Sage Journals: Discover world-class research

Abstract

Objective

We tested the performance of general machine learning and joint machine learning algorithms in the classification of bone metastasis, in patients with lung adenocarcinoma.

Methods

We used R version 3.5.3 for statistical analysis of the general information, and Python to construct machine learning models.

Results

We first used the average classifiers of the 4 machine learning algorithms to rank the features and the results showed that race, sex, whether they had surgery and marriage were the first 4 factors affecting bone metastasis. Machine learning results in the training group: for area under the curve (AUC), except for RF and LR, the AUC values of all machine learning classifiers were greater than .8, but the joint algorithm did not improve the AUC for any single machine learning algorithm. Among the results related to accuracy and precision, the accuracy of other machine learning classifiers except the RF algorithm was higher than 70%, and only the precision of the LGBM algorithm was higher than 70%. Machine learning results in the test group: Similarly, for areas under the curve (AUC), except RF and LR, the AUC values for all machine learning classifiers were greater than .8, but the joint algorithm did not improve the AUC value for any single machine learning algorithm. For accuracy, except for the RF algorithm, the accuracy of other machine learning classifiers was higher than 70%. The highest precision for the LGBM algorithm was .675.

Conclusion

The results of this concept verification study show that machine learning algorithm classifiers can distinguish the bone metastasis of patients with lung cancer. This will provide a new research idea for the future use of non-invasive technology to identify bone metastasis in lungcancer. However, more prospective multicenter cohort studies are needed.

Keywords

machine learning bone metastasis lung adenocarcinoma distinguishing AUC

Introduction

Lung cancer has become 1 of the most common malignant tumors in the world, with high incidence and mortality.¹ The skeletal system is the most common metastasis site of lung cancer. Bone metastasis seriously affects the treatment and prognosis of patients with lung cancer.² Studies have shown that 15%∼40% of patients have bone metastasis when lung cancer is first diagnosed.³ Some studies have also suggested that the prevalence of bone marrow micro metastasis in patients with lung cancer is 22%∼60%.⁴ Moreover, the latest study reports that the bone metastasis rate for lung cancer patients in some areas has reached 48% for stage IV non-small cell lung cancer and 40% for extensive stage small cell lung cancer.⁵ The regional difference in bone metastasis incidence is not only correlated with disease stage, histologic type, survival time, treatment and other factors, but is also associated with diagnostic methods. Thus, it would be of great significance to diagnose bone metastasis in the early stages, and provide timely treatment. However, the current diagnosis of bone metastasis in patients with lung cancer relies on regular bone imaging screening for patients with lung cancer; Then, suspicious patients are further diagnosed by means of imaging or pathology.⁶ A study has shown that the 2 examination methods can complement each other, and bone scans can be used for general examinations. Meanwhile, MRI can locate the lesions that are easy to miss in a bone scan, especially the isolated metastatic lesions and the metastatic lesions in the spine.⁷ A study has shown that the diagnostic rate and lesion detection rate of PET/CT in patients with bone metastasis from lung cancer were much higher than those of MRI and SPECT.⁸ In addition to PET/CT and MRI, the clinical application of ultrasound is maturing.^9,10 The use of ultrasound provides a rapid and less invasive method of diagnosis and staging for lung cancer.¹¹ Moreover, chest ultrasound can also be used to detect rib metastases from non-small cell lung cancer.¹²

However, these radiologic examinations are prone to high false positive rates and high expenses. Additionally, many medical facilities lack proper equipment. Therefore, there is an urgent need to identify a new effective, economical and routine detection method to screen for bone metastasis in patients with lung cancer in the initial stages.

