignant solitary pulmonary nodules on radiographs
Radiology
(2003)
L. Boroczky et al.
Feature subset selection for improving the performance of false positive reduction in lung nodule cad
IEEE Trans Inf Technol Biomed
(2006)
D. Yankelevitz et al.
Small pulmonary nodules: volumetrically determined growth rates based on CT evaluation
Radiology
(2000)
M. Revel et al.
Pulmonary nodules: preliminary experience with three-dimensional evaluation
Radiology
(2004)
W. Kostis et al.
Small pulmonary nodules: reproducibility of 3D volumetric measurement and estimation of time to follow-up CT
Radiology
(2004)

F. Li et al.

Radiologists’ performance for differentiating benign from malignant lung nodules on high-resolution CT using computer-estimated likelihood of malignancy

Am J Roentgenol

(2004)

S. Mehta et al.

Evaluation of voxel-based morphometry for focal lesion detection in individuals

Neuroimage

(2003)

Cited by (78)

Applications of machine learning and deep learning in SPECT and PET imaging: General overview, challenges and future prospects
2023, Pharmacological Research
The integration of positron emission tomography (PET) and single-photon emission computed tomography (SPECT) imaging techniques with machine learning (ML) algorithms, including deep learning (DL) models, is a promising approach. This integration enhances the precision and efficiency of current diagnostic and treatment strategies while offering invaluable insights into disease mechanisms. In this comprehensive review, we delve into the transformative impact of ML and DL in this domain. Firstly, a brief analysis is provided of how these algorithms have evolved and which are the most widely applied in this domain. Their different potential applications in nuclear imaging are then discussed, such as optimization of image adquisition or reconstruction, biomarkers identification, multimodal fusion and the development of diagnostic, prognostic, and disease progression evaluation systems. This is because they are able to analyse complex patterns and relationships within imaging data, as well as extracting quantitative and objective measures. Furthermore, we discuss the challenges in implementation, such as data standardization and limited sample sizes, and explore the clinical opportunities and future horizons, including data augmentation and explainable AI. Together, these factors are propelling the continuous advancement of more robust, transparent, and reliable systems.
[<sup>18</sup>F]FDG-PET/CT Radiomics and Artificial Intelligence in Lung Cancer: Technical Aspects and Potential Clinical Applications
2022, Seminars in Nuclear Medicine
Lung cancer is the second most common cancer and the leading cause of cancer-related death worldwide. Molecular imaging using [¹⁸F]fluorodeoxyglucose Positron Emission Tomography and/or Computed Tomography ([¹⁸F]FDG-PET/CT) plays an essential role in the diagnosis, evaluation of response to treatment, and prediction of outcomes. The images are evaluated using qualitative and conventional quantitative indices. However, there is far more information embedded in the images, which can be extracted by sophisticated algorithms. Recently, the concept of uncovering and analyzing the invisible data extracted from medical images, called radiomics, is gaining more attention. Currently, [¹⁸F]FDG-PET/CT radiomics is growingly evaluated in lung cancer to discover if it enhances the diagnostic performance or implication of [¹⁸F]FDG-PET/CT in the management of lung cancer. In this review, we provide a short overview of the technical aspects, as they are discussed in different articles of this special issue. We mainly focus on the diagnostic performance of the [¹⁸F]FDG-PET/CT‐based radiomics and the role of artificial intelligence in non-small cell lung cancer, impacting the early detection, staging, prediction of tumor subtypes, biomarkers, and patient's outcomes.
An Artificial Intelligence Algorithm to Predict Nodal Metastasis in Lung Cancer
2022, Annals of Thoracic Surgery
Endobronchial ultrasound (EBUS) has features that allow a high accuracy for predicting lymph node (LN) malignancy. However their clinical application remains limited because of high operator dependency. We hypothesized that an artificial intelligence algorithm (NeuralSeg; NeuralSeg Ltd, Hamilton, Ontario, Canada) is capable of accurately identifying and predicting LN malignancy based on EBUS images.
In the derivation phase EBUS images were segmented twice by an endosonographer and used as controls in 5-fold cross-validation training of NeuralSeg. In the validation phase the algorithm was tested on new images it had not seen before. Logistic regression and receiver operator characteristic curves were used to determine NeuralSeg’s capability of discrimination between benign and malignant LNs, using pathologic specimens as the gold standard.
Two hundred ninety-eight LNs from 140 patients were used for derivation and 108 LNs from 47 patients for validation. In the derivation cohort NeuralSeg was able to predict malignant LNs with an accuracy of 73.8% (95% confidence interval [CI], 68.4%-78.7%). In the validation cohort NeuralSeg had an accuracy of 72.9% (95% CI, 63.5%-81.0%), specificity of 90.8% (95% CI, 81.9%-96.2%), and negative predictive value of 75.9% (95% CI, 71.5%-79.9%). NeuralSeg showed higher diagnostic discrimination during validation compared with derivation (c-statistic = 0.75 [95% CI, 0.65-0.85] vs 0.63 [95% CI, 0.54-0.72], respectively).
NeuralSeg is able to accurately rule out nodal metastasis and can possibly be used as an adjunct to EBUS when nodal biopsy is not possible or inconclusive. Future work to evaluate the algorithm in a clinical trial is required.
