Method, device, and computer program for predicting pulmonary complications after esophagectomy using deep learning

Abstract

The present invention relates to a method for predicting pulmonary complications after esophagectomy using deep learning, the method comprising the steps of: receiving a pre-operative chest CT image of one or more patients; extracting the total lung volume from the chest CT image and detecting a lesion area by inputting the chest CT image into a trained deep learning model; calculating a degree of emphysema by dividing a percentage value of a low attenuated area having a density of specific Hounsfield Units (HU) or less in the detected area by the total lung volume; and generating a nomogram by performing multivariate analysis on the calculated degree of emphysema and one or more postoperative pulmonary complication (PPC) factors.

Description

Method, device, and computer program for predicting pulmonary complications after esophagectomy using deep learning

[0001] This invention was made under the support of the Ministry of Science and ICT under the grant number 1711183369 and grant number 2022R1A2C2093106. The research management organization of the above-mentioned project is the National Research Foundation of Korea, the research project name is "Individual Basic Research", the research project title is "Identification of changes in the tumor microenvironment and clinical significance through single-cell and spatial transcriptome analysis in the esophageal cancer carcinogenesis process", the main organization is Samsung Medical Center, and the research period is from September 1, 2022 to February 28, 2025.

[0002] In addition, the present invention was made under the support of the Ministry of Science and ICT under the task identification number 1711184924 and task number 2022R1A2C1003999, and the research management specialized organization of the said task is the National Research Foundation of Korea, the research project name is "Mid-career Researcher Support Project", the research project name is "Development and verification of a model for predicting the response and prognosis of specialized immuno-oncology drugs based on multi-omics data integration and artificial intelligence algorithm", the main organization is Samsung Seoul Hospital, and the research period is from 2022.03.01. to 2025.02.28.

[0003] The present invention relates to a method and device for predicting pulmonary complications after esophagectomy using deep learning, and more particularly, to a method and device for predicting pulmonary complications after esophagectomy using deep learning, which can select a high-risk group for developing pulmonary complications after esophageal cancer surgery through CT images.

[0004] Esophagectomy is known to have a higher incidence of complications and mortality compared to other gastrointestinal surgeries. Postoperative pulmonary complications (PPC) are among the most common and fatal complications after esophagectomy. Although the incidence of postoperative pulmonary complications has been gradually decreasing due to advancements in minimally invasive esophagectomy (MIE) and perioperative management techniques, they still occur in 16% to 23% of patients and are a major cause of hospital death after esophageal cancer surgery. Furthermore, postoperative pulmonary complications are known to decrease long-term survival rates regardless of esophageal cancer recurrence. Therefore, identifying and preventing risk factors for postoperative pulmonary complications is crucial for improving postoperative esophagectomy outcomes.

[0005] Interstitial lung abnormalities (ILAs) are incidentally discovered computed tomography (CT) findings with potential clinical implications. The prevalence of ILA varies depending on smoking status, with rates ranging from 4% to 9% in smokers and 2% to 7% in the general population. ILAs are associated with accelerated decline in lung function, increased mortality, and an increased risk of lung cancer development and worse prognosis. Although ILAs have been reported as a risk factor for postoperative pulmonary complications in various surgeries, including lung cancer surgery, the association between ILAs and postoperative pulmonary complications after esophagectomy has not been adequately studied.

[0006] Accordingly, there is a need for a method, device and computer program capable of predicting pulmonary complications after esophagectomy by combining the presence and extent of interstitial abnormalities in the lung with previously reported risk factors for postoperative pulmonary complications.

[0007] The present invention aims to provide a method, device and computer program for predicting pulmonary complications after esophagectomy using deep learning, which can predict pulmonary complications after esophagectomy by combining the presence and extent of interstitial abnormalities in the lung and previously reported risk factors for postoperative pulmonary complications.

[0008] In order to achieve the above object, the present invention is characterized in that it includes a step of receiving a preoperative chest CT image of one or more patients using deep learning, a step of extracting a total lung volume from the chest CT image, and a step of detecting a lesion area by inputting the chest CT image into a trained deep learning model, a step of calculating the degree of emphysema by dividing the percentage value of an area having a density below a specific HU (Hounsfield Units) (low attenuated area) among the detected areas by the total lung volume, and a step of generating a nomogram by multivariately analyzing the calculated degree of emphysema and one or more postoperative pulmonary complication (PPC) factors.

[0009] Preferably, the step of detecting the lesion area is such that the lesion is related to interstitial lung abnormalities (ILA), and consolidation, ground-glass opacity (GGO), reticulation, and honeycombing areas can be detected as the lesion area.

[0010] Preferably, the step of calculating the degree of said emphysema may be such that said specific HU is -950HU.

[0011] Preferably, the step of generating the nomogram may include at least one of the PPC factors being age, smoking status, FEV1 level, anastomosis level, and lung inflammation.

[0012] Preferably, the step of generating the nomogram may perform at least one of a t-test, a Mann-Whitney u-test, a chi-square test, Fisher's exact test, a Spearman correlation coefficient, and a logistic regression analysis as the multivariate analysis.

[0013] Preferably, the method may further include a step of training a deep learning model using a chest CT image as input and a lesion area among the chest CT images as output.

[0014] Preferably, the step of training the deep learning model may train the deep learning model by dividing the lesion area into consolidation, ground-glass opacity (GGO), reticulation, and honeycomb areas.