Machine learning technology can improve the diagnosis and prediction efficacy in the field of clinical medicine. Studies have shown that the prognosis of patients receiving chemotherapy can be predicted by machine learning.¹³ Other studies have shown that the immune microenvironment of prostate cancer can predict disease progression.¹⁴ Other studies have shown that novel autophagy-related lncRNA can predict the prognosis of colonic adenocarcinoma.¹⁵ Other studies have shown that the complement system is closely related to the coagulation cascade and endometriosis.¹⁶ Studies have shown that machine learning methods can distinguish people who are allergic to food by epigenetic biomarkers.¹⁷ Other studies have shown that machine learning can predict and diagnose hemodynamic instability in surgery patients by physiological waveforms and electronic health records.¹⁸ Additionally, it has been reported that a new machine learning technique for accurate diagnosis of coronary artery disease has been developed. Moreover, studies have shown that machine learning algorithms based on medical data can diagnose patients with ankylosing spondylitis.¹⁹ In addition, the current research on artificial intelligence-related bone metastasis has focused on radiologic data with small sample sizes.^20–22

Therefore, we explored 9 machine learning algorithms (including general machine learning algorithms and joint machine learning algorithms) to distinguish bone metastasis among elderly stage IV lung cancer patients.

Methods

General Information

27 627 patients with advanced lung cancer from the Surveillance, Epidemiology, and End Results database (SEER) between January 2010 and December 2015 were included in this study, and they were divided into a bone metastasis group (11 147 cases) and non-bone metastasis group (16 480 cases). Inclusion criteria were elderly patients diagnosed with stage IV lung cancer (age greater than or equal to 60 years old) (seventh edition of UICC/AJCC clinical stage); and exclusion criteria were patients with undefined staging, pathological features, or indicators. This study was exempt from review by the local institutional review board because it was a secondary data analysis of the SEER public database, and because the information in this database was anonymous and publicly available for relevant medical research worldwide.

Research Methods and Research Contents

The following clinicopathological data from patients diagnosed with stage IV lung adenocarcinoma from 2010 to 2015 were collected from the SEER database using SEER * Stat 8.3.5: sex, race, age, marital status, diagnostic age, primary location, histological grade, T stage (seventh edition of UICC/AJCC TNM stage), N stage, M stage, surgical status, tumor location, radiotherapy and chemotherapy, metastasis of liver cancer, metastasis of lung cancer, brain metastasis and bone metastasis.

Machine Learning Method

Nine machine learning algorithms (Logical regression-LR, Random forest-RF, Gradient Boosting Decision Tree-GBDT, XGBoost-XGB, LightGBM- LGBM, RF + LR, LGBM + LR, GBDT + LR, XGB + LR) were used. Logistic regression is a classical statistical learning classification method, and the basic model can be used for two-class learning. Random forest is a classifier with multiple decision trees, and its output categories are determined by the modes of the categories output by individual trees. Gradient Boosting Decision Tree is an iterative decision tree algorithm, which consists of multiple decision trees, and the conclusions of all trees are added up for the final answer. XGBoost is a lifting tree model, which integrates many tree models to form a strong classifier. Lightgbm (lightgradient boosting machine) is an open source framework for gradient lifting, and 1 of the frameworks for implementing the GBDT algorithm which supports efficient parallel training.

The study patients were divided into training and testing groups at a ratio of 7:3, and we adopted the 5 cross-verification scheme. We used the training group to train the machine learning model, and then tested the verification model’s performance in the test group. To adjust the parameters, we conducted manual parameter adjustment and grid search. In addition, in order to simplify the results of machine learning classifiers, we used 4 machine learning classifiers to calculate and quantify each feature’s ranking. All of the data were normalized, and we used the accuracy, precision, recall rate and AUC (area under the curve) value to evaluate the machine learning model’s performance. Accuracy is the ratio of correct predictions to all predictions. The precision rate is the rate that is actually labeled a among all predicted a’s. The recall rate is the predicted number of correct positive samples divided by the number of all positive samples (for positive samples). AUC is the area under the ROC curve.

Statistical Analysis

We used R version 3.5.3 for statistical analysis of the general information, and the categorical variables were expressed with count and percentage. The quantitative variables were expressed with mean ± standard deviation. We conducted a Student’s t-test with statistical significance set at P < .05, and constructed the machine learning models in Python. P > .05 indicated that there was no significant difference. .01 < P < .05 indicated significant difference and was marked with *; P < .01 indicated the most significant difference and was marked with * *.