Radiogenomics in personalized management of lung cancer patients: Where are we?
2022, Clinical Imaging
Citation Excerpt :
Some studies have reported that this perinodular tissue might harbor a different microenvironment from healthy parenchyma that could be used in radiomic models to distinguish benign from malignant lesions43 or to predict response to chemotherapy,44 survival outcomes,45 or tumoral spread through air spaces in lung cancer.46 Further applications of AI predictive models could be directed to predict mediastinal lymph node invasion and distant metastases.47,48 Recent investigations have also explored the utility of incorporating nucleic acids sequencing data to radiogenomics models.
With the rise of artificial intelligence, radiomics has emerged as a field of translational research based on the extraction of mineable high-dimensional data from radiological images to create “big data” datasets for the purpose of identifying distinct sub-visual imaging patterns. The integrated analysis of radiomic data and genomic data is termed radiogenomics, a promising strategy to identify potential imaging biomarkers for predicting driver mutations and other genomic parameters. In lung cancer, recent advances in whole-genome sequencing and the identification of actionable molecular alterations have led to an increased interest in understanding the complex relationships between imaging and genomic data, with the potential of guiding therapeutic strategies and predicting clinical outcomes. Although the integration of the radiogenomics data into lung cancer management may represent a new paradigm in the field, the use of this technique as a clinical biomarker remains investigational and still necessitates standardization and robustness to be effectively translated into the clinical practice. This review summarizes the basic concepts, potential contributions, challenges, and opportunities of radiogenomics in the management of patients with lung cancer.
A sustainable IoHT based computationally intelligent healthcare monitoring system for lung cancer risk detection
2021, Sustainable Cities and Society
A sustainable healthcare focuses on enhancing and restoring public health parameters thereby reducing gloomy impacts on social, economic and environmental elements of a sustainable city. Though it has uplifted public health, yet the rise of chronic diseases is a concern in sustainable cities. In this work, a sustainable lung cancer detection model is developed to integrate the Internet of Health Things (IoHT) and computational intelligence, causing the least harm to the environment. IoHT unit retains connectivity continuously generates data from patients. Heuristic Greedy Best First Search (GBFS) algorithm is used to select most relevant attributes of lung cancer data upon which random forest algorithm is applied to classify and differentiates lung cancer affected patients from normal ones based on detected symptoms. It is observed during the experiment that the GBFS-Random forest model shows a promising outcome. While an optimal accuracy of 98.8 % was generated, simultaneously, the least latency of 1.16 s was noted. Specificity and sensitivity recorded with the proposed model on lung cancer data are 97.5 % and 97.8 %, respectively. The mean accuracy, specificity, sensitivity, and f-score value recorded is 96.96 %, 96.26 %, 96.34 %, and 96.32 %, respectively, over various types of cancer datasets implemented. The developed smart and intelligent model is sustainable. It reduces unnecessary manual overheads, safe, preserves resources and human resources, and assists medical professionals in quick and reliable decision making on lung cancer diagnosis.
Complete fully automatic segmentation and 3-dimensional measurement of mediastinal lymph nodes for a new response evaluation criteria for solid tumors
2021, Biocybernetics and Biomedical Engineering
Citation Excerpt :
Wang at el. [37] and Gao at el. [34] achieved similar accuracy of 86% and their obtained specificity are less than compared to our proposed method, but they both achieve more sensitivity than ours.
We aimed to automatically detect and segment mediastinal lymph nodes, and to establish an objective method and reliable new response evaluation criteria to monitor the effectiveness of cancer treatment. The image processing techniques were applied and developed 3D measurement of the mediastinal lymph nodes based on automatic settings when segmenting the lymph node image. A repeatable and consistent lymph node evaluation system was created based on such features as the position of occurrence, grayness, and number of serial sections of lymph nodes. A total of 200 lymph node samples from Tri-Service General Hospital, Taiwan, were examined for statistical analysis. The proposed approach used weighted k-nearest neighbors for classification, achieving superior results with an accuracy and specificity of 97.5% and 99.4%, respectively. The volume of the lymph nodes was used as the reference index for tumor invasiveness evaluation. The error in the lymph node volume was 1.71% according to the verification results. Receiver operating characteristic (ROC) curves for each analysis were constructed and the area under the curve (AUC) was calculated with histopathology diagnosis as outcome for determining the optimal volume threshold of benign and malignant lymph nodes. It was observed that the lymph node volume was highly correlated with tumor invasion (p-value was less than 0.05). The experiment showed that the volume for the area under the ROC curve was 0.90 of tumor invasion evaluation. The lymph node volume was most effective in predicting tumor invasiveness, with the value 798.53 mm³ used as the standard for judging benignity and malignancy.