[0015] Preferably, the method may include a step of predicting whether PPC occurs by inputting the degree of emphysema of a specific patient and one or more PPC factors into the generated nomogram.

[0016] In addition, the present invention provides a device for predicting pulmonary complications after esophagectomy using deep learning, comprising: a processor including one or more cores; and a memory; wherein the processor receives a preoperative chest CT image of one or more patients, extracts a total lung volume from the received chest CT image, inputs the CT image into a trained deep learning model to detect a lesion area, divides the percentage value of an area having a density below a specific HU (Hounsfield Units) (low attenuated area) among the detected areas by the total lung volume to calculate the degree of emphysema, performs multivariate analysis of the degree of emphysema and one or more postoperative pulmonary complication (PPC) factors to generate a nomogram, and predicts whether PPC occurs by inputting the degree of emphysema and one or more other PPC factors of a specific patient into the generated nomogram.

[0017] In addition, the present invention is characterized in that it is a computer program including commands stored in a computer-readable storage medium and causing a computer to perform the following operations, wherein the operations include: an operation of receiving a preoperative chest CT image of one or more patients; an operation of extracting a total lung volume from the received chest CT image and inputting the CT image into a trained deep learning model to detect a lesion area; an operation of calculating the degree of emphysema by dividing the percentage value of an area having a density below a specific HU (Hounsfield Units) (low attenuated area) among the detected areas by the total lung volume; an operation of generating a nomogram by performing a multivariate analysis of the degree of emphysema and one or more postoperative pulmonary complication (PPC) factors; and an operation of predicting whether PPC occurs by inputting the degree of emphysema of a specific patient and one or more other PPC factors into the generated nomogram.

[0018] The present invention has the advantage of being able to select a high-risk group for developing pulmonary complications after esophageal cancer surgery (after esophagectomy).

[0019] In addition, the present invention has the advantage of being able to provide information so that, in the case of patients determined to be at high risk through prediction of the occurrence of pulmonary complications after esophageal cancer surgery, alternative treatments can be performed instead of surgery, or methods to reduce the risk of occurrence of pulmonary complications after surgery through preoperative preparation can be considered.

[0020] Figure 1 shows a flowchart of a method for predicting pulmonary complications after esophagectomy using deep learning according to an embodiment of the present invention.

[0021] Figure 2 illustrates a schematic diagram of a method for predicting pulmonary complications after esophagectomy using deep learning according to an embodiment of the present invention.

[0022] Figure 3 shows a nomogram generated according to an embodiment of the present invention.

[0023] Figure 4 shows the correlation between texture parameters and lung function according to an embodiment of the present invention.

[0024] FIG. 5 shows a correction plot of the predicted probability for postoperative pulmonary complications and the predicted probability for postoperative pulmonary complications according to an embodiment of the present invention.

[0025] Fig. 6 shows a device for predicting pulmonary complications after esophagectomy using deep learning according to an embodiment of the present invention.

[0026] Figure 7 illustrates a schematic diagram of a computing environment according to an embodiment of the present invention.

[0027] A method for predicting pulmonary complications after esophagectomy using deep learning,

[0028] A step of receiving a preoperative chest CT image of one or more patients;

[0029] A step of extracting the entire lung volume from the chest CT image and inputting the chest CT image into a trained deep learning model to detect a lesion area;

[0030] A step of calculating the degree of emphysema by dividing the percentage value of the area with a density below a specific HU (Hounsfield Units) among the detected areas by the total lung volume; and

[0031] A step of generating a nomogram by multivariate analysis of the calculated degree of emphysema and one or more postoperative pulmonary complication (PPC) factors;

[0032] A method comprising:

[0033] Hereinafter, the present invention will be described in detail with reference to the contents described in the attached drawings. However, the present invention is not limited or restricted by the exemplary embodiments. The same reference numerals in each drawing indicate components that perform substantially the same functions.

[0034] The purpose and effects of the present invention can be naturally understood or made clearer by the following description, and the purpose and effects of the present invention are not limited solely by the following description. Furthermore, in describing the present invention, if a detailed description of known technologies related to the present invention is deemed to unnecessarily obscure the gist of the present invention, such detailed description will be omitted.

[0035] The terminology used herein is merely used to describe specific embodiments and is not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this application, it should be understood that the terms "comprise" or "have" indicate the presence of a feature, number, step, operation, component, part, or combination thereof described in the description of the invention, but do not preclude the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0036] While terms like "first" and "second" may be used to describe various components, these components should not be limited by these terms. These terms are used solely to distinguish one component from another. For example, without departing from the scope of the present invention, a first component could be referred to as a "second component," and similarly, a second component could also be referred to as a "first component."

[0037] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and shall not be interpreted in an idealized or overly formal sense unless explicitly defined herein.

[0038] When interpreting components, even if there is no explicit description, it is interpreted as including the margin of error. When describing temporal relationships, for example, when temporal continuity is described with phrases such as "after," "following," "next to," or "before," this also includes cases where the relationship is not continuous, unless "immediately" or "directly" is used.

[0039] Hereinafter, the technical configuration of the present invention will be described in detail with reference to the attached drawings.

[0040] Fig. 1 is a flowchart of a method for predicting pulmonary complications after esophagectomy using deep learning according to an embodiment of the present invention. Referring to Fig. 1, the method for predicting pulmonary complications after esophagectomy using deep learning may include a step of receiving a chest CT image (S100), a step of detecting a lesion area (S200), a step of calculating the degree of emphysema (S300), a step of generating a nomogram (S400), a step of predicting whether PPC occurs (S500), and a step of training a deep learning model (S600).