Results

General Information

A total of 27 627 patients were included in this study, of which 11 147 had bone metastasis. There were significant differences in age, sex, race and surgery between the bone metastasis group and the non-metastasis group (P < .001). (Table 1).

Table 1.

Clinical Characteristics of Patients with Lung Cancer.

Bone Metastasis	No	Yes	P Value
N	16 480	11 147
Age (years)	73.0 ± 8.3	72.0 ± 7.8	* *
Race recode			* *
White	12 971 (78.7%)	8977 (80.5%)
Black	1942 (11.8%)	1111 (10.0%)
Other	1567 (9.5%)	1059 (9.5%)
Sex			* *
Male	8238 (50.0%)	6094 (54.7%)
Female	8242 (50.0%)	5053 (45.3%)
Surgery			* *
No	14 846 (90.1%)	10 284 (92.3%)
Yes	1634 (9.9%)	863 (7.7%)
Marital status at diagnosis			* *
Married	8295 (50.3%)	6156 (55.2%)
Not married	8185 (49.7%)	4991 (44.8%)
Primary site			* *
C34.1-upper lobe, lung	9435 (57.3%)	6522 (58.5%)
C34.2-middle lobe, lung	788 (4.8%)	494 (4.4%)
C34.3-lower lobe, lung	4589 (27.8%)	3273 (29.4%)
C34.9-lung, NOS	1668 (10.1%)	858 (7.7%)
Laterality			* *
Left - origin of primary	6544 (39.7%)	4597 (41.2%)
Right - origin of primary	9407 (57.1%)	6302 (56.5%)
Paired site, but no information concerning laterality	507 (3.1%)	232 (2.1%)
Only 1 side - side unspecified	22 (.1%)	16 (.1%)
T			* *
T0	328 (2.0%)	109 (1.0%)
T1	2367 (14.4%)	1822 (16.3%)
T2	4585 (27.8%)	3189 (28.6%)
T3	4154 (25.2%)	2729 (24.5%)
T4	5046 (30.6%)	3298 (29.6%)
N			* *
N0	5109 (31.0%)	2673 (24.0%)
N1	1310 (7.9%)	998 (9.0%)
N2	6985 (42.4%)	5195 (46.6%)
N3	3076 (18.7%)	2281 (20.5%)
M			* *
M1a	7332 (44.5%)	311 (2.8%)
M1b	8983 (54.5%)	10 673 (95.7%)
M1NOS	165 (1.0%)	163 (1.5%)
Radiotherapy			* *
No	14 846 (90.1%)	10 284 (92.3%)
Yes	1634 (9.9%)	863 (7.7%)
Chemotherapy			* *
No	7612 (46.2%)	4829 (43.3%)
Yes	8868 (53.8%)	6318 (56.7%)
Sequence number			* *
One primary only	12 072 (73.3%)	8447 (75.8%)
1st of 2 or more primaries	437 (2.7%)	231 (2.1%)
2nd of 2 or more primaries	3206 (19.5%)	2062 (18.5%)
3rd of 3 or more primaries	638 (3.9%)	328 (2.9%)
4th of 4 or more primaries/or more	127 (.8%)	79 (.7%)
Tumor size			* *
≤3 cm	6100 (37.0%)	3821 (34.3%)
3-7 cm	7898 (47.9%)	5707 (51.2%)
≥7 cm	2482 (15.1%)	1619 (14.5%)
Brain metastasis			* *
No	11 786 (71.5%)	8546 (76.7%)
Yes	4694 (28.5%)	2601 (23.3%)
Liver metastasis			* *
No	14 681 (89.1%)	8663 (77.7%)
Yes	1799 (10.9%)	2484 (22.3%)
Lung metastasis			* *
No	10 899 (66.1%)	7854 (70.5%)
Yes	5581 (33.9%)	3293 (29.5%)

Note: P > .05 indicated that there was no significant difference. .01 < P < .05 indicated significant difference and was marked with *; P < .01 indicated the most significant difference and was marked with * *.

