Sage Journals: Discover world-class research

Abstract

Background:

Given the rarity of tracheobronchopathia osteochondroplastica (TO), many young doctors in primary hospitals are unable to identify TO based on bronchoscopy findings.

Objectives:

To build an artificial intelligence (AI) model for differentiating TO from other multinodular airway diseases by using bronchoscopic images.

Design:

We designed the study by comparing the imaging data of patients undergoing bronchoscopy from January 2010 to October 2022 by using EfficientNet. Bronchoscopic images of 21 patients with TO at Anhui Chest Hospital from October 2019 to October 2022 were collected for external validation.

Methods:

Bronchoscopic images of patients with multinodular airway lesions (including TO, amyloidosis, tumors, and inflammation) and without airway lesions in the First Affiliated Hospital of Guangzhou Medical University were collected. The images were randomized (4:1) into training and validation groups based on different diseases and utilized for deep learning by convolutional neural networks (CNNs).

Results:

We enrolled 201 patients with multinodular airway disease (38, 15, 75, and 73 patients with TO, amyloidosis, tumors, and inflammation, respectively) and 213 without any airway lesions. To find multinodular lesion images for deep learning, we utilized 2183 bronchoscopic images of multinodular lesions (including TO, amyloidosis, tumor, and inflammation) and compared them with images without any airway lesions (1733). The accuracy of multinodular lesion identification was 98.9%. Further, the accuracy of TO detection based on the bronchoscopic images of multinodular lesions was 89.2%. Regarding external validation (using images from 21 patients with TO), all patients could be diagnosed with TO; the accuracy was 89.8%.

Conclusion:

We built an AI model that could differentiate TO from other multinodular airway diseases (mainly amyloidosis, tumors, and inflammation) by using bronchoscopic images. The model could help young physicians identify this rare airway disease.

Keywords

artificial intelligence bronchoscopy tracheobronchopathia osteochondroplastica TO

Introduction

Tracheobronchopathia osteochondroplastica (TO), a rare, benign disorder involving the lumen of the trachea-bronchial tree, is characterized by multiple cartilaginous and submucosal osseous nodules in the trachea and bronchus. As well as the TO nodules are located along the airway cartilage side but not along the posterior membranous wall of the large airways.^1,2 The classical radiological features for TO are shown as follows. On the one hand, the thickness of the airway wall is increased. On the other hand, multiple calcified shadows can be found along the airway cartilage side but not along the posterior wall. However, computed tomography (CT) was unable to detect calcified nodules in the majority of TO patients.³ Although TO is rare, there have been many reports in which it was occasionally diagnosed.^1,4–27 To date, only a limited number of studies have utilized extensive data to conduct a comprehensive investigation of this disease.^3,28

Nowadays, artificial intelligence (AI) has been used to assist physicians in better detecting lesions under gastrointestinal endoscopy and making a diagnosis.^3,29–34 However, AI has not been frequently utilized to assist physicians by using bronchoscopy data, which has been reported to be used only for identifying between vocal cord and tracheal, as well as between the carina and main bronchi.^35,36 Recently, Li et al.³⁷ conducted a study on using AI to recognize comprehensively anatomical positions by using bronchoscopy data.

TO could be asymptomatic or manifest in the form of common respiratory symptoms such as cough, sputum production, and dyspnea. Its diagnosis is based on bronchoscopy and pathology findings, and its treatment, in turn, is based on its diagnosis, for which a specific strategy is lacking.³⁸ In the past 10 years, the use of flexible bronchoscopy increased approximately five-fold in China.³⁹ However, given the rarity of TO, many young doctors in primary hospitals are unable to identify TO based on bronchoscopy findings. Because many young doctors in primary hospitals do not have chances to meet patients with TO. Even in some high-level hospitals, advanced respiratory doctors do not have chances to touch the bronchoscopes. Further, as one of the multinodular airway diseases, TO should also be distinguished from other airway diseases (including tumor, inflammation, and amyloidosis), and the sample figures of inflammation and tumor could find that they were difficult to differentiate (Supplemental Figures 1 and 2). Therefore, this deep learning-based study was aimed at building an AI model by using a convolutional neural network (CNN) to help young physicians identify TO from other multinodular airway diseases by using bronchoscopic images so as to avoid misdiagnoses.

