JP2020057053A - Postoperative complication prediction method, postoperative complication prediction program, and postoperative complication prediction device - Google Patents

Abstract

To predict the occurrence of postoperative complications with high accuracy.SOLUTION: A postoperative complication prediction method includes a determination step and a display step. In the determination step, a computer follows a predictive model for predicting a risk of the occurrence of postoperative complications constructed based on background factors of a first sample in which postoperative complications have occurred within a predetermined period and a second sample in which postoperative complications have not occurred and determines a risk of the occurrence of postoperative complications in a subject from the background factors of the subject measured preoperatively. In the display step, the computer displays a risk determination result on a display device.SELECTED DRAWING: Figure 13

Description

  The present invention relates to a postoperative complication prediction method, a postoperative complication prediction program, and a postoperative complication prediction device.

  After surgery with a high degree of surgical invasion, the risk of complications is relatively high. On the other hand, there is a technique for generating a discriminator (discrimination factor set) having high discrimination accuracy for a disease from a sample data file (for example, see Patent Document 1). Regarding complications, the technique of Patent Document 1 mathematically identifies the types of independent discriminant factors (risk factors).

JP 2008-90833 A

  Discriminants for complications vary. Also, in the real world, individual patients have a variety of different background factors. Even if discriminants are identified, it is difficult to predict the occurrence of complications in individual patients with high accuracy.

  In order to solve the above-mentioned problems, a method for predicting postoperative complications according to one aspect of the present invention includes a computer, wherein a first sample in which postoperative complications occur within a predetermined period and the postoperative complications occur. According to a prediction model constructed based on background factors with no second sample and predicting the risk of occurrence of post-operative complications, the subject is subjected to surgery based on background factors of the subject measured before surgery. A determination step of determining a risk of occurrence of a post-complication; and a display step of the computer displaying the determination result of the risk on a display device.

  According to the present invention, it is possible to predict the occurrence of postoperative complications with high accuracy.

FIG. 1 is a diagram showing a configuration of a postoperative complication predicting apparatus according to the present embodiment. FIG. 2 shows a cohort of patient samples. FIG. 3 is a diagram showing a comparison of HRQoL between the group with postoperative complications and the group without postoperative complications. FIG. 4 is a diagram showing the results of a univariate risk analysis for postoperative complications for a training sample. FIG. 5 is a diagram showing the results of a multivariate risk analysis for postoperative complications for a training sample. FIG. 6 is a diagram illustrating the accuracy rates of the five decision tree models according to the present embodiment. FIG. 7 is a diagram illustrating the overall accuracy rate of the decision tree model for the training sample and the test sample. FIG. 8A is a diagram illustrating a first decision tree model according to the present embodiment. FIG. 8B is a diagram showing a continuation of FIG. 8A. FIG. 8C is a diagram showing a continuation of FIG. 8B. FIG. 8D is a diagram showing a continuation of FIG. 8B. FIG. 8E is a diagram showing a continuation of FIG. 8A. FIG. 9A is a diagram illustrating a second decision tree model according to the present embodiment. FIG. 9B is a diagram showing a continuation of FIG. 9A. FIG. 9C is a diagram showing a continuation of FIG. 9B. FIG. 10A is a diagram illustrating a third decision tree model according to the present embodiment. FIG. 10B is a diagram showing a continuation of FIG. 10A. FIG. 10C is a diagram showing a continuation of FIG. 10B. FIG. 10D is a diagram showing a continuation of FIG. 10C. FIG. 11A is a diagram illustrating a fourth decision tree model according to the present embodiment. FIG. 11B is a diagram showing a continuation of FIG. 11A. FIG. 11C is a diagram showing a continuation of FIG. 11B. FIG. 11D is a diagram showing a continuation of FIG. 11B. FIG. 11E is a diagram showing a continuation of FIG. 11A. FIG. 12A is a diagram illustrating a fifth decision tree model according to the present embodiment. FIG. 12B is a diagram showing a continuation of FIG. 12A. FIG. 12C is a diagram showing a continuation of FIG. 12B. FIG. 12D is a diagram showing a continuation of FIG. 12C. FIG. 13 is a diagram showing a typical flow of the first danger determination process by the processing circuit of FIG. FIG. 14 is a diagram showing an example of a display screen of the determination result displayed in step SA4 of FIG. FIG. 15 is a diagram showing a typical flow of the second danger determination process by the processing circuit of FIG.

  Hereinafter, a postoperative complication prediction method, a postoperative complication prediction program, and a postoperative complication prediction device according to an embodiment of the present invention will be described with reference to the drawings.

  The postoperative complication prediction method according to the present embodiment is a method in which a person or a computer predicts whether or not a patient to be operated will develop complications after the operation. The postoperative complication prediction program is a program that causes a computer to realize the function of the postoperative complication prediction method. The postoperative complication prediction device is a computer that executes a postoperative complication prediction method according to a postoperative complication prediction program. The postoperative complication predicting device may be a computer having a circuit configuration equivalent to the program and executing the postoperative complication predicting method using the circuit configuration.

  FIG. 1 is a diagram illustrating an example of the configuration of the postoperative complication predicting device 1. As shown in FIG. 1, the postoperative complication predicting device 1 includes a processing circuit 11, a storage device 12, a display device 13, an input device 14, and a communication device 15. The processing circuit 11, the storage device 12, the display device 13, the input device 14, and the communication device 15 are configured to be able to send and receive signals to and from each other via a bus.

  The processing circuit 11 has, as hardware resources, a processor such as a CPU (Central Processing Unit) and a memory such as a ROM (Read Only Memory) and a RAM (Random Access Memory). The processor functions as a patient factor input unit 111, an image measurement unit 112, a model selection unit 113, a risk determination unit 114, and a display control unit 115 by executing the post-operative complication prediction program stored in the memory. Having.

  The patient factor input unit 111 inputs information on factors (hereinafter, referred to as patient factors) used for predicting postoperative complications for a patient whose postoperative complications are to be predicted. Patient factors include at least patient background factors. The background factors include a patient basic factor and an image measurement factor. The patient basic factor is demographic information of the patient such as age and gender. The image measurement factor is information that can be measured by a radiographic image with respect to the form, positional relationship, and the like of the surgical target site. Patient factors may further include surgical invasive factors. Surgical invasive factors are factors that characterize the surgical procedure performed on a patient. The information of the patient basic factor and the image measurement factor is measured or collected before the operation. Surgical invasive factor information is measured or collected during or after surgery.

