Construction method of ALS prognosis model based on chest CT skeletal muscle quantitative parameters

Abstract

The invention discloses a construction method of an ALS prognosis model based on chest CT skeletal muscle quantitative parameters. The method comprises the following steps: firstly, performing regular follow-up visit on an ALS receptor which is clearly diagnosed; imageJ software is used for quantifying skeletal muscle area / index / attenuation values (SMD / SMD / SMI) and paravertebral muscle area / attenuation values (PMA / PMD) of the first lumbar vertebra (L1) level of chest CT of all receptors and performing inter-group difference analysis, and it is found that the consistency of L1SMA / SMD / SMI / PMA / PMD of an ALS receptor group is reduced along with staging progress. According to the method, an ALS death risk prediction model is constructed through a single-factor and multi-factor COX proportional risk model, and a column diagram is drawn, so that L1SMA (HR is equal to 0.96, 95% CI is equal to 0.94-0.98, p is equal to 0.001), L1SMD (HR is equal to 0.92, 95% CI is equal to 0.88-0.96, plt; 0.001) and L1PMA (HR = 1.06, 95% CI = 1.01-1.11, plt; 0.022) significantly influences the survival rate of ALS, and areas under curves (AUC) of one-year survival prediction and three-year survival prediction are 0.687 and 0.851 respectively.

Description

Technical Field

[0001] The present invention relates to a method for constructing an ALS prognostic model based on chest CT skeletal muscle quantitative parameters; specifically, to a skeletal muscle quantitative parameter model at the first lumbar vertebrae level of chest CT based on ALS receptors and a method for constructing the same. Background Art

[0002] Amyotrophic lateral sclerosis (ALS) is a rapidly progressive and fatal neurodegenerative disease whose prognosis is closely related to skeletal muscle status. Existing evidence suggests that quantitative chest CT parameters at the L1 vertebral level (including skeletal muscle index, density, and area) are significantly associated with respiratory prognosis. This is due to the critical role of the muscles at this level (psoas major, rectus abdominis, and paraspinal muscles) in maintaining respiratory function. Although electrophysiological and imaging studies have suggested that denervation of lower thoracic muscles is associated with ventilatory impairment and that abnormal lumbar muscle metabolism is associated with mortality, the prognostic value of L1 CT skeletal muscle parameters in ALS remains unclear. Summary of the Invention

[0003] Purpose of the invention: The purpose of the present invention is to provide a method for constructing an ALS prognostic model based on chest CT skeletal muscle quantitative parameters, which can verify the correlation between L1 level chest CT skeletal muscle quantitative parameters and the risk of ALS death.

[0004] Technical solution: The method for constructing an ALS prognostic model based on chest CT skeletal muscle quantitative parameters described in the present invention comprises the following steps:

[0005] (1) Obtain chest CT and clinical evaluation data from multiple ALS recipients;

[0006] (2) identify and quantify skeletal muscle quantitative parameters at the first lumbar vertebral level based on chest CT of multiple ALS recipients;

[0007] (3) Based on the skeletal muscle quantitative parameters, determine the skeletal muscle quantitative parameters related to survival through univariate Cox regression analysis;

[0008] (4) Screening of skeletal muscle quantitative parameters or clinical indicators related to survival by multivariate Cox regression analysis to determine the prognostic indicators and risk scoring model for constructing the risk scoring model;

[0009] (5) A nomogram was constructed based on the indicators screened by the above multivariate Cox regression model, and the predictive performance of the risk scoring model was evaluated.

[0010] Furthermore, the operation process of step (2) is:

[0011] (21) L1 SMA: ImageJ software was used to adjust the threshold range between -29 and 150 HU to cover all muscles within the region of interest and automatically correct the boundaries; then the boundaries of the ROI were manually corrected by selecting muscle groups; at the same time, the average density within the ROI was calculated and recorded;

[0012] (22) L1 PMA: Based on L1 SMA, the bilateral paraspinal muscles were distinguished by manually correcting the boundaries of the ROI using ImageJ; subsequently, the ROI region was automatically generated.

[0013] Furthermore, the muscle groups selected in step (21) include the psoas major, erector spinae, quadratus lumborum, latissimus dorsi, transverse abdominal muscles, internal oblique muscles and lateral abdominal muscles.

[0014] Furthermore, in step (21), the L1 SMI is obtained by dividing the L1 SMA by the square of the subject's height.

[0015] Furthermore, the PMA in step (22) includes the erector spinae and quadratus lumborum muscles; the average density within the ROI is recorded as L1 PMD.

[0016] Furthermore, the skeletal muscle quantitative parameters in step (3) include: L1 SMA, L1 SMD, L1SMI, L1 PMA, and L1 PMD.