Methods

Patients who were found to have diffuse and multinodular airway lesions under bronchoscopy and based on pathologic testing at the First Affiliated Hospital of Guangzhou Medical University from January 2010 to October 2022 were included. Bronchoscopic image data were used to distinguish between TO and other diffuse and multinodular airway diseases (such as amyloidosis, tumors, and inflammation). The sample bronchoscopic images of tumors and inflammation are shown in Supplemental Figures 1 and 2, respectively, which were extremely similar to those of TO. As a rare airway disease, amyloidosis is always discussed with TO, which is why we considered amyloidosis, tumors, and inflammation in the differential diagnosis of TO.

As a famous department for respiratory medicine in China, since 2009, the First Affiliated Hospital of Guangzhou Medical University has been the top respiratory medicine center in the country, with the average annual number of hospitalizations reaching 70,000 and the average number of outpatient visits reaching over 1,000,000. Therefore, data for many cases of rare diseases (such as TO and amyloidosis) could be collected. The data for external validation (patients with TO) were collected from Anhui Chest Hospital, a famous provincial hospital in China. The external validation dataset was collected by doctor FT from Anhui Chest Hospital, who was blinded to the process of internal AI model building and internal/external validating.

Five groups of participants were included in the study: patients with TO, amyloidosis, tumors, and inflammation, and healthy individuals. The bronchoscopic images of patients with multinodular diseases (TO, amyloidosis, tumors, and inflammation) were compared with those of individuals without any airway lesions. Second, the bronchoscopic images of patients with TO were compared with those of patients with other multinodular diseases (amyloidosis, tumors, and inflammation). Lastly, the external dataset of bronchoscopic images obtained for patients with TO was used for validation.

Inclusion and exclusion criteria

The inclusion criteria were as follows: (1) patients older than 14 years of age, (2) patients diagnosed with multinodular airway diseases based on bronchoscopic findings, (3) patients with multinodular airway disease with a pathological diagnosis of TO, amyloidosis, tumors, and inflammation, and (4) patients without any overlapping airway disease (such as overlapping TO and tumors or overlapping TO and inflammation). Patients without clear bronchoscopic image data (probably due to the date of examinations) were excluded.

The inclusion criteria for bronchoscopic images were as follows: (1) the scope of the lesions was more than one-third of the picture, (2) the scope of the blood was less than one-quarter of the picture, (3) the absence of any operative apparatus in the picture, (4) the absence of sputum in the picture, (5) and adequate image clarity. All images were selected by two independent investigators (CXC and FT), and duplicate or similar images were deleted. Finally, any differences in opinion between the first two investigators were resolved by a third investigator (SYL).

Statistical analysis

Categorical variables are expressed as numbers and percentages, and continuous variables are expressed as mean ± standard deviation. Python3.7 was used for AI analysis. EfficientNet (a CNN) was used for deep learning to differentiate TO from other diffusive nodular diseases (amyloidosis, tumor, and inflammation). The EfficientNet could ignore the influence of the pixel.⁴⁰ We also tried several different AI models (Inception, ResNet, VGG) to distinguish TO from other diseases before utilizing EfficientNet model. The sensitivity, specificity, positive predicted value (PPV), and negative predicted value (NPV) for finding TO by Inception model were 73.5%, 95.8%, 80.3%, and 93.9%, respectively (Supplemental Figure 3). Moreover, the sensitivity, specificity, PPV, and NPV for finding TO by ResNet were 79.5%, 97.5%, 88.0%, and 95.3%, respectively (Supplemental Figure 4). Lastly, the sensitivity, specificity, PPV, and NPV for finding TO by VGG were 85.5%, 97.2%, 87.7%, and 96.6%, respectively (Supplemental Figure 5). Image classification is one of the most basic jobs in pattern identification and computer vision; it translates obtainable features from the image into feature vectors that could be known by computers.⁴¹ So, the method of CNNs was to figure out the features of the figures by computers, which could not be explained in detail.⁴¹ The bronchoscopic images of all the airway anatomical positions were pre-processed using the Gaussian filter, graphic lightening, and normalizing. The confusion matrix and receiver operating characteristic curves (ROCs) were plotted for determining accuracy. The findings of this study adhere to the guidelines outlined in the Standards for Reporting of Diagnostic Accuracy (STARD) statement.⁴²