  The image measurement unit 112 performs image processing on the radiation image of the patient to measure information of an image measurement factor. For example, a radiation image is collected in advance by a medical image diagnostic apparatus such as an X-ray diagnostic apparatus, an X-ray computed tomography apparatus, a magnetic resonance imaging apparatus, an ultrasonic diagnostic apparatus, a nuclear medicine diagnostic apparatus, and stored in the storage device 12. . Note that when the image measurement information has already been measured by another device, the image measurement by the image measurement unit 112 is unnecessary.

  The model selection unit 113 selects a prediction model 121 used for predicting a postoperative complication of the target patient. The prediction model 121 is generated in advance by another computer and stored in the storage device 12. The prediction model 121 is an algorithm for inputting at least a background factor and outputting a risk of occurrence of a postoperative complication. The model selection unit 113 may automatically select a prediction model based on a patient factor or the like, or may select the prediction model according to a user's instruction via the input device 14.

  The risk determination unit 114 determines the risk of a postoperative complication occurring in the prediction target patient from at least background factors related to the prediction target patient collected before the operation, in accordance with the prediction model selected by the model selection unit 113. . The risk determining unit 114 may determine the risk based on the background factor and the surgical invasive factor.

  The display control unit 115 displays various information on the display device 13. For example, the display control unit 115 displays, on the display device 13, the determination result of the postoperative complication by the risk determination unit 114.

  The storage device 12 is a storage device such as a hard disk drive (HDD), a solid state drive (SSD), or an integrated circuit storage device for storing various information. For example, the storage device 12 stores the prediction model 121. The storage device 12 may store a plurality of prediction models 121, or may store a single prediction model 121. As hardware, the storage device 12 may be a drive device that reads and writes various information from and to a portable storage medium such as a CD-ROM drive, a DVD drive, or a flash memory.

  The display device 13 displays various information. As the display device 13, for example, a CRT display, a liquid crystal display, an organic EL display, an LED display, a plasma display, or any other display known in the art can be appropriately used. The display device 13 may be a projector.

  The input device 14 receives various commands from a user. As the input device 14, a keyboard, a mouse, various switches, and the like can be used. An output signal from the input device 14 is supplied to the processing circuit 11 via a bus.

  The communication device 15 performs data communication with a medical image diagnostic apparatus or another computer via a cable or wireless (not shown).

  Hereinafter, prediction of a postoperative complication by the postoperative complication prediction device 1 according to the present embodiment will be described. In addition, the prediction model 121 according to the present embodiment can be any type of postoperative invasive surgical operation such as amputation, resection, excision, anastomosis, transplantation, plastic surgery, correction, etc. It can also be used to predict disease. In the following description, the prediction model 121 is a model for predicting a complication after orthopedic surgery, specifically, a surgical surgery for adult spinal deformity (ASD). And

  ASD is a heterogeneous and complex spinal disease that often requires surgery. Surgical spine correction for ASD is invasive but effective as a symptomatic treatment for ASD. On the other hand, conservative therapy for ASD often fails. ASD surgery has been shown to significantly improve the health-related quality of life (HRQoL) for two years after surgery. On the other hand, the degree of relief of pain and disease by conservative treatment is lower than that of surgery. Although ASD surgery provides such favorable results, it requires large incisions in the body, intervertebral spinal fusion, osteotomy, blood transfusions and prolonged hospital stays, and is associated with both perioperative and postoperative long-term. Risk of postoperative complications.

The post-operative complications of ASD predicted by the predictive model 121 include all complications that can occur during the two years after the operation. Major postoperative complications include nerve damage, implant-related (proximal and distal junctional kyphosis, rod breakage, pseudoarthrosis, implant misalignment, screw breakage) , Surgical site infection, other infections (such as urinary tract infection), excessive bleeding (> 2000 mL), delirium, cardiopulmonary disease (hemodynamic), myocardial infarction, deep vein thrombosis, pulmonary embolism, lung dilatation Insufficiency, congestive heart failure, etc.), gastrointestinal disorders (intestinal obstruction, cholecystitis), and kidney disorders (acute renal failure). Re-operation to cope with post-operative complications should be avoided as far as possible from the viewpoint of the physical and mental burden of patients and doctors, medical economics, and the like. Therefore, it is very useful to predict preoperatively whether the patient is at risk of causing postoperative complications.

  Next, generation of the prediction model 121 according to the present embodiment will be described. The predictive model 121 was generated based on information on risk factors for patients who underwent surgical spine correction for ASD. These multiple patients include patients who have suffered at least one complication during the two years following surgery and patients who have not suffered from one complication. Specifically, 170 patient samples were used. The information of 170 patients was used for the construction and internal verification of the prediction model, and the information of 25 patients was used for external verification.

  FIG. 2 shows a cohort of patient samples. These patients have undergone surgery for older ASDs aged 50 and over. The ASD according to the present embodiment is limited to those satisfying a preset standard. The criterion is a Cobb angle of 20 degrees or more, a C7 sagittal vertical axis (C7SVA) of 5 cm or more, or a pelvic tilt (PT) of 25 degrees or more. In addition, these ASD patients have a minimum of five or more vertebral body fusions (fused vertebral levels) and a segmental pedicle screw from an upper-instrumented vertebral (UIV) level to a lower-instrumented vertebral (LIV) level. The patient has undergone pedicle screw fixation and has been able to be followed up for a full two years after surgery. Patients with inadequate radiographic images or spinal deformities associated with syndromic, neuromyogenic or other pathological conditions are excluded.

As shown in FIG. 2, regarding the cohort of the patient sample according to the present embodiment, the number of patients, age, gender, BMI [kg / m ² ], BMD (T score), number of fixed vertebral bodies, vertebral osteotomy (PSO) ), Reoperation, Schwab-SRS classification TypeT, Schwab-SRS classification TypeD, Schwab-SRS classification TypeL, Schwab-SRS classification TypeN, C7SVA [cm], PI-LL [°], PT [°], TPA [°] And the median and standard deviation were calculated for the items of the number of complications. The number in parentheses in the value column indicates the percentage of the whole. Subjects show severe deformation in the sagittal plane as represented by C7VSA and PI-LL. The postoperative complication rate is 48% (81/170) two years after the operation and 28% (48/81) during the perioperative period (30 days after the operation). In the two years after surgery, the most common postoperative complications are implant-related (50 cases, 30% of the total), followed by nervous (17 cases, 10% of the total).