[0017] Furthermore, the univariate cox method in step (3) is to use the coxph function of the survival package in R to perform regression modeling on all skeletal muscle quantitative parameters or clinical characteristics respectively, and screen the prognosis-related skeletal muscle quantitative parameters or clinical characteristics with p<0.05 to enter the multivariate cox analysis.

[0018] Furthermore, the multivariate Cox regression analysis in step (4) is as follows: the cph function in the R package rms is used to construct a Cox proportional hazard regression model of skeletal muscle quantitative parameters and clinical characteristics, and then the survival package is used to calculate the survival probability.

[0019] Furthermore, the step (5) is to construct a nomogram using the nomogram function, and draw a time-dependent receiver operating characteristic curve of the training data set to analyze and evaluate the goodness of fit of the scoring model.

[0020] Furthermore, the skeletal muscle quantitative parameters ultimately used to construct the risk scoring model in step (5) include: L1SMA, L1 SMD, and L1 PMA.

[0021] Beneficial effects: Compared with the existing technology, the present invention has the following significant advantages: the quantitative skeletal muscle parameters of chest CT of the amyotrophic lateral sclerosis prognosis model constructed by the present invention are accurate, non-invasive and rapid, which increases the value of quantitative skeletal muscle evaluation of chest CT in opportunistic examination for ALS prognosis evaluation; it provides recipients with new disease progression monitoring and evaluation indicators. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 Schematic diagram of specific regions for skeletal muscle quantification at the first lumbar vertebra level of a recipient's chest CT scan in an embodiment of the present invention; wherein (a) skeletal muscle area; (b) paraspinal muscle area;

[0023] Figure 2 This is a comparison chart showing the consistent progression of skeletal muscle quantitative parameters at the first lumbar vertebra level of the recipient's chest CT scan along with the disease stage in the embodiment of the present invention;

[0024] Figure 3 This is a nomogram constructed based on the multivariate COX regression results of skeletal muscle quantitative parameters at the first lumbar vertebra level of chest CT in an embodiment of the present invention;

[0025] Figure 4 Receiver operating characteristic curves and decision curves for the nomogram model of the training set for 1 year and 3 years in the embodiment of the present invention. DETAILED DESCRIPTION

[0026] The specific technical solutions of the present invention are further described in detail below with reference to specific examples.

[0027] like Figure 1-3 As shown; the present invention provides a method for constructing an ALS prognosis model based on chest CT skeletal muscle quantitative parameters, comprising the following steps:

[0028] (1) Obtain chest CT and clinical assessment data for multiple amyotrophic lateral sclerosis (ALS) recipients;

[0029] (2) Identify and quantify skeletal muscle quantitative parameters at the first lumbar vertebral level based on chest CT of multiple amyotrophic lateral sclerosis receptors (ALS receptors);

[0030] (3) Based on the skeletal muscle quantitative parameters (L1 SMA, L1 SMD, L1 SMI, L1 PMA, L1 PMD), determine the skeletal muscle quantitative parameters related to survival by univariate Cox regression analysis;

[0031] (4) Screening of skeletal muscle quantitative parameters or clinical indicators related to survival by multivariate Cox regression analysis to determine the prognostic indicators and risk scoring model for constructing the risk scoring model;

[0032] (5) A nomogram was constructed based on the indicators screened by the above multivariate Cox regression model, and the predictive performance of the risk scoring model was evaluated.

[0033] The operation process of step (2) is: identifying and separating the CT plain film that best represents the L1 vertebra and importing it into ImageJ software,

[0034] (21) L1 SMA: ImageJ software was used to adjust the threshold range between -29 and 150 HU to cover all muscles within the region of interest (ROI) and automatically correct the boundaries; the boundaries of the ROI were then manually corrected by selecting muscle groups, including the psoas major, erector spinae, quadratus lumborum, latissimus dorsi, transverse abdominis, internal oblique, and lateralis abdominis; the size of the generated ROI was measured in square centimeters ( Figure 1 a); At the same time, the average density within the ROI, namely the average attenuation density (L1 SMD inHU), was calculated and recorded; L1 SMI was calculated by dividing L1 SMA by the square of the subject's height (cm 2 / m 2 ) obtained;

[0035] (22) L1 PMA: Based on L1 SMA, the bilateral paraspinal muscles were distinguished by manually correcting the boundaries of the ROI using ImageJ; the PMA mainly included the erector spinae and quadratus lumborum muscles; then, the ROI region was automatically generated ( Figure 1 b), the average density within the ROI was recorded as L1 PMD;

[0036] The univariate Cox method in step (3) is: use the coxph function of the survival package in R to perform univariate Cox regression modeling on all skeletal muscle quantitative parameters or clinical characteristics, and enter the multivariate Cox analysis with prognosis-related skeletal muscle quantitative parameters or clinical characteristics with p < 0.05.