Results

We included 201 patients with diffuse and nodular airway disease: 38, 15, 75, and 73 patients were diagnosed with TO, amyloidosis, tumors, and inflammation, respectively, based on pathologic testing. Furthermore, 213 patients without any nodular lesions were included as a comparator group. For external validation, the Anhui Chest Hospital offered 177 bronchoscopic images of 21 patients with TO (Table 1).

Table 1.

The baseline characteristics of the enrolled patients and images.

Classification	Inflammation	Tumor	Amyloidosis	TO (internal)	TO (external)	Without nodular lesions
Number of patients	73	75	15	38	21	213
Sex (Male)	38 (52.05%)	53 (70.67%)	9 (60%)	23 (60.53%)	13 (61.90%)	109 (51.17%)
Age	45.27 ± 16.12	58.68 ± 10.06	54.07 ± 10.50	46.68 ± 9.67	54.50 ± 11.43	50.83 ± 14.22
Number of images	836	561	370	416	177	1733

TO, tracheobronchopathia osteochondroplastica.

Among the included patients with tumors, 22, 16, 5, 10, 8, and 5 patients were diagnosed with lung adenocarcinoma, lung squamous cell carcinoma, non-small-cell lung cancer (NSCLC, un-classified), small-cell lung cancer, metastatic carcinoma (of the breast, digestive tract, and thymus gland), and other types of tumor (such as big-cell lung cancer, lymphoma, lymphoepithelioma, esothelioma, and adenoid cystic carcinoma), respectively. The remaining nine patients were diagnosed as having cancer without classification via pathologic testing. Among the included patients with inflammation, 32, 6, and 35 were diagnosed with granulomatous inflammation, eosinophilic inflammation, and other types of common acute/chronic inflammation, respectively. The types of tumors and inflammation are shown in Table 2.

Table 2.

Classifications of inflammation and tumors included in the study.

Classification	Number (%)
Tumor
Lung adenocarcinoma	22 (29.33%)
Lung squamous cell carcinoma	16 (21.33%)
Non–small-cell lung cancer (un-classified)	5 (6.67%)
Small-cell lung cancer	10 (13.33%)
Metastatic carcinoma (breast, digestive tract, and thymus)	8 (10.67%)
Others (big-cell lung cancer, lymphoma, lymphoepithelioma, esothelioma, and adenoid cystic carcinoma)	5 (6.67%)
Cancer without classifications via pathology	9 (12.00%)
Total number	75 (100%)
Inflammation
Granulomatous inflammation	32 (43.84%)
Eosinophilic inflammation	6 (8.22%)
Other common acute/chronic types of inflammation	35 (47.94%)
Total number	73 (100%)

The total number of the images used for training and validation from our hospital was 3916: the numbers of the bronchoscopic images of TO, amyloidosis, tumors, inflammation, and without any nodular lesions were 416, 370, 561, 836, and 1733, respectively. In every group, 80% of bronchoscopic images were randomly chosen as the training dataset, and the remaining 20% were chosen as the internal validation dataset (Table 1).

The included patients were divided into two groups: patients with multinodular airway disease and patients without any nodular airway disease. The model’s accuracy for recognizing the bronchoscopic images of multinodular lesions was 98.9% (Figure 1). Figure 2 shows the area under the ROC for recognizing multinodular airway lesions for the AI model; the results were both 0.999. The train and validation losses associated with the AI model for recognizing multinodular airway disease are shown in Figure 3.

Figure 1.