Correlation and Feature Ranking Analysis

M stage, metastasis and tumor size were significantly correlated with bone metastasis. There was a weak negative correlation between being female and radiotherapy and bone metastasis (Figure 1). We first used the average classifiers of the 4 machine learning algorithms to rank the features. The results showed that race, sex, whether they had had surgery and marital status were the first 4 factors affecting bone metastasis (Figure 2).

Figure 1.

Correlation between Clinical Characteristics Data.

Figure 2.

Ranking Results for Bone Metastasis Feature Weights of Average Algorithm.

Machine Learning Results in the Training Group

In this study, we used general machine learning algorithms and a joint machine learning algorithm. For area under the curve (AUC), except for RF and LR, the AUC values for all machine learning classifiers were greater than .8, but the joint algorithm could not improve the AUC of any single machine learning algorithm. Among the results related to accuracy and precision, the accuracy of machine learning classifiers other than the RF algorithm was higher than 70%, and only the LGBM algorithm’s precision was higher than 70%. For recall rate, the GBDT algorithm’s recall rate was the highest at 79.5% (Table 2 and Figure 3).

Table 2.

Model Results for Training Group.

	Accuracy	Precision	Recall	AUC
RF	.670	.637	.425	.780
GBDT	.740	.644	.795	.810
LGBM	.757	.709	.675	.842
XGB	.751	.662	.782	.827
LR	.727	.640	.738	.781
RF + LR	.743	.651	.780	.812
GBDT + LR	.737	.645	.773	.810
LGBM + LR	.754	.675	.750	.834
XGB + LR	.753	.676	.745	.833

Abbreviations: (Logical regression-LR, Random forest-RF; Gradient Boosting Decision Tree-GBDT, XGBoost-XGB; LightGBM- LGBM, RF + LR; LGBM + LR, GBDT + LR; XGB + LR).

Figure 3.

Machine Learning Algorithm Results for Training Group for Bone Metastasis Abbreviations: (Logical regression-LR, Random forest-RF, Gradient Boosting Decision Tree-GBDT, XGBoost-XGB, LightGBM- LGBM, RF + LR, LGBM + LR, GBDT + LR, XGB + LR).

Machine Learning Results in the Test Group

Similarly, for area under the curve (AUC), except for RF and LR, the AUC values of all of the machine learning classifiers were greater than .8. However, the joint algorithm did not improve the AUC value of any single machine learning algorithm. For accuracy, except for the RF algorithm, the accuracy of all machine learning classifiers was higher than 70%. The highest precision was that of the LGBM algorithm at .675. The highest recall rate was that of the GBDT algorithm at 79.2%. (Table 3 and Figure 4).

Table 3.

Model Results for Testing Group.

	Accuracy	Precision	Recall	AUC
RF	.678	.654	.426	.786
GBDT	.743	.649	.792	.814
LGBM	.733	.675	.652	.813
XGB	.748	.660	.775	.816
LR	.732	.648	.737	.790
RF + LR	.747	.656	.784	.812
GBDT + LR	.739	.649	.769	.813
LGBM + LR	.724	.644	.707	.800
XGB + LR	.730	.650	.715	.803

Abbreviations: (Logical regression-LR, Random forest-RF; Gradient Boosting Decision Tree-GBDT, XGBoost-XGB; LightGBM- LGBM, RF + LR; LGBM + LR, GBDT + LR; XGB + LR).

Figure 4.

Machine Learning Algorithm Results for Testing Group for Bone Metastasis Abbreviations: (Logical regression-LR, Random forest-RF, Gradient Boosting Decision Tree-GBDT, XGBoost-XGB, LightGBM- LGBM, RF + LR, LGBM + LR, GBDT + LR, XGB + LR).