  The patient samples consisted of a patient sample group with complications occurring two years after ASD surgery (hereinafter referred to as a group with complications) and a patient sample group without complications occurring for two years after ASD surgery (hereinafter referred to as complications). (Referred to as a group without symptoms).

  FIG. 3 is a diagram showing a comparison of HRQoL between a group with a complication and a group without a complication. As shown by the ODI (Oswestry Disability Index) (P <0.01) measured before the operation (Baseline) and 2 years after the operation, and SRS22r, the HRQoL was significantly improved in both groups. The function, pain, self-image, mental health, satisfaction, and total of SRS22r were uncomplicated in 2 years after surgery in the uncomplicated group. The SRS22 score and ODI are deteriorated as compared with the group with the disease. On the other hand, preoperative HRQoL is similar between the two groups.

  Analysis of complication risk factors in patients undergoing ASD surgery. Risk factors include the above-mentioned basic patient factors, surgical invasion, image measurement, and other factors. As basic patient factors and risk factors for surgical invasion, specifically, the patient's age, gender, BMI, bone mineral density (BMD), spinal surgery history, frailty, coexistence, Schwab- The ARS classification and subclassification of SRS, application of pedicle subtraction osteotomy (PSO), high UIV, high LIV, and the number of posterior fused vertebrae (variables) were defined by the present inventors. . Frailty and co-morbidity were assessed by the modified frailty index (mFI) and the Charlesson co-morbidity index (CCI). BMD was calculated from the dual energy X-ray absorptiometry (DXA) score of the right femoral neck. Radiographic images of the standing whole spine were analyzed before surgery, 6 weeks after surgery and 2 years after surgery. The risk factors relating to image measurement include, specifically, Cobb angle, C7SVA, T5-T12 thoracic kyphosis (TK), lumbar lordosis angle (LL), sacral slope (SS), pelvic slope (PT) Spinal column alignment variables such as pelvic incidence angle (PI) and PI-LL were defined by the inventor. In the case of another surgical operation, a factor corresponding to the surgical operation is appropriately selected as a risk factor relating to the image measurement factor. In addition, instead of HRQoL, a variable between ODI and SRS-22r (Scoliosis Research Society-22r questionnaire) was defined by the present inventors as a risk factor. ODI and SRS-22r were recorded at baseline and two years after surgery.

  Among these risk factors, the factors used to construct a predictive model were identified by computer using logistic regression analysis. The relationship between risk factors and the occurrence of postoperative complications was analyzed by computer by univariate and multivariate logistic regression analysis using all information of the training samples for 117 persons.

  The risk factors that were evaluated to be strongly associated with postoperative complications by univariate and multivariate logistic regression analysis are, specifically, (1) age, (2) gender, (3) BMI, (4) BMD (5) mFI, (6) CCI, (7) UIV high and LIV high, (8) levels involved, (9) PSO applied, (10) C7SVA, (11) PI-LL (12) Baseline ODI, (13) Baseline SRS22r total score, (14) Baseline SRS22r nociceptive domain score, (15) Baseline SRS22r functional domain score, (16) Baseline SRS22r mental status score, (17) Estimation It is blood loss (estimated blood loss) and (18) time of surgery (TOS). Some of the above risk factors were classified as follows. Age: 50-74 and above 75, BMD (T score): below -1.5 and above, Frailty: Robust (mFI = 0), pre-weak (mFI <0.21), frail (mFI> 0.21), UIV High: T1-T6 (proximal thoracic: PT) and T9-T11 (lower thoracic: LT), LIV High: L5 and above, pelvis, EBL: 1000 mL and below, TOS: 5 hours and above. In the present embodiment, the risk factors evaluated to be strongly associated with postoperative complications are synonymous with the patient factors.

  In univariate logistic regression analysis, Overall summary statics were calculated. Global summary statistics include the mean and standard deviation for continuous values, frequencies and percentages for classification variables. After descriptive analysis, univariate comparisons analyzed independent associations between potential risk factors and major postoperative complications. As a univariate risk analysis, an unpaired t-test and an HSD (Tukey's Honest Significant Difference) test or a Wilcoxon ranking test were performed.

  FIG. 4 is a diagram showing the results of a univariate risk analysis for postoperative complications for a training sample. As shown in FIG. 4, the risk factors analyzed for univariate risk analysis were: age, gender, BMI, BMD, frailty, CCI, C7VSA, PI-LL, revision surgery, number of fixed vertebral bodies, high UIV, LIV High, PSO application and operative time. Odds ratios, P values, and 95% confidence intervals were calculated for each risk factor. Univariate risk analysis showed elderly (age> 75), gender (male), frailty (mFI> 0.21), low bone density (T score ≤-1.5), high LIV (pelvis) and PI- LL> 30 ° was determined to be a significant risk factor for major postoperative complications.

  In a multivariate analysis, a multivariate binary logistic regression model was generated to evaluate the adjusted relevance of each potential explanatory variable and predict the likelihood of occurrence of major postoperative complications . Multivariate logistic regression analysis was performed on clinically relevant variables and variables with a univariate significance level of 0.25 or less.

  FIG. 5 is a diagram showing the results of a multivariate analysis of post-operative complications for a training sample. As shown in FIG. 5, the risk factors subjected to the multivariate analysis are BMD, PI-LL, and frailty. Multivariate analysis showed that BMD (T score <-1.5), spinopelvic alignment PI-LL> 30 ° and weakness (mFI class) were independent risk factors for major postoperative complications after ASD surgery Was found to be.

  After the risk analysis, a predictive model was built to predict the risk of developing postoperative complications after ASD surgery. As a prediction model, any machine learning model such as a neural network, a support vector machine, or a decision tree, provided that at least a background factor is input and an algorithm that outputs a determination result with respect to the risk of occurrence of postoperative complications is provided. May be used. Hereinafter, for the sake of specific description, it is assumed that the algorithm of the prediction model is a decision tree model having a decision tree structure.

  Several decision tree models were generated by computer based on the patient factors of the training samples. To build a decision tree model, the objective variables of the decision tree model are at risk or no risk of developing postoperative complications two years after ASD surgery. That is, the decision tree model according to the present embodiment is configured to be able to predict whether or not a complication will occur within two years after the operation. Note that the decision tree model according to the present embodiment may be generated by the processing circuit 11 of the postoperative complication predicting device 1.