The AI model’s accuracy in recognizing multinodular airway lesions.

Figure 2.

The receiver operating characteristic curve for the recognition of multinodular airway lesions by the AI model.

Figure 3.

The train and validation loss curves for multinodular airway lesion recognition by the AI model.

With regard to the identification of TO-related images from among the images of multinodular disease, the bronchoscopic images were classified into two groups: TO and non-TO. The bronchoscopic images for inflammation and tumors were added to the control group. The results showed that the sensitivity, specificity, PPV, and NPV were 89.2%, 97.5%, 89.2%, and 97.5%, respectively (Figure 4). Figure 5 shows an AUC of 0.986 for the recognition of TO from among other multinodular airway diseases. The train and validation losses of the AI model for recognizing TO from among the bronchoscopic images of multinodular diseases are shown in Figure 6.

Figure 4.

The accuracy of TO recognition by the AI model.

Figure 5.

The receiver operating characteristic curve of TO recognition by the AI model.

Figure 6.

The train and validation loss curves for TO recognition by the AI model.

For external validation, 177 bronchoscopic images obtained from 21 patients with TO were used for validating the AI model (Table 1), and the accuracy for TO identification was 89.8%. Supplemental Figure 9 shows the sample bronchoscopic images chosen by the AI model as those corresponding to TO in the external dataset.

Discussion

Our AI study showed that the AI model could be used to recognize diffuse and multinodular airway disease by using bronchoscopy data. Moreover, the AI model trained by us could distinguish TO from other diffuse and nodular diseases (amyloidosis, inflammation, and tumor) with relatively high accuracy.⁴³

We also conducted three two-class image classifications (TO versus malignancy, TO versus amyloidosis, and TO versus inflammation). The sensitivity, specificity, PPV, and NPV of AI model for differentiating between TO and Malignancy were 92.8%, 97.3%, 96.3%, and 94.8%, respectively. (Supplemental Figure 6) The sensitivity, specificity, PPV, and NPV of AI model for differentiating between TO and amyloidosis were 97.6%, 98.6%, 98.8%, and 97.3%, respectively. (Supplemental Figure 7) The sensitivity, specificity, PPV, and NPV of the AI model for differentiating between TO and inflammation were 96.4%, 98.8%, 97.6%, and 98.2%, respectively. (Supplemental Figure 8).

The reasons for incorrect recognition are shown as follows. Firstly, the incidences of TO and amyloidosis were rare, as a two-center study, the enrolled number of patients was still small. Secondly, as one of the multinodular airway diseases, TO should also be distinguished from other multinodular airway diseases (tumor, inflammation, or amyloidosis). The sample figures of inflammation and tumor we proposed could find that they were difficult to differentiate (Supplemental Figures 1 and Figure 2).

TO is a rare disease characterized by diffuse nodular airway lesions, and related tissue compositions are mainly calcareous. Many theories about TO pathogenesis have been proposed; however, the exact cause of the disease remains unknown. These theories include chronic airway inflammation, a congenital basis, chemical or mechanical irritation, metabolic disturbance, degenerative processes, ecchondrosis/exostosis, and metaplasia of elastic tissue. A recent study by Hong et al.³⁸ demonstrated that dysfunction of airway basal stem cells is a significant factor in the pathophysiology of TO. In our view, The AI model probably recognized patients with these conditions based on bronchoscopic images by recognizing the distribution and color characteristics of nodules. However, this is only our presumption, and the precise underlying disease mechanism remains unknown.