Discussion

As of this writing, lung cancer remains a form of malignant tumor with high incidence and mortality throughout the world, and 50% −70% of patients suffer bone metastasis complications.²³ The overall prognosis of patients with bone metastasis lung cancer is still poor, and the incidence of lung cancer with bone metastasis remains on the rise.²⁴ Some lung cancer patients with bone metastases may not show any clinical symptoms, and the diagnosis relies primarily on radiologic examinations, such as CT, ECT and MRI. These diagnostics are not only expensive, but there is radiation risk, and low sensitivity and specificity. They also cannot dynamically monitor changes in bone metabolism. This results in delayed diagnosis and treatment of bone metastasis.^25–27 Therefore, there is an urgent need to establish a new early warning system to indicate the risk and presence of bone metastasis, so as to promote the prevention and treatment of lung cancer. We rank the features based on the differentiation performance of the average classifiers of the 4 machine learning algorithms. The results show that race, surgery condition, whether they had surgery and marriage were the 4 factors most affecting bone metastasis. Moreover, machine learning algorithm classifiers can distinguish bone metastasis from lung cancer, but the joint algorithm cannot improve the performance of a single machine learning algorithm.

Race, sex and marital factors are associated with bone metastasis in cancer patients. However, the effect of race on patient prognosis with non-small cell lung cancer is still controversial.^28,29 Studies have shown that race is associated with bone metastasis in patients with bladder cancer.³⁰ Studies have also shown strong associations between sex and race and bone metastasis in patients with nasopharyngeal carcinoma.³¹ Other studies have shown that race does not predict bone metastasis in men with non-metastatic castrated prostate cancer.³² Additional studies have shown that women have a lower risk of bone metastasis and good prognosis in non-sex-specific cancers.³³ Additionally, it has also been shown that there is a positive correlation between being male and lymphoid metastasis and bone metastasis of lung cancer.³⁴ In addition, studies have shown that the independent risk factors for BM in patients with liver cancer are sex, T stage and N phase.³⁵ Moreover, studies have shown that sex and marriage are independent risk factors for brain metastasis in patients with melanoma.³⁶ Our results also indicate that race and sex influence bone metastasis in patients with lung cancer.

The prognostic benefits of surgical treatment for patients with advanced lung cancer are still controversial.³⁷ Previous studies based on SEER program analysis have shown that further surgical treatment of patients with advanced lung cancer is not recommended.³⁸ However, some studies have shown that the quality of life and 5-year survival rate of stage IV NSCLC patients undergoing pneumonectomy and extended chest wall resection have been improved.³⁹ In addition, studies have shown that surgery may improve the quality of life and survival rate of patients with bone metastasis from gastric cancer.⁴⁰ Our study has shown that surgery is 1 of the main factors affecting bone metastasis in patients with lung cancer.

The advantages and disadvantages of machine learning and other examination tools (such as MRI/CT/ultrasound) in distinguishing the bone metastasis of lung cancer include the following: Advantages: Compared with other tools, the machine learning tool is less invasive for patients, the cost for large-scale screening in the general population is lower, and the environmental requirements for testing are lower; Disadvantages: At present, the clinical application of machine learning in this field remains immature, and its application value in the evaluation of therapeutic effects on bone metastases and dynamic monitoring of disease progression remains unclear, thus this warrants further research.

This study has several limitations. First, the factors above limit the use of other factors in the SEER database. Only a limited amount of information can be extracted from the SEER database, so it is impossible to use multimodal data to construct bone metastasis models. Moreover, it is not possible to include more specific details on bone metastasis if there is no relevant data. Finally, the sequence of different metastatic sites could not be determined. Multicenter prospective cohort study is needed in the future.

Conclusion

The results of this concept verification study show that machine learning algorithm classifiers can distinguish the bone metastasis of patients with lung cancer. This will provide a new research direction for identifying bone metastasis of lung cancer by non-invasive technology in the future. However, more prospective multicenter cohort studies are required. In addition, the performance of these algorithms needs further improvement.