[0041] Figure 2 illustrates a schematic diagram of a method for predicting pulmonary complications after esophagectomy using deep learning according to an embodiment of the present invention.

[0042] The step of receiving a chest CT image (S100) may receive a preoperative chest CT image of one or more patients. The step of receiving a chest CT image (S100) may exclude, from among the received chest CT images, any volumetric images that are insufficient or thin cross-sectional images less than 2.5 mm in size.

[0043] The step of receiving a chest CT image (S100) may receive a chest CT image of a patient who will undergo therapeutic esophagectomy and esophageal reconstruction. In one embodiment, the patients who are the subjects of the chest CT image may be patients who have undergone esophagectomy with reconstruction through a general surgical procedure for esophageal cancer. In one embodiment, the patients who are the subjects of the chest CT image may be patients who have undergone thoracotomy or minimally invasive esophagectomy. In one embodiment, the patients who are the subjects of the chest CT image may be patients who have undergone an intrathoracic anastomosis, and if the resection margin is insufficient, the patients who have undergone a cervical anastomosis. In one embodiment, the patients who are the subjects of the chest CT image may be patients who have upper thoracic esophageal cancer or clinically metastatic lymph nodes in the neck, in which case the patients may have undergone a three-zone lymphadenectomy (cervical, thoracic, and abdominal).

[0044] In addition, the step of receiving a chest CT image (S100) may exclude chest CT images of patients who received neoadjuvant therapy from among the received chest CT images. This is because neoadjuvant therapy itself may reduce FEV1 and DLCo (Diffusing capacity of the Lung for CO) after esophagectomy, which may be a risk factor for pulmonary complications.

[0045] In the step (S100) of receiving a chest CT image according to the present invention, the chest CT image received may be taken by two different 64-channel multi-slice CT scanners. The chest CT image may be obtained with the patient lying in the supine position at a voltage of 120 kVp and a slice thickness of 1-2 mm.

[0046] The step of detecting a lesion area (S200) extracts the entire lung volume from the received chest CT image, and inputs the CT image into a trained deep learning model to detect the lesion area.

[0047] The step of detecting a lesion area (S200) can automatically extract the entire lung volume to specifically describe the lung parenchyma area excluding pulmonary blood vessels or airways.

[0048] The extent of interstitial lung abnormalities (ILA) at the whole lung level can be automatically quantified using a deep learning-based approach. The extent of ILA can refer to the lesion area.

[0049] The step of detecting a lesion area (S200) is related to interstitial lung abnormalities (ILA), and consolidation, ground-glass opacity (GGO), reticulation, and honeycombing can be detected as parameters for the lesion area. The step of detecting a lesion area (S200) can identify texture patterns according to reticulation and honeycombing.

[0050] The step of detecting the lesion area (S200) can obtain the extent of ILA (extent of ILA %, i.e., lesion area) by dividing the sum of the detected parameters by the total lung volume.

[0051] The step of calculating the degree of emphysema (S300) can calculate the degree of emphysema by dividing the percentage value of the area (low attenuated area) with a density lower than a specific HU (Hounsfield Units) among the detected areas by the total lung volume (LAA%). In the step of calculating the degree of emphysema (S300), the specific HU may be -950 HU.

[0052] The entire quantitative process of kernel transformation, lung segmentation, and texture analysis can be performed based on deep learning algorithms.

[0053] The step of generating a nomogram (S400) can generate a nomogram by multivariately analyzing the degree of emphysema and one or more postoperative pulmonary complication (PPC) factors.

[0054] Figure 3 illustrates a nomogram generated according to an embodiment of the present invention. Referring to Figure 3, in the step (S400) of generating the nomogram, one or more other PPC factors may be at least one of age, smoking status, FEV1 value, anastomosis level, and lung inflammation level.

[0055] The step of generating a nomogram (S400) is a multivariate analysis, and at least one or more of a t-test, a Mann-Whitney u test, a chi-square test, a Fisher's exact test, a Spearman correlation coefficient, and a logistic regression analysis can be performed.

[0056] The nomogram generation step (S400) can utilize a t-test in the analysis of two samples. The nomogram generation step (S400) can summarize distorted variables into a median with an interquartile range (IQR). The nomogram generation step (S400) can utilize a chi-square test or Fisher's exact test for categorical variables. The nomogram generation step (S400) can analyze the relationship between parameters measured by lung function and texture analysis by calculating the Spearman correlation coefficient. The nomogram generation step (S400) performed multivariate and multivariate logistic regression analyses, which can detect risk factors for PPC using stepwise selection.

[0057] When evaluating the performance of the logistic regression model, the area under the curve (AUC) was used as an indicator, and internal validity was evaluated using three methods: 10-fold cross-validation, bootstrap, and LGOCV (Leave-Group-Out CV).

[0058] The step of generating a nomogram (S400) can compare the performance of the nomogram with the performance of the original logistic regression model, and if the p-value is less than 0.05, it can be considered statistically significant.

[0059] The step (S500) of predicting whether PPC occurs can predict whether PPC occurs by inputting the degree of emphysema of a specific patient and one or more other PPC factors into the generated nomogram.

[0060] The step (S600) of training a deep learning model can train a deep learning model using a chest CT image as input and a lesion area among the chest CT images as output.