  The computer constructs a decision tree model by randomly changing the type of the variable of the patient factor, the threshold of the determination condition, and the order of the determination condition. The computer compares the overall accuracy between the constructed decision tree models and the area under a receiver operating characteristic curve (AUC), and determines the overall accuracy of the constructed decision tree models. A plurality of decision tree models having a high correct answer rate and an AUC value are selected as the prediction model 121 according to the present embodiment.

  FIG. 6 is a diagram illustrating the correct answer rates of the five decision tree models. FIG. 7 is a diagram illustrating the overall accuracy rate of the decision tree model for the training sample and the test sample. As shown in FIG. 6, the correct answer rate is 90.2% for the first decision tree model, 79.6% for the second decision tree model, 86.3% for the third decision tree model, 88.4% for the fourth decision tree model and 77.3% for the fifth decision tree model. Thus, the correct answer rate of about 75% or more was secured for all decision tree models. As shown in FIG. 7, the overall accuracy rate of the prediction model using all the decision tree models in FIG. 6 is 92.3% for the training sample. Internal verification of the test sample for the prediction model indicated a correct answer rate of 92.5% and an AUC of 0.963. For samples from independent centers different from the patient sample used for the decision tree model, the predictive model showed a 84% accuracy rate. As described above, the decision tree model according to the present embodiment has a high accuracy rate and is highly reliable.

Next, a decision tree model according to the present embodiment will be described with reference to FIGS. 8A to 8E are diagrams illustrating a first decision tree model. The first decision tree model is generated based on information of background factors and surgical invasive factors of the training sample. As shown in FIG. 8, the first decision tree model includes a plurality of nodes provided in a tree shape. Each node is defined by a determination condition regarding each of a plurality of variables included in the background factor and the surgical invasive factor. The types of the variables of the background factors are age, gender, BMI [kg / m ² ], BMD, frailty, CCI, C7SVA [cm], PI-LL [°], TPA [°], number of fixed vertebral bodies, UIV At least one of high and LIV high. The types of variables included in the surgical invasive factor include at least one of the presence or absence of application of PSO, the operation time [minutes], and the EBL [mL].

  In each node, the number of patients and the ratio satisfying the determination condition are defined for each of the group without complications and the group with complications in the training sample. Each node determines whether the determination condition is satisfied (true) or not (false). Until the terminal node is reached, the judgment of the judgment condition is repeated at each node. Similarly, the true / false determination of the determination condition set for the terminal node is repeated at the terminal node. After the true / false determination, it is determined whether or not the ratio of the group having a complication or the group having no complication to the determination result exceeds a threshold. For example, when the ratio of the group with a complication exceeds a first threshold (hereinafter, referred to as a threshold), it is determined that there is a high risk of developing postoperative complications, and the ratio of the group without a complication is the second. If it exceeds the threshold value (hereinafter referred to as no threshold value), it is determined that the risk of developing postoperative complications is low. In other cases, it is determined that the determination is impossible. The threshold value and the non-threshold value may be set, for example, to a value of 90% to 100%, but may be set to values greater than 50%. In the following description, the threshold with and without the threshold is assumed to be 90%.

  Specifically, as shown in FIG. 8A, it is determined whether or not PSO has been applied at start node 0. Note that there is 3CO (three column osteotomy) as an example of the PSO. In the first decision tree model, when 3CO is applied, it is considered that PSO is applied, and when 3CO is not applied, it is considered that PSO is not applied. If it is determined that the PSO has been applied, the process moves to the node 1. At the node 1, it is determined whether the UIV high rank is UT. If the UIV high is UT, it is moved to the end node 2. As shown in terminal node 2, the percentage of the uncomplicated group is 100% for UT patients, which is larger than the no threshold. Therefore, when it is determined in the node 1 that the UIV high rank is UT, it is determined that there is no danger.

  If it is determined that the UIV high rank is not a UT at the node 1, in other words, it is an LT, the process proceeds to the node 3. In the node 3, it is determined whether the PI-LL is equal to or less than 34.100. If the PI-LL is less than or equal to 34.100, the process moves to the terminal node 4. As shown in the terminal node 4, for the patients with PI-LL of 34.100 or less, the ratio of the group without complication is 66.667%, which is smaller than the no threshold, and the ratio of the group with complication is 33.333. %, Which is smaller than the threshold value. Therefore, when it is determined in the node 3 that the PI-LL is equal to or less than 34.100, it is determined that “the determination is impossible”. If PI-LL is greater than 34.100, it is moved to the end node 5. As shown in the terminal node 5, for the patient whose PI-LL is larger than 34.100, the ratio of the uncomplicated group is 11.111%, which is smaller than the no threshold, and the ratio of the group with a complication is 88. Since it is 889%, it is smaller than the threshold value. Therefore, when it is determined in the node 5 that the PI-LL is greater than 34.100, it is determined that “the determination is impossible”.

  If it is determined in the start node 0 that PSO is not applied, the process proceeds to node 6. In the node 6, it is determined whether or not C7SVA is 11.0 or less. If C7SVA is less than or equal to 11.0, the process moves to the terminal node 7. As shown in the end node 7, for the patient whose C7SVA is 11.0 or less, the ratio of the uncomplicated group is 98.630%, which is larger than the no threshold. Therefore, when it is determined that C7SVA is 11.0 or less at the node 6, it is determined that there is no danger.

  If it is determined in the node 6 that C7SVA is larger than 11.0, the process proceeds to the node 8. At node 8, it is determined whether the number of fixed vertebral bodies is less than or equal to 13.0. If the number of fixed vertebral bodies is less than or equal to 13.0, the process moves to node 9. At the node 9, it is determined whether the BMI is 17.898 or less. If the BMI is less than or equal to 17.898, the process moves to the terminal node 10. As shown in the terminal node 10, for the patient whose BMI is 17.898 or less, the ratio of the uncomplicated group is 100%, which is larger than the no threshold. Therefore, when it is determined that the BMI is 17.898 or less at the node 9, it is determined that there is no danger.