Many associations between TO and other clinical conditions have been described. Hence, caution should be exercised in the evaluation of TO since it could co-exist with sarcoidosis, tumors, and amyloidosis.^44,45 We excluded three patients with TO and tumors (lung adenocarcinoma, lung squamous cell carcinoma, and malignant fibroma). While bronchoscopy can be used to diagnose TO, biopsy is essential for excluding other potential causes.²⁸ This is why our research focused on obtaining a pathological diagnosis. In the inflammation group, we comprehensively considered disorders including tuberculosis,⁴⁶ and sarcoidosis.⁴⁷ The tumor group included patients such as those with lung cancer, breast cancer, and lymphoma.⁴⁸

We previously used an AI model to differentiate nine anatomical positions in the airway based on bronchoscopy data.⁴⁹ In our view, various backgrounds of bronchoscopy (many bronchi and lobes) could bring problems to the AI study. The lesions of multinodular airway disease in our study were mainly located in the trachea and main bronchi (large airways). The use of images of diffuse and multinodular airway disease for differential diagnosis could address the problems posed by various background factors. Because multinodular lesions were widely distributed in the airway, we did not annotate the lesions on the images.

Up to now, many AI-associated studies about pulmonary diseases have shown up. These previous studies used AI for managing lung cancer,⁵⁰ chronic obstructive pulmonary disease (COPD),^51–53 infectious pulmonary diseases,^54,55 and acute respiratory distress syndrome.⁵⁶ However, most of them were analyzed by using CT, and ultrasound.^57,58 And there were few studies using bronchoscopic images for AI study.

After the supervised machine learning, the training model could classify the images that could promptly and automatically collect image data during examination. The algorithm research of CNN went through ALexNet,⁵⁹ VGG,⁶⁰ ResNet,⁶¹ and EfficientNet.⁴⁰ Now, the EfficientNet added the attention mechanism with high efficiency, and could ignore the influence of pixel.⁴⁰

The limitations of this study were as follows. Firstly, we only included 38 patients with TO for training and testing from our hospital, and the AI model trained was mainly aimed at differentiating TO. The accuracy of tumor and inflammation detection was limited because of the varying types of tumors (including lung adenocarcinoma, lung squamous cell carcinoma, and small-cell lung cancer) and inflammation (including due to granuloma, fungal infection, and tuberculosis) present. In the future, we could include more bronchoscopic images from patients with tumors or inflammation to differentiate patients (for subgroup analysis). Secondly, image inclusion and exclusion were mainly based on the pathophysiological characteristics of the multinodular disease (such as the size of the lesions). Thirdly, this study only included two centers with Chinese patients; future studies should include larger populations with a variety of patients. However, although our study only included 59 patients because of the rarity of TO, it was still the largest population of patients with TO discussed thus far. Lastly, we did not calculate the sample size selected in this study.

In the future, AI could be a teaching tool for young doctors in primary hospitals. We hope to conduct a worldwide multi-center study based on bronchoscopy image data, which could help us increase the effects of AI.

Conclusion

The AI model trained by us could distinguish between TO and other multinodular diseases caused by tumors or inflammation. Thus, it can be used to help teach young physicians and remote clinical workers to recognize this rare disease by using bronchoscopy data.

Supplemental Material

sj-doc-1-tar-10.1177_17534666241253694 – Supplemental material for Building and validating an artificial intelligence model to identify tracheobronchopathia osteochondroplastica by using bronchoscopic images

Supplemental material, sj-doc-1-tar-10.1177_17534666241253694 for Building and validating an artificial intelligence model to identify tracheobronchopathia osteochondroplastica by using bronchoscopic images by Chongxiang Chen, Fei Tang, Felix J. F. Herth, Yingnan Zuo, Jiangtao Ren, Shuaiqi Zhang, Wenhua Jian, Chunli Tang and Shiyue Li in Therapeutic Advances in Respiratory Disease

Supplemental Material

sj-docx-2-tar-10.1177_17534666241253694 – Supplemental material for Building and validating an artificial intelligence model to identify tracheobronchopathia osteochondroplastica by using bronchoscopic images

Supplemental material, sj-docx-2-tar-10.1177_17534666241253694 for Building and validating an artificial intelligence model to identify tracheobronchopathia osteochondroplastica by using bronchoscopic images by Chongxiang Chen, Fei Tang, Felix J. F. Herth, Yingnan Zuo, Jiangtao Ren, Shuaiqi Zhang, Wenhua Jian, Chunli Tang and Shiyue Li in Therapeutic Advances in Respiratory Disease

Footnotes

Acknowledgements

We acknowledge all contributed authors.