[0061] The step (S600) of training a deep learning model can train the deep learning model by dividing the lesion area into consolidation, ground-glass opacity (GGO), reticulation, and honeycomb areas.

[0062]

[0063] Below, the simulation results of the present invention are presented.

[0064] Baseline characteristics of patients

[0065] Among the 765 patients studied, 129 (16.9%) experienced PPC. The incidence of pneumonia, acute respiratory distress syndrome (ALI), and acute respiratory distress syndrome (ARDS) was 112 (14.6%), 29 (3.8%), and 8 (1.0%), respectively. Of these, 20 patients (14 ALI, 6 ARDS) experienced both pneumonia and lung injury. Patient demographics and baseline characteristics are detailed in Table 1. Patients with PPC had a higher current morbidity and lower FEV1 (%) and DLCO (%) compared to those without PPC. Additionally, patients with PPC were more likely to have undergone open esophagectomy and endothoracic anastomosis.

[0066] No PPC (n=636)PPC (n=129)pAge64.48 ± 8.0965.84 ± 9.020.089Male576 (90.6%)122 (94.6%)0.142Smoking status<0.001Never smoker81 (12.7%)6 (4.7%)Ex-smoker300(47.2%)43 (33.3%)Current smoker255 (40.1%)80 (62.0%)FEV1%90.69 ± 14.2885.74 ± 16.400.0005DLCo%87.25 ± 17.9178.04 ± 15.14<0.001FEV1 / FVC73.53 ± 8.5469.70 ± 10.25<0.001Pathology0.325Squamous cell carcinoma598 (94.03%)125 (96.9%)Adenocarcinoma31 (4.87%)4 (3.10%)Others7 (1.10%)0Location of lesion0.476Cervical5 (0.79%)Upper81 (12.74%)15 (11.63%)Mid271 (42.61%)50 (38.76%)Lower244 (38.36%)45.74%)EG junction35 (5.5%)5 (3.88%)pT0.167Tis7 (1.1%)0T1a92 (14.5%)14 (10.9%)T1b348 (54.7%)67 (51.9%)T284 (13.2%)27 (20.9%)T3100 (15.8%)21 (16.3%)T4a4 (0.7%)0pN0.164N0399 (62.8%)82 (63.6%)N1162 (25.6%)25 (19.4%)N254 (8.5%)19 (14.7%)N318 (2.8%)3 (2.3%)Nx2 (0.3%)0Minimally invasive esophagectomy237 (37.3%)25 (19.4%)<0.001Level of anastomosis<0.001Intrathoracic anastomosis380 (59.8%)100 (77.5%)Cervical anastomosis256 (40.3%)29 (22.5%)Operative mortality8 (1.3%)1 (0.8%)0.643.

[0067]

[0068] Lung tissue parameters by automated software

[0069] Lung tissue parameters determined by automated software are described in Table 2 below. Although total lung volume did not differ between patients with and without PPC, ground-glass opacity, reticulation, honeycombing, reduced area, and ILA degree were significantly higher in patients with PPC than in those without. ILA degree (0.149 [0.062–0.502] vs. 0.342 [0.126–0.821], p<0.001) and reduced area (0.123 [0.036–0.462] vs. 0.236 [0.033–1.367], p=0.005) were greater in patients with PPC (p<0.001).

[0070]

[0071] No PPC (n=636)PPC (n=129)p * Whole lung volume (mL)4920.561 (4247.186~5629.445)4794.276 (4106.011~5608.299)0.804Ground glass opacity (%), unit: x10 -3 98.837 (41.056~289.550)211.111 (69.399~588.326)<0.001Reticulation (%), unit: x10 -3 39.094 (11.932~131.261)96.273 (29.302~272.917)<0.001Honeycombing (%), unit: x10 -3 0.098 (0~0.533)0.313 (0.028~1.333)<0.001Extent of ILA (%)0.149 (0.062~0.502)0.342 (0.126~0.821)<0.001Low attenuated area (%)0.123 (0.036~0.462)0.236 (0.033~1.367)0.005

[0072]

[0073] Figure 4 illustrates the correlation between texture parameters and lung function according to an embodiment of the present invention. Referring to Figure 4, FEV1(%) and DLCo(%) showed a significant correlation with each other. FEV1(%) showed a correlation with low attenuation area and ILA degree. ILA degree showed a correlation with DLCo(%) and other texture parameters.

[0074] Risk Factors of PPC

[0075] Univariate and multivariate analyses of PPC are described in Table 3 below. Because the ILA and LAA extents were skewed rather than normal, they were log-transformed to normal distributions and then applied to the univariate analysis. In the univariate analysis, smoking status (current smoker), access method (thoracotomy), FEV1 (%), DLCo (%), and the extent of ILA and LAA were associated with the development of PPC. Age was not associated with PPC in the univariate analysis (p=0.090), but was a significant risk factor for PPC in patients aged 65 years or older in the univariate analysis. Because FEV1 (%) and DLCo (%) were correlated, only DLCo (%) was used in the multivariate analysis. In patients aged 65 years or older, age did not significantly affect the development of PPC (odds ratio 1.054, p=0.062). In multivariate analysis, both the degree of ILA and the low attenuation area, regardless of whether it was fibrous or non-fibrous ILA, were associated with the development of PPC.