  As illustrated in FIG. 8B, when it is determined that the BMI is greater than 17.898 at the node 9, the processing shifts to the node 11. At node 11, it is determined whether the number of fixed vertebral bodies is less than or equal to 7.0. If the number of fixed vertebral bodies is less than or equal to 7.0, the process moves to node 12. The node 12 determines whether the TPA is 13.0 or less. If the TPA is less than or equal to 13.0, the process moves to the terminal node 13. As shown in the terminal node 13, the ratio of the uncomplicated group is 21.595% for the patient whose TPA is 13.0 or less, which is smaller than the no threshold value, and the ratio of the complication group is 78.405%. Therefore, it is smaller than the threshold value. Therefore, when it is determined in the node 12 that the TPA is 13.0 or less, it is determined that “the determination is impossible”. If it is determined in the node 12 that the TPA is larger than 13.0, the process proceeds to the terminal node 14. As shown in the terminal node 14, the proportion of the uncomplicated group is 92.872% for the patients with TPA greater than 13.0, which is larger than the no threshold. Therefore, when it is determined in the node 12 that the TPA is larger than 13.0, it is determined that there is no danger.

  If it is determined in the node 11 that the number of fixed vertebral bodies is greater than 7, the process proceeds to the node 15. In the node 15, it is determined whether the PI-LL is 36.0 or less. As shown in FIG. 8C, when PI-LL is equal to or less than 36.0, the processing is shifted to the node 16. At node 16, it is determined whether the LIV high is the pelvis, L5 and above. If the LIV high is the pelvis, the process moves to node 17. At node 17, it is determined whether the number of fixed vertebral bodies is less than or equal to 12.0. If the number of fixed vertebral bodies is less than or equal to 12.0, the process moves to node 18. At the node 18, it is determined whether the EBL is less than or equal to 323.0. If the EBL is 323.0 or less, the process moves to the terminal node 19. As shown in the end node 19, the ratio of the uncomplicated group is 89.969% for the patients with an EBL of 323.0 or less, which is smaller than the no threshold value, and the ratio of the group with a complication is 10.031%. Therefore, it is smaller than the threshold value. Therefore, when it is determined in the node 18 that the EBL is 323.0 or less, it is determined that “the determination is impossible”.

  If it is determined in the node 18 that the EBL is larger than 323.0, the process moves to the terminal node 20. As shown in the end node 20, for the patients whose EBL is greater than 323.0, the ratio of the uncomplicated group is 19.247%, which is smaller than the no threshold value, and the ratio of the complication group is 80.755%. Therefore, it is smaller than the threshold value. Therefore, when it is determined in the node 18 that the EBL is larger than 323.0, it is determined that “the determination is impossible”.

  When it is determined in the node 17 that the number of fixed vertebral bodies is larger than 12.0, the process shifts to the terminal node 21. As shown in the end node 21, for the patient having a fixed vertebral body number larger than 12.0, the ratio of the uncomplicated group is 100%, which is larger than the no threshold. Therefore, when it is determined in the node 17 that the number of fixed vertebral bodies is larger than 12.0, it is determined that there is no risk.

  If it is determined at the node 16 that the LIV high order is L5 or higher, the process moves to the node 22. At the node 22, it is determined whether the BMI is equal to or less than 19.044. When it is determined that the BMI is equal to or less than 19.044, the process moves to the terminal node 23. As shown in the terminal node 23, for the patients with a BMI of 19.044 or less, the ratio of the group without complication is 33.333%, which is smaller than the no threshold, and the ratio of the group with complication is 66.667%. Therefore, it is smaller than the threshold value. Therefore, when it is determined in the node 22 that the BMI is equal to or more than 19.044, it is determined that “the determination is impossible”. If the node 22 determines that the BMI is larger than 19.044, the process moves to the terminal node 24. As shown in the terminal node 24, the ratio of the uncomplicated group is smaller than the no-threshold value for the patient with a BMI of greater than 19.044, which is 85.326%, and the ratio of the group with a complication is 14.674%. Therefore, it is smaller than the threshold value. Therefore, when it is determined in the node 22 that the BMI is larger than 19.044, it is determined that “the determination is impossible”.

  As illustrated in FIG. 8D, when it is determined in the node 15 that the PI-LL is greater than 36.0, the processing shifts to the node 25. At node 25, it is determined whether the gender is male or female. If it is determined that the gender is male, the process moves to the terminal node 26. As shown in the terminal node 26, the ratio of the group with a complication is 100% for a patient whose gender is male, which is larger than the threshold value. Therefore, when it is determined in the node 26 that the gender is male, it is determined that there is a risk.

  If it is determined that the gender is female at the node 25, the process proceeds to the node 27. At the node 27, it is determined whether or not the BMD (T score) is equal to or less than -1.700. If it is determined that the BMD is equal to or less than −1.700, the process moves to the terminal node 28. As shown in the end node 28, the ratio of the group with a complication is 95.470% for a patient with a BMD of −1.700 or less, which is larger than the threshold value. Therefore, when it is determined that the BMD is equal to or less than −1.700 at the node 27, it is determined that there is “danger”.

  If it is determined in the node 27 that the BMD is larger than −1.700, the process proceeds to the node 29. At the node 29, it is determined whether the PI-LL is equal to or less than 52.900. If the PI-LL is less than or equal to 52.900, the process moves to the node 30. At the node 30, it is determined whether the operation time is 231.0 minutes or less. If the operation time is 231.0 minutes or less, the process moves to the end node 31. As shown in the end node 31, for the patient whose operation time is 231.0 minutes or less, the ratio of the group without complication is 76.891%, which is smaller than the no threshold value, and the ratio of the group with complication is 23.109. %, Which is smaller than the threshold value. Therefore, when it is determined in the node 30 that the operation time is 231.0 minutes or less, it is determined that “the determination is impossible”. If it is determined in the node 30 that the operation time is longer than 231.0 minutes, the process shifts to the terminal node 32. As shown in the terminal node 32, the ratio of the uncomplicated group is 11.790% for the patient whose operation time is longer than 231.0 minutes, which is smaller than the no threshold, and the ratio of the group with a complication is 88.210. %, Which is smaller than the threshold value. Therefore, when it is determined in the node 30 that the operation time is longer than 231.0 minutes, it is determined that “the determination is impossible”.