Declarations

ORCID iD

Chongxiang Chen

Supplemental material

Supplemental material for this article is available online.

References

Kamangar

. Tracheobronchopathia Osteochondroplastica. Mayo Clin Proc 2019; 94: 949–950.

Riva

Girolami

Luchini

, et al. Tracheobronchopathia osteochondroplastica: a case report illustrating the importance of multilevel workup clinical, endoscopic and histological assessment in diagnosis of an uncommon disease. Am J Case Rep 2019; 20: 74–77.

Liu

Lei

, et al. Clinical characteristics of tracheobronchopathia osteochondroplastica: a retrospective study of 33 patients. Int J Gen Med 2023; 16: 3447–3455.

Agarwal

Klair

Joiner

, et al. Nodular trachea: tracheobronchopathia osteochondroplastica. BMJ Case Rep 2015; 2015: bcr2015209860.

Bachy

Saroul

Darcha

, et al. An unusual cause of tracheal stenosis: diagnosis and management? Tracheopathia osteochondroplastica. Eur Ann Otorhinolaryngol Head Neck Dis 2012; 129: 211–213.

Barata

Gomes

Moreira

, et al. Tracheobronchopathia osteochondroplastica: two different facets of a rare entity. Eur J Case Rep Intern Med 2020; 7: 1925.

Basharzad

Mojtabaee

Fadaei

. Nephroquiz 9: tracheobronchopathia osteochondroplastica presenting with dyspnea in a patient with end-stage renal disease. Iran J Kidney Dis 2017; 11: 74–78.

Chien

Wang

Chang

, et al. Uncommon chronic cough caused by tracheobronchopathia osteochondroplastica. Thorax 2012; 67: 1021–1022.

Guo

Zhou

Wei

, et al. Clinical characteristics of six cases of tracheobronchopathia osteochondroplastica. Can Respir J 2020; 2020: 8685126.

10.

Ihara

Tajima

Yoshikawa

, et al. Autofluorescence imaging bronchoscopy as a novel approach to the management of tracheobronchopathia osteochondroplastica: a case report. J Thorac Dis 2016; 8: E1195–E1198.

11.

Jiang

Yang

Guo

. Multi-nodule of large airway: tracheobronchopathia osteochondroplastica, two cases report and literature review. Ear Nose Throat J 2021: 1549797006.

12.

Juanola

Rejon

Ortega

. Tracheobronchopathia osteochondroplastica. Arch Bronconeumol (Engl Ed) 2020; 56: 172.

13.

Kafili

Sampson

Tolhurst

. Difficult intubation in an asymptomatic patient with tracheobronchopathia osteochondroplastica. Respirol Case Rep 2020; 8: e526.

14.

Jin

Nan

, et al. Multi-nodule of large airway: tracheobronchopathia osteochondroplastica. Ann Palliat Med 2021; 10: 1115–1121.

15.

Lim

Abu

Pereirasamy

, et al. Tracheobronchopathia osteochondroplastica-stalactite of airways. Respirol Case Rep 2021; 9: e790.

16.

Luo

Zhou

Meng

. Clinical characteristics of tracheobronchopathia osteochondroplastica. Respir Care 2019; 64: 196–200.

17.

Matsuki

Takeda

Kida

, et al. Tracheobronchopathia osteochondroplastica: a cause of difficult tracheal intubation. Intern Med 2018; 57: 909–910.

18.

Michaelides

Bablekos

Ionas

, et al. Right upper lobe shadow and tracheobronchopathia osteochondroplastica confined to right main bronchus: a case Report and literature review. Case Rep Med 2015; 2015: 368485.

19.

Nikandish

Fallahi

Ziaian

, et al. Repeated tracheostomy tube cuff rupture due to tracheobronchopathia osteochondroplastica: a case report. Iran J Otorhinolaryngol 2015; 27: 387–390.

20.

Porzezinska

Janowicz

Janowiak

, et al. Tracheobronchopathia osteochondroplastica-case report and literature review. Pneumonol Alergol Pol 2015; 83: 135–139.