[0076]

[0077] Univariable analysisMultivariable analysisOdds ratio(95% confidence interval)pOdds ratio(95% confidence interval)pSex0.551 (0.246~1.234)0.147 Age1.020 (0.997~1.044)0.090 Age(< 65, 1 year increase)0.988 (0.935~1.045)0.6740.967 (0.912~1.025)0.258Age(≥65, 1 year increase)1.044 (0.993~1.097)0.09061.054 (0.997~1.114)0.062Smoking statusEx-smoker (vs. non-smoker)1.935 (0.795~4.705)0.1451.680 (0.667~4.236)0.271Current smoker (vs. non-smoker)4.235 (1.780~10.073)0.0014.887 (1.962~12.169)0.001Open thoractomy (vs. minimally invasive esophagectomy)0.430 (0.276~0.670)<0.0010.485 (0.304~0.775)0.003FEV1 (%)0.977 (0.965~0.990)0.0010.981 (0.966~0.997)0.019Extent of ILA (%, log scale)1.049(1.002~1.097)0.0391.364 (1.182~1.573)<0.001Low attenuated area (%, log scale)1.114 (1.045~1.189)0.0011.158 (1.049~1.279)0.004

[0078]

[0079] PPC 및 검증을 위한 노모그램

[0080] A visually interpretable nomogram score model was developed from the logistic model presented in Table 3 (Figure 3). This model incorporates a linear relationship and clinically acceptable FEV1 levels, which are prominent in patients aged 65 years and older. Scores in the nomogram were assigned based on the estimated regression coefficients and the range of data values ​​in the logistic model presented in Table 3. The log-transformed ILA, the variable with the largest range, was assigned a maximum score of 10, and other variables were assigned scores proportional to this value. After constructing the nomogram, the range values ​​for ILA and LAA were converted to the initial scale.

[0081] FIG. 5 shows a correction plot of the predicted probability for postoperative pulmonary complications and the predicted probability for postoperative pulmonary complications according to an embodiment of the present invention.

[0082]

[0083] The AUC of the nomogram was 0.748, and the average AUCs by the internal 10-fold CV, bootstrap, and LGOCV were 0.729, 0.719, and 0.727, respectively. The nomogram calibration curve for PPC prediction also showed excellent performance (Figure 5A). The total score, which is the sum of the scores for each variable, directly corresponds to the predicted probability of PPC, allowing for the estimation of PPC risk. The correlation coefficient between the predicted probability of the score model in Table 3 and the logistic regression model was 0.99, demonstrating a high level of agreement (Figures 5A and 5B).

[0084] Patients were classified based on the nomogram score (Fig. 5B). Low-risk group (nomogram score 11 points) For 93 patients, the predicted probability of PPC occurrence was 3.3% and the observed PPC occurrence rate was 0%. For 88 very high-risk patients (nomogram score > 18 points), the predicted probability and observed PPC occurrence rate were 42.4% and 43.2%, respectively.

[0085] Fig. 6 illustrates a device (100) for predicting pulmonary complications after esophagectomy using deep learning according to an embodiment of the present invention. Referring to Fig. 6, the configuration of the device (100) for predicting pulmonary complications after esophagectomy using deep learning illustrated is merely a simplified example. In one embodiment of the present invention, the device (100) for predicting pulmonary complications after esophagectomy using deep learning may include other configurations for performing the computing environment of the device (100), and only some of the disclosed configurations may constitute the device (100).

[0086] A device (100) for predicting pulmonary complications after esophagectomy using deep learning may include a processor (110) including one or more cores, a memory (120), and a network (130).

[0087] The processor (110) may be configured with one or more cores, and may include a processor for data analysis and deep learning, such as a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), and a tensor processing unit (TPU) of a computing device. The processor (110) may read a computer program stored in the memory (120) and perform data processing for machine learning according to an embodiment of the present disclosure. According to an embodiment of the present disclosure, the processor (110) may perform operations for learning a neural network. The processor (110) may perform calculations for learning a neural network, such as processing input data for learning in deep learning (DL), extracting features from input data, calculating errors, and updating weights of a neural network using backpropagation. At least one of the CPU, GPGPU, and TPU of the processor (110) may process learning of a network function. For example, a CPU and a GPGPU can jointly process network function learning and data classification using network functions. Furthermore, in one embodiment of the present disclosure, processors of multiple computing devices can be jointly used to process network function learning and data classification using network functions. Furthermore, a computer program executed on a computing device according to one embodiment of the present disclosure may be a CPU, GPGPU, or TPU executable program.

[0088] The processor (110) can receive a preoperative chest CT image of one or more patients. The processor (110) can perform the step (S100) of receiving the chest CT image described above.

[0089] The processor (110) can extract the entire lung volume from the received chest CT image and input the CT image into a trained deep learning model to detect a lesion area. The processor (110) can perform the step (S200) of detecting the lesion area described above.

[0090] The processor (110) can calculate the degree of emphysema by dividing the percentage value of the area (low attenuated area) with a density below a specific HU (Hounsfield Units) among the detected areas by the total lung volume. The processor (110) can perform the step (S300) of calculating the degree of emphysema described above.

[0091] The processor (110) can generate a nomogram by multivariately analyzing the degree of emphysema and one or more other postoperative pulmonary complication (PPC) factors. The processor (110) can perform the step (S400) of generating the aforementioned nomogram.

[0092] The processor (110) can predict whether PPC occurs by inputting the degree of emphysema of a specific patient and one or more other PPC factors into the generated nomogram. The processor (110) can perform the step of predicting whether PPC occurs as described above.

[0093] The memory (120) can store any form of information generated or determined by the processor (110) and any form of information received by the network (130).