  When it is determined in the node 29 that the PI-LL is larger than 52.900, the processing shifts to the node 33. At the node 33, it is determined whether the age is 60 or less. If the age is 60 or less, the process moves to the terminal node 34. As shown in the terminal node 34, the ratio of the uncomplicated group is 12.699% for the patients whose age is 60 or less, which is smaller than the no threshold, and the ratio of the group with the complication is 87.301%. It is smaller than the threshold. Therefore, when it is determined in the node 33 that the age is 60 or less, it is determined that “the determination is impossible”. If the node 33 determines that the age is higher than 60, the process shifts to the terminal node 35. As shown in the terminal node 35, the proportion of the uncomplicated group is 88.444% for the patients over the age of 60, which is smaller than the no threshold, and the proportion of the group with the complication is 11.556%. Therefore, it is smaller than the threshold value. Therefore, when it is determined in the node 35 that the age is higher than 60, it is determined that “the determination is impossible”.

  As shown in FIG. 8E, if it is determined that the number of fixed vertebral bodies is greater than 13.0 at node 8, the process moves to node 36. At the node 36, it is determined whether the age is 68 or less. If it is determined that the age is 68 or less, the process moves to the terminal node 37. As shown in the terminal node 37, the percentage of the group with a complication is 100% for a patient with an age of 68 or less, which is larger than the threshold. Therefore, when it is determined in the node 36 that the age is 68 or less, it is determined that there is a danger.

  If it is determined at node 36 that the age is above 68, the process moves to node 38. At the node 38, it is determined whether the BMI is equal to or less than 22.680. If it is determined that the BMI is equal to or less than 22.680, the process moves to the terminal node 39. As shown in the terminal node 39, the ratio of the uncomplicated group is 100% for patients with a BMI of 22.680 or less, which is larger than the no threshold. Therefore, if the node 38 determines that the BMI is 22.680 or less, it is determined that there is no danger. If the node 38 determines that the BMI is greater than 22.680, the process moves to the terminal node 40. As shown in the terminal node 40, the ratio of the group with a complication is 100% for a patient with a BMI greater than 22.680, which is larger than the threshold. Therefore, when it is determined in the node 38 that the BMI is larger than 22.680, it is determined that “there is a risk”.

  9A to 9C are diagrams illustrating a configuration of the second decision tree model. As shown in FIG. 9A, at the start node 41, it is determined whether the gender is male or female. If it is a man, it is transferred to the terminal node 42. As shown in the terminal node 42, when it is determined in the start node 41 that the gender is male, it is determined that “the determination is impossible”.

  If the gender is determined to be female at the node 41, the process proceeds to the node 43. At the node 43, it is determined whether the LIV high level is the pelvis, L5 and higher. If it is determined that the LIV high is the pelvis, the process moves to node 44. At the node 44, it is determined whether the TPA is 52.0 or less. When it is determined that the TPA is 52.0 or less, the process moves to the terminal node 45. As shown in the terminal node 45, when it is determined in the node 44 that the TPA is 52.0 or less, it is determined that “the determination is impossible”. When it is determined in the node 44 that the TPA is larger than 52.0, the process is shifted to the terminal node 46. As shown in the terminal node 46, when it is determined in the node 44 that the TPA is larger than 52.0, it is determined that “the determination is impossible”.

  When it is determined that the LIV high level is L5 or higher at the node 43, the process proceeds to the node 47. At the node 47, it is determined whether the BMD is equal to or less than -1.500. If the BMD is determined to be −1.500 or less, the process moves to the terminal node 48. As shown in the terminal node 48, when it is determined in the node 47 that the BMD is equal to or less than −1.500, it is determined that “the determination is impossible”.

  As illustrated in FIG. 9B, when the node 47 determines that the BMD is greater than −1.500, the process moves to the node 49. At node 49, it is determined whether the operation time is 432.0 minutes or less. If it is determined that the operation time is 432.0 minutes or less, the process proceeds to node 50. At node 50, it is determined whether the UIV high is UT or LT. If the UIV high rank is UT, the process moves to the terminal node 51. As shown in the terminal node 51, when it is determined that the UIV high rank is UT in the node 51, it is determined that "determination is impossible".

  If the node 50 determines that the UIV high order is LT, the process moves to the node 52. At the node 52, it is determined whether the PI-LL is 38.0 or less. When it is determined that the PI-LL is 38.0 or less, the processing shifts to the node 53. At the node 53, it is determined whether the PI-LL is 0.700 or less. When it is determined that the PI-LL is 0.700 or less, the processing is shifted to the terminal node 54. As shown in the terminal node 54, when it is determined in the node 53 that the PI-LL is 0.700 or less, it is determined that “the determination is impossible”. When it is determined in the node 53 that the PI-LL is larger than 0.700, the processing shifts to the terminal node 55. As shown in the terminal node 55, the ratio of the uncomplicated group is 100% with respect to the patient whose PI-LL is greater than 0.700, which is larger than the no threshold. Therefore, when it is determined in the node 53 that the PI-LL is greater than 0.700, it is determined that there is no danger.

  As shown in FIG. 9C, when it is determined that the PI-LL is greater than 38.0 in the node 52, the process proceeds to the node 56. At the node 56, it is determined whether or not the BMD is equal to or less than -1.300. When it is determined that the BMD is equal to or less than −1.300, the process moves to the terminal node 57. As shown in the terminal node 57, for the patient whose BMD is −1.300 or less, the ratio of the group without complication is 100%, which is larger than the threshold. Therefore, when the node 56 determines that the BMD is equal to or less than -1.300, it is determined that there is no danger.

  If the node 56 determines that the BMD is greater than −1.300, the process moves to the node 58. At node 58, it is determined whether the age is less than or equal to 62.312. If the age is less than or equal to 62.312, the process moves to node 59. At the node 59, it is determined whether or not C7SVA is equal to or less than 145.0. If it is determined that C7SVA is equal to or less than 145.0, the process moves to the terminal node 60. As shown in the terminal node 60, when C7SVA is determined to be 145.0 or less at the node 60, it is determined to be “undeterminable”. When it is determined in the node 59 that C7SVA is larger than 145.0, the process proceeds to the terminal node 61. As shown in the terminal node 61, the ratio of the group with a complication is 95.463% for patients with C7SVA greater than 145.0, which is larger than the threshold value. Therefore, when it is determined in the node 61 that C7SVA is greater than 145.0, it is determined that there is a risk.