View all citing articles on Scopus

^☆: This research was supported by the general program of National Natural Science Fund of China (Grant No. 81171405).

View full text

The method and efficacy of support vector machine classifiers based on texture features and multi-resolution histogram from 18F-FDG PET-CT images for the evaluation of mediastinal lymph nodes in patients with lung cancer☆

Highlights

Abstract

Objectives

Methods

Results

Conclusion

Introduction

Section snippets

Patients

The diagnostic results from radiologists

The diagnostic results from SVMS

Discussion

Conflict of interest

Acknowledgements

Diagnostic value of fluorine 18 fluorodeoxyglucose positron emission tomography/computed tomography for the detection of metastases in non-small-cell lung cancer patients

Int J Cancer

The maximum standardized 18F-fluorodeoxyglucose uptake on positron emission tomography predicts lymph node metastasis and invasiveness in clinical stage IA non-small cell lung cancer

Interact Cardiovasc Thorac Surg

Prediction of true-negative lymph node metastasis in clinical IA non-small cell lung cancer by measuring standardized uptake values on positron emission tomography

Surg Today

Increased standardized uptake value in the primary lesion predicts nodal or distant metastases at presentation in lung cancer

Clin Lung Cancer

Computer aided differential diagnosis of pulmonary nodules using curvature based analysis

Computer-aided diagnosis to distinguish benign from malignant solitary pulmonary nodules on radiographs

Radiology

Feature subset selection for improving the performance of false positive reduction in lung nodule cad

IEEE Trans Inf Technol Biomed

Small pulmonary nodules: volumetrically determined growth rates based on CT evaluation

Radiology

Pulmonary nodules: preliminary experience with three-dimensional evaluation

Radiology

Small pulmonary nodules: reproducibility of 3D volumetric measurement and estimation of time to follow-up CT

Radiology

Radiologists’ performance for differentiating benign from malignant lung nodules on high-resolution CT using computer-estimated likelihood of malignancy

Am J Roentgenol

Evaluation of voxel-based morphometry for focal lesion detection in individuals

Neuroimage

The method and efficacy of support vector machine classifiers based on texture features and multi-resolution histogram from ¹⁸F-FDG PET-CT images for the evaluation of mediastinal lymph nodes in patients with lung cancer☆

The maximum standardized ¹⁸F-fluorodeoxyglucose uptake on positron emission tomography predicts lymph node metastasis and invasiveness in clinical stage IA non-small cell lung cancer