[0094] The memory (120) may include at least one type of storage medium among a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (e.g., SD or XD memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, and an optical disk. The computing device (100) may also operate in relation to web storage that performs the storage function of the memory (120) on the internet. The description of the above-described memory is merely an example, and the present disclosure is not limited thereto.

[0095] The network (130) may use any known wired or wireless communication system. The network (130) may receive chest CT images, etc. from related devices or systems.

[0096] The network (130) can transmit and receive information, user interfaces, etc. processed by the processor (110) through communication with other terminals. For example, the network (130) can provide a user interface generated by the processor (100) to a client (e.g., a user terminal). In addition, the network (130) can receive external input from a user authorized as a client and transmit it to the processor (110). At this time, the processor (110) can process operations such as outputting, modifying, changing, and adding information provided through the user interface based on the external input of the user received from the network (130).

[0097] Meanwhile, a device (100) for predicting post-esophagectomy pulmonary complications using deep learning according to one embodiment of the present disclosure may include a server as a computing system that transmits and receives information through communication with a client. In this case, the client may be any type of terminal capable of accessing the server.

[0098] In an additional embodiment, a device (100) for predicting pulmonary complications after esophagectomy using deep learning may include any type of terminal that receives data resources generated from any server and performs additional information processing.

[0099] Another embodiment of the present invention, a computer program for predicting pulmonary complications after esophagectomy using deep learning, may include operations for receiving a chest CT image, detecting a lesion area, calculating the degree of emphysema, generating a nomogram, and predicting whether PPC occurs.

[0100] The operation of receiving a chest CT image may receive a preoperative chest CT image of one or more patients. The operation of receiving a chest CT image refers to the operation performed in the step (S100) of receiving a chest CT image described above.

[0101] The operation of detecting a lesion area can extract the entire lung volume from a received chest CT image and input the CT image into a trained deep learning model to detect the lesion area. The operation of detecting a lesion area refers to the operation performed in the step of detecting a lesion area (S200) described above.

[0102] The operation of calculating the degree of emphysema can be performed by dividing the percentage value of the area with a density below a specific HU (Hounsfield Units) (low attenuated area) among the detected areas by the total lung volume to calculate the degree of emphysema. The operation of calculating the degree of emphysema refers to the operation performed in the step (S300) of calculating the degree of emphysema described above.

[0103] The operation of generating a nomogram can generate a nomogram by multivariately analyzing the degree of emphysema and one or more other postoperative pulmonary complication (PPC) factors. The operation of generating a nomogram refers to the operation performed in the aforementioned nomogram generating step (S400).

[0104] The operation of predicting whether PPC occurs can be performed by inputting the degree of emphysema of a specific patient and one or more other PPC factors into the generated nomogram to predict whether PPC occurs. The operation of predicting whether PPC occurs refers to the operation performed in the step (S500) of predicting whether PPC occurs described above.

[0105] Figure 7 illustrates a schematic diagram of a computing environment according to an embodiment of the present invention.

[0106] Although the present disclosure has been described above as being generally implemented by a computing device, those skilled in the art will appreciate that the present disclosure may be implemented in combination with computer-executable instructions and / or other program modules that may be executed on one or more computers and / or as a combination of hardware and software.

[0107] Generally, program modules include routines, programs, components, data structures, and the like that perform particular tasks or implement particular abstract data types. Furthermore, those skilled in the art will appreciate that the methods of the present disclosure can be implemented with other computer system configurations, including single-processor or multiprocessor computer systems, minicomputers, mainframe computers, as well as personal computers, handheld computing devices, microprocessor-based or programmable consumer electronics, and the like, each of which may be operatively connected to one or more associated devices.

[0108] The described embodiments of the present disclosure can also be practiced in distributed computing environments, where certain tasks are performed by remote processing devices that are connected through a communications network. In a distributed computing environment, program modules may be located in both local and remote memory storage devices.

[0109] Computers typically include a variety of computer-readable media. Computer-readable media can be any media that can be accessed by a computer, and includes both volatile and nonvolatile media, transitory and non-transitory media, removable and non-removable media. By way of example, and not limitation, computer-readable media can include computer-readable storage media and computer-readable transmission media. Computer-readable storage media includes both volatile and nonvolatile media, transitory and non-transitory media, removable and non-removable media implemented in any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer-readable storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital video disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be accessed by a computer and used to store the desired information.

[0110] Computer-readable transmission media typically includes any information delivery media that embodies computer-readable instructions, data structures, program modules, or other data in a modulated data signal, such as a carrier wave or other transport mechanism. The term modulated data signal means a signal that has one or more of its characteristics set or changed so as to encode information in the signal. By way of example, and not limitation, computer-readable transmission media includes wired media, such as a wired network or direct-wired connection, and wireless media, such as acoustic, RF, infrared, or other wireless media. Combinations of any of the above are also intended to be included within the scope of computer-readable transmission media.

[0111] An exemplary environment for implementing various aspects of the present disclosure is illustrated, including a computer (1000), which includes a processing unit (1020), a system memory (1030), and a system bus (1010). The system bus (1010) connects system components, including but not limited to the system memory (1030), to the processing unit (1020). The processing unit (1020) may be any of a variety of commercially available processors. Dual processors and other multiprocessor architectures may also be utilized as the processing unit (1020).