  If it is determined at node 58 that the age is above 62.312, the process moves to node 62. At the node 62, it is determined whether or not C7SVA is equal to or less than 184.0. If it is determined that C7SVA is equal to or less than 184.0, the process proceeds to the terminal node 63. As shown in the terminal node 63, the ratio of the group with a complication is 91.975% for patients with a C7SVA of 184.0 or less, which is larger than the threshold value. Therefore, when it is determined in the node 62 that C7SVA is equal to or less than 184.0, it is determined that there is a risk. If it is determined in the node 62 that C7SVA is greater than 184.0, the process proceeds to the terminal node 64. As shown in the terminal node 64, the ratio of the uncomplicated group is 100% for patients with a C7SVA of greater than 184.0, which is larger than the threshold. Therefore, when it is determined in the node 62 that C7SVA is larger than 184.0, it is determined that there is no danger.

  If it is determined in the node 49 that the operation time is longer than 432.0 minutes, the process shifts to the terminal node 65. As shown in the terminal node 65, the ratio of the group with a complication is 98.803% for a patient whose operation time is longer than 432.0 minutes, which is larger than the complication existence threshold. Therefore, when it is determined in the node 49 that the operation time is longer than 432.0, it is determined that there is a risk.

  FIG. 10A to FIG. 10D are diagrams illustrating the configuration of the third decision tree model. As shown in FIG. 10A, it is determined at the start node 66 whether PSO has been applied. If it is determined that the PSO has been applied, the process moves to the node 67. At the node 67, it is determined whether the LIV high level is UT or LT. If it is determined to be a UT, the process moves to the terminal node 68. As shown in terminal node 68, the percentage of the uncomplicated group is 100% for patients with UT, which is greater than the no threshold. Therefore, when it is determined in the node 67 that the LIV high is UT, it is determined that there is no risk of the postoperative complication.

  If it is determined in the node 67 that the LIV high order is LT, the process moves to the node 69. At the node 69, it is determined whether the PI-LL is 52.0 or less. When it is determined that the PI-LL is 52.0 or less, the process proceeds to the node 70. At node 70, it is determined whether the age is 59.252 or less. If it is determined that the age is 59.252 or less, the process moves to the terminal node 71. As shown in the terminal node 71, the percentage of the group with a complication is 100% for a patient whose age is 59.252 or less, which is larger than the threshold value. Therefore, when it is determined in the node 70 that the age is 59.252 or less, it is determined that there is a danger. If it is determined at the node 70 that the age is greater than 59.252, the process moves to the terminal node 72. As shown in the terminal node 72, when it is determined in the node 70 that the age is higher than 59.252, it is determined that “the determination is impossible”.

  When it is determined in the node 69 that the PI-LL is larger than 52.0, the processing shifts to the terminal node 73. As shown in the terminal node 72, the ratio of the group with the complication is 100% for the patient whose PI-LL is larger than 52.0, which is larger than the threshold value. Therefore, when it is determined in the node 69 that the PI-LL is larger than 52.0, it is determined that there is a risk.

  If it is determined in the node 66 that PSO is not applied, the processing shifts to the node 74. At the node 74, it is determined whether or not C7SVA is 11.0 or less. When it is determined that C7SVA is equal to or less than 11.0, the process proceeds to the terminal node 75. As shown in the terminal node 75, the ratio of the uncomplicated group is 98.073% for patients with C7SVA of 11.0 or less, which is larger than the no threshold. Therefore, when it is determined in the node 74 that C7SVA is 11.0 or less, it is determined that there is no danger.

  As shown in FIG. 10B, when it is determined that C7SVA is greater than 11.0 at the node 74, the process proceeds to the node 76. At node 76, it is determined whether the UIV high is UT or LT. If it is determined that it is a UT, the process moves to the node 77. At the node 77, it is determined whether the BMI is equal to or less than 22.600. When it is determined that the BMI is equal to or less than 22.600, the processing shifts to the node 78. At node 78, it is determined whether the EBL is less than or equal to 1000.0. When it is determined that the EBL is equal to or less than 1000.0, the process moves to the terminal node 79. As shown in the terminal node 79, when it is determined in the node 78 that the EBL is equal to or less than 1000.0, it is determined that “the determination is impossible”. If the node 78 determines that the EBL is greater than 1000.0, the process moves to the terminal node 80. As shown in the terminal node 80, the ratio of the group with a complication is 95.811% for a patient with an EBL greater than 1000.0, which is larger than the threshold value. Therefore, when it is determined that the EBL is larger than 1000.0 at the node 78, it is determined that there is a risk.

  If the node 77 determines that the BMI is larger than 22.600, the process moves to the terminal node 81. As shown in the terminal node 81, the ratio of the group with a complication is 100% for a patient with a BMI of greater than 22.600, which is larger than the threshold. Therefore, when it is determined in the node 77 that the BMI is greater than 22.600, it is determined that there is a risk.

  As illustrated in FIG. 10C, when it is determined that the high level of the UIV is LT at the node 76, the process proceeds to the node 82. At node 82, it is determined whether the operation time is less than or equal to 254.0. If the operative time is less than or equal to 254.0, the process moves to node 83. At the node 83, it is determined whether the EBL is 890.0 or less. If the EBL is less than or equal to 890.0, the process moves to the node 84. At node 84, it is determined whether the age is 70.533 or less. When it is determined that the age is 70.533 or less, the process moves to the terminal node 87. As shown in the terminal node 87, when it is determined that the age is 70.533 or less, it is determined that “the determination is impossible”.

  If the node 84 determines that the age is above 70.553, it is determined whether the age is 71.408 or less. When it is determined that the age is 71.408 or less, the process moves to the terminal node 87. As shown in the terminal node 87, the ratio of the group with a complication is 93.398% for a patient whose age is 71.408 or less, which is larger than the threshold value. Therefore, when it is determined in the node 86 that the age is 71.408 or less, it is determined that there is a risk. If it is determined at node 86 that the age is above 71.408, the process moves to terminal node 88. As shown in terminal node 88, the percentage of the uncomplicated group is 91.970% for patients older than 71.408, which is larger than the no threshold. Therefore, when it is determined in the node 86 that the age is 71.408 or less, it is determined that there is no danger.

  If the EBL is greater than 890.0 at the node 83, the process moves to the terminal node 89. As shown in the terminal node 89, when it is determined in the node 83 that the EBL is larger than 890.0, it is determined that “the determination is impossible”.

  If it is determined at the node 82 that the operation time is longer than 254.0, the process moves to the node 90. At node 90, it is determined whether the LIV high is the pelvis, L5 and above. When it is determined that the LIV high position is the pelvis, the process shifts to the terminal node 91. As shown in the terminal node 91, when it is determined in the node 91 that the LIV high level is the pelvis, it is determined that “the determination is impossible”.