21.

Ribeiro

Natal

Silva

, et al. Tracheobronchopathia osteochondroplastica: computed tomography, bronchoscopy and histopathological findings. Radiol Bras 2016; 49: 56–57.

22.

Sakaguchi

Matsumoto

Nishioka

, et al. Bronchoscopic surgery for a solitary tracheal tumor of tracheobronchopathia osteochondroplastica. Ann Thorac Surg 2020; 109: e419–e421.

23.

Sharma

Kamath

Acharya

, et al. Pebbled trachea: a case series and literature review of tracheobronchopathia osteochondroplastica. J Glob Infect Dis 2021; 13: 180–182.

24.

Sun

Xie

, et al. Tracheobronchopathia osteochondroplastica: case report and literature review. Respir Med Case Rep 2015; 15: 14–17.

25.

Thakur

Yang

Chen

, et al. Atypical presentation of tracheobronchopathia osteochondroplastica: is chronic inflammation a perpetrator? Med Princ Pract 2013; 22: 503–505.

26.

Wang

Liu

, et al. Tracheobronchopathia osteochondroplastica: five cases report and literature review. Ear Nose Throat J 2020; 99: P111–P118.

27.

Xing

Liu

, et al. Tracheobronchopathia osteochondroplastica: imaging and bronchoscopic findings. Ann Thorac Surg 2015; 99: e97.

28.

Nair

Gupta

Irodi

, et al. Tracheobronchopathia osteochondroplastica - to biopsy or not to biopsy? A relook at the rare disease. J Bronchology Interv Pulmonol 2024; 31: 57–62.

29.

Yang

Wang

, et al. A deep learning method for delineating early gastric cancer resection margin under chromoendoscopy and white light endoscopy. Gastric Cancer 2020; 23: 884–892.

30.

Dong

, et al. Real-time use of artificial intelligence for diagnosing early gastric cancer by magnifying image-enhanced endoscopy: a multicenter diagnostic study (with videos). Gastrointest Endosc 2022; 95: 671–678.

31.

Sutton

Zai

Goebel

, et al. Artificial intelligence enabled automated diagnosis and grading of ulcerative colitis endoscopy images. Sci Rep 2022; 12: 2748.

32.

Tontini

Rimondi

Vernero

, et al. Artificial intelligence in gastrointestinal endoscopy for inflammatory bowel disease: a systematic review and new horizons. Therap Adv Gastroenterol 2021; 14: 1088200706.

33.

Liu

, et al. Evaluation of the effects of an artificial intelligence system on endoscopy quality and preliminary testing of its performance in detecting early gastric cancer: a randomized controlled trial. Endoscopy 2021; 53: 1199–1207.

34.

Jiang

, et al. Real-time artificial intelligence for detecting focal lesions and diagnosing neoplasms of the stomach by white-light endoscopy (with videos). Gastrointest Endosc 2022; 95: 269–280.

35.

Matava

Pankiv

Raisbeck

, et al. A convolutional neural network for real time classification, identification, and labelling of vocal cord and tracheal using laryngoscopy and bronchoscopy video. J Med Syst 2020; 44: 44.

36.

Yoo

Kang

Park

, et al. Deep learning for anatomical interpretation of video bronchoscopy images. Sci Rep 2021; 11: 23765.

37.

Zheng

Xie

, et al. Development and validation of the artificial intelligence (AI)-based diagnostic model for bronchial lumen identification. Transl Lung Cancer Res 2022; 11: 2261–2274.

38.

Hong

Shan

, et al. Malfunction of airway basal stem cells plays a crucial role in pathophysiology of tracheobronchopathia osteoplastica. Nat Commun 2022; 13: 1309.

39.

Shi

Wang

, et al. The development of bronchoscopy in China: a national cross-sectional study. J Cancer 2020; 11: 5547–5555.

40.

Souedet

Jan

, et al. A general deep learning framework for neuron instance segmentation based on Efficient UNet and morphological post-processing. Comput Biol Med 2022; 150: 106180.