[0112] The system bus (1010) may be any of several types of bus structures that may be additionally interconnected to a memory bus, a peripheral bus, and a local bus using any of a variety of commercial bus architectures. The system memory (1030) includes read-only memory (ROM) (1034) and random access memory (RAM) (1032). A basic input / output system (BIOS) is stored in non-volatile memory (1034), such as ROM, EPROM, or EEPROM, and includes basic routines that help transfer information between components within the computer (1000), such as during start-up. The RAM (1032) may also include high-speed RAM, such as static RAM, for caching data.

[0113] The computer (1000) also includes an internal hard disk drive (HDD) (1050) (e.g., EIDE, SATA) - which may also be configured for external use within a suitable chassis (not shown), a magnetic floppy disk drive (FDD) (1060) (e.g., for reading from or writing to removable diskettes), and an optical disk drive (1070) (e.g., for reading from or writing to CD-ROM disks or other high-capacity optical media such as DVDs). The hard disk drive (1050), the magnetic disk drive (1060), and the optical disk drive (1070) may be connected to the system bus (1010) by a hard disk drive interface, a magnetic disk drive interface, and an optical drive interface, respectively. Interfaces for implementing external drives include at least one or both of Universal Serial Bus (USB) and IEEE 1394 interface technologies.

[0114] These drives and their associated computer-readable media provide non-volatile storage of data, data structures, computer-executable instructions, and the like. In the case of the computer (1000), the drives and media correspond to storing any data in a suitable digital format. While the description of computer-readable media above refers to HDDs, removable magnetic disks, and removable optical media such as CDs or DVDs, those skilled in the art will appreciate that other types of computer-readable media, such as zip drives, magnetic cassettes, flash memory cards, cartridges, and the like, may also be used in the exemplary operating environment, and that any such media may contain computer-executable instructions for performing the methods of the present disclosure.

[0115] A number of program modules, including an operating system (1092), one or more application programs (1094), other program modules (1096), and a database (1098), may be stored in the drive and RAM (1032). All or portions of the operating system, applications, modules, and / or data may also be cached in RAM (1032). It will be appreciated that the present disclosure may be implemented in various commercially available operating systems or combinations of operating systems.

[0116] A user may enter commands and information into the computer (1000) via one or more wired / wireless input devices (1042), such as a keyboard and a pointing device such as a mouse. Other input devices (not shown) may include a microphone, an IR remote control, a joystick, a game pad, a stylus pen, a touch screen, and the like. These and other input devices are often connected to the processing unit (1020) via an input / output interface (1040) that is connected to the system bus (1010), but may be connected by other interfaces such as a parallel port, an IEEE 1394 serial port, a game port, a USB port, an IR interface, and the like.

[0117] A monitor or other type of display device is also connected to the system bus (1010) via an interface such as a video adapter. In addition to the monitor, the computer typically includes other peripheral output devices (not shown) such as speakers, a printer, and so on.

[0118] The computer (1000) may operate in a networked environment using logical connections to one or more remote computers, such as remote computer(s) (1082), via wired and / or wireless communications. The remote computer(s) (1082) may be a workstation, a computing device computer, a router, a personal computer, a portable computer, a microprocessor-based entertainment device, a peer device, or other conventional network node, and may generally include many or all of the components described for the computer (1000). The logical connections include wired / wireless connections to a local area network (LAN) and / or a larger network, such as a wide area network (WAN). Such LAN and WAN networking environments are common in offices and companies and facilitate enterprise-wide computer networks, such as intranets, all of which may be connected to a worldwide computer network, such as the Internet.

[0119] When used in a LAN networking environment, the computer (1000) is connected to a local network (not shown) via a wired and / or wireless communication network interface or adapter (not shown). The adapter (not shown) may facilitate wired or wireless communication to the LAN (not shown), which may also include a wireless access point installed therein for communicating with the wireless adapter (not shown). When used in a WAN networking environment, the computer (1000) may include a modem (not shown), be connected to a communication computing device on the WAN (not shown), or have other means for establishing communications over the WAN (not shown), such as via the Internet. The modem (not shown), which may be internal or external and wired or wireless, is connected to the system bus (1010) via a serial port interface (not shown). In a networked environment, program modules described for the computer (1000), or portions thereof, may be stored in a remote memory / storage device (not shown). It will be appreciated that the network connections shown are exemplary and that other means of establishing a communications link between computers may be used.

[0120] The computer (1000) operates to communicate with any wireless device or object that is arranged and operates via wireless communication, such as a printer, a scanner, a desktop and / or portable computer, a portable data assistant (PDA), a communication satellite, any equipment or location associated with a radio-detectable tag, and a telephone. This includes at least Wi-Fi and Bluetooth wireless technologies. Accordingly, the communication may be a predefined structure, as in a conventional network, or simply an ad hoc communication between at least two devices.

[0121] Wi-Fi (Wireless Fidelity) enables connections to the Internet and other devices without wires. Wi-Fi is a wireless technology that allows devices, such as computers, to send and receive data anywhere within the coverage area of ​​a base station, both indoors and outdoors, similar to cell phones. Wi-Fi networks use wireless technologies called IEEE 802.11 (a, b, g, etc.) to provide secure, reliable, and high-speed wireless connections. Wi-Fi can be used to connect computers to each other, to the Internet, and to wired networks (using IEEE 802.3 or Ethernet). Wi-Fi networks can operate in the unlicensed 2.4 and 5 GHz radio bands, at data rates of, for example, 11 Mbps (802.11a) or 54 Mbps (802.11b), or in products that include both bands (dual-band).