[0037] The multivariate Cox regression analysis in step (4) is as follows: using the cph function in the R package rms to construct a multivariate Cox proportional hazard regression model including skeletal muscle quantitative parameters and clinical characteristics, and then using the survival package to calculate the survival probability; the multivariate Cox regression analysis screens the skeletal muscle quantitative parameters related to survival, and finally determines the prognostic indicators and risk scoring model used to construct the risk scoring model;

[0038] In step (5), a nomogram function is used to construct a nomogram, and a time-dependent receiver operating characteristic curve of the training data set is drawn to analyze and evaluate the goodness of fit of the scoring model; finally, the skeletal muscle quantitative parameters used to construct the risk scoring model include: L1 SMA, L1 SMD, and L1 PMA.

[0039] The construction process is as follows:

[0040] Recipient enrollment and assessment methods: Chest CT scans were obtained from consecutive recipients diagnosed with ALS; the diagnosis of definite, probable, or possible ALS was based on the modified Awaji criteria; if multiple scans were available, the CT scan closest to the onset of ALS symptoms was included; after excluding one recipient with concurrent tumor, three recipients with poor CT image quality, and four recipients lost to follow-up, 102 ALS recipients were enrolled in the study ( Figure 1 ); A total of 102 healthy controls (HC) were included in the physical examination center;

[0041] The following statistical and medical data were collected from the recipients' medical records, including sex, age, height, weight, BMI, onset date, onset site, revised ALS Functional Rating Scale-R (ALSFRS-R) score, ALSFRS-R respiratory subgroup (ALSFRS-RR), disease progression rate calculated as ([48-ALSFRS-R] / [time from onset to clinical assessment]), King clinical stage, smoking history, and tracheostomy date; data on auxiliary examinations were collected, including forced vital capacity (FVC); in addition, the date and indication of CT examination were also recorded; complete clinical data of 102 ALS recipients were available except FVC; only 40 recipients had complete and valid FVC data; the sample size of 102 cases met the principle of 10 events per variable for Cox regression analysis, and the final calculation required 84-100 recipients;

[0042] CT quantitative method:

[0043] Chest CT examinations were performed using a high-resolution CT scanner (Discovery CT750 HD scanner; chest CT scanning parameters were defined as follows: tube voltage under automatic exposure control was 120 kVp, tube current was 260 mA, and rotation time was 1.0 s; CT scan reconstruction section width was 5 mm;

[0044] To ensure inter-reader reproducibility and consistency of slices, chest CT scans of ALS recipients and HCs were reviewed using a standardized image viewing platform. CT slice numbers that best represented the L1 vertebra were systematically identified, separated, and randomly assigned. Subsequently, the numbered slices were imported into ImageJ software for semi-automatic muscle quantification. Hounsfield unit (HU) thresholds were used at the L1 vertebral level to identify different tissues and calculate SMA, SMD, SMI, PMA, and PMD. The HU boundaries of muscles were set between -29 and +150. Referring to the previously adopted method, the procedure was as follows: (1) L1 SMA: ImageJ software was used to adjust the threshold range between -29 and 150 HU to cover all muscles within the region of interest (ROI) and automatically correct the boundaries. The boundaries of the ROI were then manually corrected by selecting muscle groups, including the psoas major, erector spinae, quadratus lumborum, latissimus dorsi, transverse abdominis, internal oblique, and lateralis abdominis muscles. The size of the generated ROI was measured in square centimeters ( Figure 1 a); At the same time, the average density within the ROI, namely the mean attenuation density (L1 SMD in HU), was calculated and recorded; L1 SMI was calculated by dividing L1 SMA by the square of the subject's height (cm 2 / m 2 ); (2) L1 PMA: Based on L1 SMA, the boundaries of the ROI were manually corrected using ImageJ to distinguish the bilateral paraspinal muscles; the PMA mainly included the erector spinae and quadratus lumborum muscles; then, the ROI area was automatically generated ( Figure 1 b), the average density within the ROI was recorded as L1PMD.

[0045] Statistical analysis: Statistical methods included the use of the Kolmogorov-Smirnov test to assess normality, expressing continuous variables as mean ± standard deviation (SD) for normal distribution and median as interquartile range (IQR) for nonnormal distribution; recipients with missing data were excluded; categorical variables were expressed as frequencies and percentages; comparisons were made using the t-test for normally distributed continuous data, the Mann-Whitney U test for nonnormal data, and the chi-square test for categorical variables; analysis of variance and post hoc tests were used to analyze the normal distribution of King stage; Kruskal-Wallis and Bonferroni corrections were performed for nonnormal and ordered categorical variables; Spearman correlations were used to assess relationships; survival analysis defined the endpoint as death or tracheostomy initiation; Cox regression identified predictive factors; p < 0.05 was entered into the multivariate model; nomograms were tested, and discrimination (concordance index, area under the curve) and decision curve analysis (DCA) were assessed; statistical significance: p < 0.05; R software (v4.1.2) was used.