41.

Zhou

Wang

, et al. A graph neural network model for the diagnosis of lung adenocarcinoma based on multimodal features and an edge-generation network. Quant Imaging Med Surg 2023; 13: 5333–5348.

42.

Bossuyt

Reitsma

Bruns

, et al. STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies. BMJ 2015; 351: h5527.

43.

Fang

Zhang

Wang

, et al. [Relationship between etiology and morphological classification of benign central airway stenosis and its prognosis]. Zhonghua Jie He He Hu Xi Za Zhi 2022; 45: 768–774.

44.

Karlikaya

Yuksel

Kilicli

, et al. Tracheobronchopathia osteochondroplastica. Respirology 2000; 5: 377–380.

45.

Abu-Hijleh

Lee

Braman

. Tracheobronchopathia osteochondroplastica: a rare large airway disorder. Lung 2008; 186: 353–359.

46.

Marom

Goodman

McAdams

. Diffuse abnormalities of the trachea and main bronchi. AJR Am J Roentgenol 2001; 176: 713–717.

47.

White

Kong

Khoo

, et al. Computed tomography diagnosis of tracheobronchopathia osteochondroplastica. Australas Radiol 2005; 49: 319–321.

48.

Joshi

Chan

Bruch

, et al. Thyroid lymphoma and airway obstruction – is there a rationale for surgical management? Int J Clin Pract 2009; 63: 1647–1652.

49.

Chen

Herth

Zuo

, et al. Distinguishing bronchoscopically observed anatomical positions of airway under by convolutional neural network. Ther Adv Chronic Dis 2023; 14: 384230071

50.

Adams

Stone

Baldwin

, et al. Lung cancer screening. Lancet 2023; 401: 390–408.

51.

Diaz

Nardelli

Wang

, et al. Artificial intelligence-based CT assessment of bronchiectasis: the COPDGene study. Radiology 2023; 307: e221109.

52.

Schiebler

Seo

. Artificial intelligence analysis of bronchiectasis is predictive of outcomes in chronic obstructive pulmonary disease. Radiology 2023; 307: e222675.

53.

Gonzalez

Ash

Vegas-Sanchez-Ferrero

, et al. Disease staging and prognosis in smokers using deep learning in chest computed tomography. Am J Respir Crit Care Med 2018; 197: 193–203.

54.

Ying

, et al. T-SPOT with CT image analysis based on deep learning for early differential diagnosis of nontuberculous mycobacteria pulmonary disease and pulmonary tuberculosis. Int J Infect Dis 2022; 125: 42–50.

55.

Trovato

Russo

. Artificial Intelligence (AI) and lung ultrasound in infectious pulmonary disease. Front Med (Lausanne) 2021; 8: 706794.

56.

Rashid

Ramakrishnan

Chandran

, et al. Artificial intelligence in acute respiratory distress syndrome: a systematic review. Artif Intell Med 2022; 131: 102361.

57.

Gottlieb

Patel

Viars

, et al. Comparison of artificial intelligence versus real-time physician assessment of pulmonary edema with lung ultrasound. Am J Emerg Med 2023; 70: 109–112.

58.

Diaz-Escobar

Ordonez-Guillen

Villarreal-Reyes

, et al. Deep-learning based detection of COVID-19 using lung ultrasound imagery. PLos One 2021; 16: e255886.

59.

Wang

Zhang

, et al. Diagnostic performance of corona virus disease 2019 chest computer tomography image recognition based on deep learning: systematic review and meta-analysis. Medicine (Baltimore) 2022; 101: e31346.

60.

Yang

Lai

Wang

, et al. Diversity-learning block: conquer feature homogenization of multibranch. IEEE Trans Neural Netw Learn Syst 2022. DOI: 10.1109/TNNLS.2022.3214993.

61.

Islam

Jones

Faiz

, et al. Improving performance of breast lesion classification using a ResNet50 model optimized with a novel attention mechanism. Tomography 2022; 8: 2411–2425.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

6.39 MB

0.02 MB