Footnotes

Acknowledgments

None.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Availability of Data and Information

Data sets generated and/or analyzed during the current study are available in the Surveillance, Epidemiology and final results (SEER) repository ().

Ethical Approval and Consent to Participate

This study was exempt from review by the local institutional review board because it was a secondary data analysis of the SEER public database, and because the information in this database was anonymous and publicly available for relevant medical research worldwide.

ORCID iD

Cheng-Mao Zhou

References

Bray

Ferlay

Soerjomataram

Siegel

Torre

Jemal

. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2018;68:394-424. doi:10.3322/caac.21492

Oliveira

MBDR

Mello

FCDQ

Paschoal

MEM

. The relationship between lung cancer histology and the clinicopathological characteristics of bone metastases. Lung Cancer. 2016;96:19-24. doi:10.1016/j.lungcan.2016.03.014

Drzymalski

Werner

Regan

Kantoff

Tuli

. Predictors of survival in patients with prostate cancer and spinal metastasis. Presented at the 2009 Joint Spine Section Meeting. Clinical article. J Neurosurg Spine. 2010;13:789-794. doi:10.3171/2010.6.spine10167

Coello

Luketich

Litle

Godfrey

. Prognostic significance of micrometastasis in non-small-cell lung cancer. Clin Lung Cancer. 2004;5:214-225. doi:10.3816/clc.2004.n.002

Katakami

Kunikane

Takeda

Takayama

Sawa

Saito

, et al. Prospective Study on the Incidence of Bone Metastasis (BM) and Skeletal-Related Events (SREs) in Patients (pts) with Stage IIIB and IV Lung Cancer—CSP-HOR 13. J Thorac Oncol 2014;9:231-238. doi:10.1097/jto.0000000000000051

Kanakis

Nikolaou

Pectasides

Kiamouris

Karamanos

. Determination and biological relevance of serum cross-linked type I collagen N-telopeptide and bone-specific alkaline phosphatase in breast metastatic cancer. J Pharm Biomed Anal. 2004;34:827-832. doi:10.1016/s0731-7085(03)00567-3

Lucic

Nikoletic

Peter

Lucic

Jovanović

. Magnetic resonance imaging and bone scintigraphy in bone metastasis detection: A comparative study. Vojnosanit Pregl. 2010;67:453-458. doi:10.2298/vsp1006453l

Takenaka

Ohno

Matsumoto

Aoyama

Onishi

Koyama

, et al. Detection of bone metastases in non-small cell lung cancer patients: comparison of whole-body diffusion-weighted imaging (DWI), whole-body MR imaging without and with DWI, whole-body FDG-PET/CT, and bone scintigraphy. J Magn Reson Imag 2009;30:298-308. doi:10.1002/jmri.21858

Han

Zhang

. Ultrasonography-Guided Radiofrequency Ablation for Painful Stump Neuromas to Relieve Postamputation Pain: A Pilot Study. J Pain Res 2020;13:3437-3445. doi:10.2147/jpr.s283986

10.

Han

Chen

, et al. Ultrasound-Guided Extraforaminal Thoracic Nerve Root Block Through the Midpoint of the Inferior Articular Process and the Parietal Pleura: A Clinical Application of Thoracic Paravertebral Nerve Block. J Pain Res 2022;15:533-544. doi:10.2147/jpr.s351145

11.

Hoosein

Barnes

Khan

Peake

Bennett

Purnell

, et al. The importance of ultrasound in staging and gaining a pathological diagnosis in patients with lung cancer--a two year single centre experience. Thorax 2011;66:414-417. doi:10.1136/thx.2010.153288

12.

Tomos

Tziolos

Raptakis

Kavatha

. Thoracic ultrasound for the detection of rib metastases of non-small cell lung cancer. Adv Respir Med. 2018;86:101-102. doi:10.5603/arm.2018.0014

13.