[0046] Results: Univariate Cox regression analysis identified statistically significant survival indicators (p < 0.05): ALSFRS-R, King clinical stage, BMI, disease site, L1 SMA, L1 SMD, L1 SMI, L1 PMA, and L1 PMD. In subsequent multivariate analysis, King clinical stage (stage 3 HR = 2.3, 95% CI = 1.15-4.59, p = 0.018; stage 4a-b HR = 3.43, 95% CI = 1.64-7.16, p = 0.001), L1 SMA (HR = 0.96, 95% CI = 0.94-0.98, p = 0.000), L1 SMD (HR = 0.92, 95% CI = 0.88-0.95, p < 0.001), and L1 PMA (HR=1.06, 95% CI=1.01-1.11, p=0.022) was identified as an independent prognostic factor affecting survival.

Claims

1. A method for constructing an ALS prognostic model based on chest CT skeletal muscle quantitative parameters, characterized in that: The steps are as follows: (1) Obtain chest CT and clinical characteristics data of multiple ALS recipients; (2) identify and quantify skeletal muscle quantitative parameters at the first lumbar vertebral level based on chest CT of multiple ALS recipients; (3) Based on the skeletal muscle quantitative parameters, the skeletal muscle quantitative parameters related to the recipient survival period were determined by univariate Cox regression analysis; (4) Screening of skeletal muscle quantitative parameters or clinical characteristic indicators related to survival through multivariate Cox regression analysis to determine the prognostic indicators and risk scoring model for constructing the risk scoring model; (5) A nomogram was constructed based on the indicators screened by the above multivariate Cox regression model, and the predictive performance of the risk scoring model was evaluated.

2. The method for constructing an ALS prognostic model based on chest CT skeletal muscle quantitative parameters according to claim 1, characterized in that: The operation process of step (2) is: (21) L1 SMA: ImageJ software was used to adjust the threshold range between -29 and 150 HU to cover all muscles within the region of interest and automatically correct the boundaries; then the boundaries of the ROI were manually corrected by selecting muscle groups; at the same time, the average density within the ROI was calculated and recorded; (22) L1 PMA: Based on L1 SMA, the bilateral paraspinal muscles were distinguished by manually correcting the boundaries of the ROI using ImageJ; subsequently, the ROI region was automatically generated.

3. The method for constructing an ALS prognostic model based on chest CT skeletal muscle quantitative parameters according to claim 2, characterized in that: The muscle groups selected in step (21) include the psoas major, erector spinae, quadratus lumborum, latissimus dorsi, transverse abdominal muscles, internal oblique muscles and lateral abdominal muscles.

4. The method for constructing an ALS prognostic model based on chest CT skeletal muscle quantitative parameters according to claim 2, characterized in that: In step (21), the L1 SMI is obtained by dividing the L1 SMA by the square of the subject's height.

5. The method for constructing an ALS prognostic model based on chest CT skeletal muscle quantitative parameters according to claim 2, characterized in that: The PMA in step (22) includes the erector spinae and quadratus lumborum muscles; the average density within the ROI is recorded as L1PMD.

6. The method for constructing an ALS prognostic model based on chest CT skeletal muscle quantitative parameters according to claim 1, characterized in that: The skeletal muscle quantitative parameters in step (3) include: L1 SMA, L1 SMD, L1 SMI, L1 PMA, and L1PMD.

7. The method for constructing an ALS prognostic model based on chest CT skeletal muscle quantitative parameters according to claim 1, characterized in that: The single-factor Cox method in step (3) is to use the coxph function of the survival package in R to perform regression modeling on all skeletal muscle quantitative parameters or clinical characteristics respectively, and screen the prognosis-related skeletal muscle quantitative parameters or clinical characteristics with p<0.05 to enter the multivariate Cox analysis.

8. The method for constructing an ALS prognostic model based on chest CT skeletal muscle quantitative parameters according to claim 1, characterized in that: The multivariate Cox regression analysis in step (4) is as follows: the cph function in the R package rms is used to construct a Cox proportional hazard regression model of skeletal muscle quantitative parameters and clinical characteristics, and then the survival package is used to calculate the survival probability.

9. The method for constructing an ALS prognostic model based on chest CT skeletal muscle quantitative parameters according to claim 1, characterized in that: The step (5) is to construct a nomogram using the nomogram function, and draw a time-dependent receiver operating characteristic curve of the training data set to analyze and evaluate the goodness of fit of the scoring model.

10. The method for constructing an ALS prognostic model based on chest CT skeletal muscle quantitative parameters according to claim 1, characterized in that: The skeletal muscle quantitative parameters ultimately used to construct the risk scoring model in step (5) include: L1SMA, L1 SMD, and L1 PMA.

Images