[0122] Those skilled in the art will appreciate that information and signals may be represented using any of a variety of different technologies and techniques. For example, the data, instructions, commands, information, signals, bits, symbols, and chips referenced in the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any combination thereof.

[0123] Those skilled in the art will appreciate that the various illustrative logical blocks, modules, processors, means, circuits, and model steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, various forms of programs or design code (referred to herein, for convenience, as software), or a combination of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Those skilled in the art may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

[0124] The various embodiments presented herein can be implemented as a method, apparatus, or article of manufacture using standard programming and / or engineering techniques. The term article of manufacture includes a computer program, carrier, or media accessible from any computer-readable storage device. For example, computer-readable storage media include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, magnetic strips, etc.), optical disks (e.g., CDs, DVDs, etc.), smart cards, and flash memory devices (e.g., EEPROMs, cards, sticks, key drives, etc.). Furthermore, various storage media presented herein include one or more devices and / or other machine-readable media for storing information.

[0125] It should be understood that the specific order or hierarchy of steps in the presented processes is merely an example of exemplary approaches. It should be understood that the specific order or hierarchy of steps in the processes may be rearranged within the scope of the present disclosure based on design priorities. The appended method claims provide elements of various steps in a sample order, but are not intended to be limited to the specific order or hierarchy presented.

[0126] The description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments without departing from the scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the embodiments disclosed herein, but is to be construed in the broadest scope consistent with the principles and novel features disclosed herein.

[0127] The embodiments of the present invention described above are not implemented solely through devices and methods. They may also be implemented through programs that implement functions corresponding to the configurations of the embodiments of the present invention, or through recording media containing such programs. Such recording media may be executed not only on servers but also on user terminals.

[0128] Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements made by those skilled in the art using the basic concept of the present invention defined in the following claims also fall within the scope of the present invention.

[0129]

[0130] The present invention aims to provide a method, device and computer program for predicting pulmonary complications after esophagectomy using deep learning, which can predict pulmonary complications after esophagectomy by combining the presence and extent of interstitial abnormalities in the lung and previously reported risk factors for postoperative pulmonary complications.

Claims

1. A method for predicting pulmonary complications after esophagectomy using deep learning. A step of receiving a preoperative chest CT image of one or more patients; A step of extracting the entire lung volume from the chest CT image and inputting the chest CT image into a trained deep learning model to detect a lesion area; A step of calculating the degree of emphysema by dividing the percentage value of the area with a density below a specific HU (Hounsfield Units) among the detected areas by the total lung volume; and A step of generating a nomogram by multivariate analysis of the calculated degree of emphysema and one or more postoperative pulmonary complication (PPC) factors; A method comprising:

2. In paragraph 1, The step of detecting the above lesion area is: A method wherein the above lesion is related to interstitial lung abnormality (ILA), and the parameters of consolidation, ground-glass opacity (GGO), reticulation, and honeycombing are detected in the lesion area.

3. In paragraph 1, The steps for calculating the degree of the above emphysema are: A method wherein the above specific HU is -950HU.

4. In paragraph 1, The steps for generating the above nomogram are: A method wherein at least one of the PPC factors is at least one of age, smoking status, FEV1 level, anastomosis level, and lung inflammation.

5. In paragraph 1, The steps for generating the above nomogram are: A method for performing at least one of the above multivariate analysis, a t-test, a Mann-Whitney u-test, a chi-square test, a Fisher's exact test, a Spearman correlation coefficient, and a logistic regression analysis.

6. In paragraph 1, A method further comprising a step of training a deep learning model using a chest CT image as input and a lesion area among the chest CT images as output.

7. In paragraph 6, The step of training the above deep learning model is: A method for training the deep learning model by dividing the above lesion area into consolidation, ground-glass opacity (GGO), reticulation, and honeycombing areas.

8. In paragraph 1, A method comprising: a step of predicting whether PPC occurs by inputting the degree of emphysema of a specific patient and one or more PPC factors into a generated nomogram; 9. A device for predicting pulmonary complications after esophagectomy using deep learning. a processor comprising one or more cores; and memory; Including, The above processor, Receiving preoperative chest CT images of one or more patients, Extract the entire lung volume from the received chest CT image, input the CT image into the trained deep learning model to detect the lesion area, The degree of emphysema is calculated by dividing the percentage of areas with a density below a certain number of HU (Hounsfield Units) among the detected areas by the total lung volume. A nomogram was created by multivariate analysis of the degree of emphysema and one or more postoperative pulmonary complication (PPC) factors, and A device for predicting the occurrence of PPC by inputting the degree of emphysema of a specific patient and one or more other PPC factors into a generated nomogram.

10. A computer program stored in a computer-readable storage medium and including commands that cause a computer to perform the following operations, wherein the operations are: An action of receiving preoperative chest CT images of one or more patients; An operation of extracting the entire lung volume from a received chest CT image and inputting the CT image into a trained deep learning model to detect a lesion area; An operation to calculate the degree of emphysema by dividing the percentage value of the area with a density below a certain HU (Hounsfield Units) (low attenuated area) among the detected areas by the total lung volume; An operation to create a nomogram by multivariately analyzing the degree of emphysema and one or more postoperative pulmonary complication (PPC) factors; and A computer program stored in a computer-readable storage medium, comprising: an operation for predicting whether PPC occurs by inputting the degree of emphysema of a specific patient and one or more other PPC factors into a generated nomogram;

Images