Huang

Zhang

Lin

Song

. The impact of chemotherapy and survival prediction by machine learning in early Elderly Triple Negative Breast Cancer (eTNBC): a population based study from the SEER database. BMC Geriatr. 2022;22:268. doi:10.1186/s12877-022-02936-5

14.

Ren

Chen

Zhang

Jiang

Zhang

, et al. Immune Microenvironment and Response in Prostate Cancer Using Large Population Cohorts. Front Immunol 2021;12:686809. doi:10.3389/fimmu.2021.686809

15.

Yin

Meng

, et al. Identification of novel autophagy-related lncRNAs associated with a poor prognosis of colon adenocarcinoma through bioinformatics analysis. Sci Rep 2021;11:8069. doi:10.1038/s41598-021-87540-0

16.

Shen

Ren

Wang

Zhu

Zheng

, et al. Multi-omics analysis reveals the interaction between the complement system and the coagulation cascade in the development of endometriosis. Sci Rep 2021;11:11926. doi:10.1038/s41598-021-90112-x

17.

Alag

. Machine learning approach yields epigenetic biomarkers of food allergy: A novel 13-gene signature to diagnose clinical reactivity. PLoS One. 2019;14:e0218253. doi:10.1371/journal.pone.0218253

18.

Cannesson

Hofer

Rinehart

Lee

Subramaniam

Baldi

, et al. Machine learning of physiological waveforms and electronic health record data to predict, diagnose and treat haemodynamic instability in surgical patients: protocol for a retrospective study. BMJ Open 2019;9:e031988. doi:10.1136/bmjopen-2019-031988

19.

Walsh

Rozycki

Park

. Application of machine learning in the diagnosis of axial spondyloarthritis. Curr Opin Rheumatol. 2019;31:362-367. doi:10.1097/bor.0000000000000612

20.

Lin

Luo

Gao

Man

Cao

, et al. Deep learning based automatic segmentation of metastasis hotspots in thorax bone SPECT images. PLoS One 2020;15:e0243253. doi:10.1371/journal.pone.0243253

21.

Ellmann

Seyler

Evers

Heinen

Bozec

Prante

, et al. Prediction of early metastatic disease in experimental breast cancer bone metastasis by combining PET/CT and MRI parameters to a Model-Averaged Neural Network. Bone 2019;120:254-261. doi:10.1016/j.bone.2018.11.008

22.

Papandrianos

Papageorgiou

Anagnostis

Papageorgiou

. Bone metastasis classification using whole body images from prostate cancer patients based on convolutional neural networks application. PLoS One. 2020;15:e0237213. doi:10.1371/journal.pone.0237213

23.

Zou

Zhu

Lian

Wang

Chen

Zhang

, et al. miR-192-5p suppresses the progression of lung cancer bone metastasis by targeting TRIM44. Sci Rep 2019;9:19619. doi:10.1038/s41598-019-56018-5

24.

Liu

. Lung cancer with bone metastases in the United States: an analysis from the Surveillance, Epidemiologic, and End Results database. Clin Exp Metastasis. 2018;35:753-761. doi:10.1007/s10585-018-9943-5

25.

Ebert

Muley

Herb

Schmidt-Gayk

. Comparison of bone scintigraphy with bone markers in the diagnosis of bone metastasis in lung carcinoma patients. Anticancer Res. 2004;24:3193-3201.

26.

Swart

Scholtens

Tanis

Nieman

Bogers

A JJC

Verzijlbergen

, et al. 18F-fluorodeoxyglucose positron emission/computed tomography and computed tomography angiography in prosthetic heart valve endocarditis: from guidelines to clinical practice. Eur Heart J 2018;39:3739-3749. doi:10.1093/eurheartj/ehx784

27.

Yen

Chen

Cheng

Shiau

, et al. The diagnostic and prognostic effectiveness of F-18 sodium fluoride PET-CT in detecting bone metastases for hepatocellular carcinoma patients. Nucl Med Commun 2010;31:637-645. doi:10.1097/mnm.0b013e3283399120.

28.

White

Joseph

Rim

Johnson

Coleman

Allemani

. Colon cancer survival in the United States by race and stage (2001-2009): Findings from the CONCORD-2 study. Cancer. 2017;123(suppl 24):5014-5036. doi:10.1002/cncr.31076

29.

Tannenbaum

Koru-Sengul

Zhao

Miao

Byrne

. Survival disparities in non-small cell lung cancer by race, ethnicity, and socioeconomic status. Cancer J. 2014;20:237-245. doi:10.1097/ppo.0000000000000058

30.

Zhang

Liu

Tao

Guo

Feng

Chen

, et al. Bone Metastases Pattern in Newly Diagnosed Metastatic Bladder Cancer: A Population-Based Study. J Cancer 2018;9:4706-4711. doi:10.7150/jca.28706

31.

Yang

Ren

Yang

. Bone Metastases Pattern in Newly Diagnosed Metastatic Nasopharyngeal Carcinoma: A Real-World Analysis in the SEER Database. BioMed Res Int 2020;2020:2098325. doi:10.1155/2020/2098325

32.

Whitney

Howard

Amling

Aronson

Cooperberg

Kane

, et al. Race does not predict the development of metastases in men with nonmetastatic castration-resistant prostate cancer. Cancer 2016;122:3848-3855. doi:10.1002/cncr.30221

33.

Peltzer

Liu

Wang

, et al. Female sex is associated with a lower risk of bone metastases and favourable prognosis in non-sex-specific cancers. BMC Cancer 2019;19:1001. doi:10.1186/s12885-019-6168-1

34.

Zhang

Mao

Guo

Cui

Zhang

, et al. Nomogram based on homogeneous and heterogeneous associated factors for predicting bone metastases in patients with different histological types of lung cancer. BMC Cancer 2019;19:238. doi:10.1186/s12885-019-5445-3

35.

Yang

Huang

Liu

Lin

Tong

, et al. Diagnostic and prognostic nomograms for bone metastasis in hepatocellular carcinoma. BMC Cancer 2020;20:494. doi:10.1186/s12885-020-06995-y

36.

Liu

Guo

Dong

, et al. Predictive value of a nomogram for melanomas with brain metastases at initial diagnosis. Cancer Med 2019;8:7577-7585. doi:10.1002/cam4.2644

37.

Schmelzle

Eisenberger

am Esch

JS,

2nd Matthaei

Krausch

Knoefel

. Non-colorectal, non-neuroendocrine, and non-sarcoma metastases of the liver: resection as a promising tool in the palliative management. Langenbeck's Arch Surg. 2010;395:227-234. doi:10.1007/s00423-009-0580-y

38.

Abdel-Rahman

. Outcomes of Surgery as Part of the Management of Metastatic Non–Small-Cell Lung Cancer: A Surveillance, Epidemiology and End Results Database Analysis. Cancer Invest. 2018;36:238-245. doi:10.1080/07357907.2018.1466895

39.

Setzer

Robinson

Vrionis

. Management of locally advanced pancoast (superior sulcus) tumors with spine involvement. Cancer Control. 2014;21:158-167. doi:10.1177/107327481402100209

40.

Choi

Kim

Han

Son

Kim

, et al. Long-term survival after gastrectomy and metastasectomy for gastric cancer with synchronous bone metastasis. World J Gastroenterol 2018;24:150-156. doi:10.3748/wjg.v24.i1.150

Differentiation of Bone Metastasis in Elderly Patients With Lung Adenocarcinoma Using Multiple Machine Learning Algorithms

Abstract

Objective

Methods

Results

Conclusion

Keywords

Introduction

Methods

General Information

Research Methods and Research Contents

Machine Learning Method

Statistical Analysis

Results

General Information

Correlation and Feature Ranking Analysis

Machine Learning Results in the Training Group

Machine Learning Results in the Test Group

Discussion

Conclusion

Footnotes

Acknowledgments

Declaration of Conflicting Interests

Funding

Availability of Data and Information

Ethical Approval and Consent to Participate

ORCID iD

References