  As illustrated in FIG. 10D, when it is determined that the LIV high level is L5 or higher at the node 90, the process moves to the node 92. At the node 92, it is determined whether the operation time is 432.0 or less. If the operation time is 432.0 or less, the process moves to the node 93. At the node 93, it is determined whether or not C7SVA is 108.0 or less. If it is determined that C7SVA is equal to or less than 108.0, the process proceeds to the node 94. At node 94, it is determined whether the number of fixed vertebral bodies is less than or equal to 10.0. If it is determined that the number of fixed vertebral bodies is 10.0 or less, the process proceeds to node 95. At the node 95, it is determined whether the age is 62.493 or less. If the age is 62.493 or less, the process moves to the terminal node 96. As shown in the terminal node 96, the percentage of the uncomplicated group is 100% for the patients whose age is 62.493 or less, which is larger than the no threshold. Therefore, when it is determined in the node 95 that the age is 62.493 or less, it is determined that there is no danger. If it is determined in the node 95 that the age is higher than 62.493, the process moves to the terminal node 97. As shown in the terminal node 79, when it is determined in the node 95 that the age is higher than 62.493, it is determined that “the determination is impossible”.

  If it is determined at node 94 that the number of fixed vertebral bodies is greater than 10.0, the process moves to terminal node 98. As shown in the terminal node 98, the ratio of the uncomplicated group is 100% for patients with a fixed vertebral body number larger than 10.0, which is larger than the no threshold value. Therefore, when it is determined in the node 94 that the number of fixed vertebral bodies is larger than 10.0, it is determined that there is no risk.

  If it is determined in the node 93 that C7SVA is greater than 108.0, the process shifts to the terminal node 99. As shown in the terminal node 99, when it is determined in the node 93 that C7SVA is larger than 108.0, it is determined that “the determination is impossible”.

  If it is determined at the node 92 that the operation time is longer than 432.0, the process shifts to the terminal node 100. As shown in the terminal node 100, the ratio of the group with a complication is 95.784% for a patient whose operation time is longer than 432.0, which is larger than the threshold value. Therefore, when it is determined in the node 92 that the operation time is longer than 432.0, it is determined that there is a risk.

  11A to 11E are diagrams illustrating a configuration of a fourth decision tree model. As shown in FIG. 11A, it is determined whether or not PSO has been applied at the start node 101. If it is determined that the PSO has been applied, the process moves to the node 102. The node 102 determines whether the LIV high is UT or LT. If it is determined to be a UT, the process moves to the terminal node 103. As shown in the terminal node 103, the percentage of the group without complication is 100% for the UT patient, which is larger than the no threshold. Therefore, when it is determined that the LIV high rank is UT at the node 103, it is determined that there is no risk of the postoperative complication.

  If it is determined in the node 102 that the LIV high order is LT, the process moves to the node 104. At node 104, it is determined whether the gender is male or female. If it is determined that the gender is male, the process moves to the terminal node 105. As shown in the terminal node 105, the ratio of the group with a complication is 100% for a male patient with a gender, which is larger than the threshold value. Therefore, when it is determined in the node 104 that the gender is male, it is determined that there is a risk.

  If the gender is determined to be female at the node 104, the process proceeds to the node 106. At the node 106, it is determined whether the BMI is equal to or less than 30.700. When the BMI is equal to or less than 30.700, the processing shifts to the terminal node 107. As shown in the terminal node 107, it is determined that the patient whose BMI is equal to or less than 30.700 is "indeterminable". For the patient whose BMI is greater than 30.700 at the terminal node 106, the ratio of the uncomplicated group is 100%, so that it is greater than the no threshold. Therefore, when it is determined in the node 106 that the BMI is greater than 30.700, it is determined that there is no danger.

  If it is determined in the start node 101 that PSO is not applied, the process moves to the node 109. At node 109, it is determined whether the number of fixed vertebral bodies is less than or equal to 10.0. As shown in FIG. 10B, if the number of fixed vertebral bodies is 10.0 or less, the process moves to node 110. At node 110, it is determined whether the LIV high is the pelvis, L5 and above. When it is determined that the LIV high position is the pelvis, the process proceeds to the node 111. At the node 111, it is determined whether the age is 77.0 or less. If it is determined that the age is 77.0 or less, it is determined whether the number of fixed vertebral bodies is 9.0 or less. If the number of fixed vertebral bodies is less than or equal to 9.0, the process moves to node 113. At node 113, it is determined whether the age is 65 or less. If the age is 65 or less, the process moves to the terminal node 114. At the terminal node 114, the percentage of the group without complication is 100% for the patient whose age is 65 or less, which is larger than the no threshold. Therefore, when it is determined in the node 113 that the age is 65 or less, it is determined that there is no danger.

  As shown in FIG. 11C, when it is determined that the age is higher than 65 at the node 113, the process proceeds to the node 115. At the node 115, it is determined whether or not C7SVA is 62.0 or less. If it is determined that C7SVA is 62.0 or less, the process moves to the terminal node 116. As shown in the terminal node 116, it is determined that the patient whose C7SVA is equal to or less than 62.0 is “indeterminable”.

  When it is determined in the node 115 that C7SVA is larger than 62.0, the process proceeds to the node 117. At the node 117, it is determined whether the BMI is equal to or less than 20.135. If the BMI is less than or equal to 20.135, the process moves to the terminal node 118. As shown in the terminal node 118, the proportion of the group with a complication is 91.592% for the patient with a BMI of 20.135 or less, which is larger than the threshold value. Therefore, when it is determined in the node 117 that the BMI is equal to or less than 20.135, it is determined that there is a risk. When it is determined in the node 117 that the BMI is larger than 20.135, the process shifts to the terminal node 119. As shown in the terminal node 119, the proportion of the uncomplicated group is 100% for patients with a BMI greater than 20.135, which is larger than the no threshold. Therefore, when it is determined in the node 119 that the BMI is larger than 20.135, it is determined that there is no danger.

  If it is determined in the node 112 that the number of fixed vertebral bodies is larger than 9.0, the process shifts to the terminal node 119. As shown in the terminal node 119, the ratio of the group having a complication is 100% for the patient having the number of fixed vertebral bodies larger than 9.0, so that the ratio is larger than the threshold value. Therefore, when it is determined in the node 112 that the number of fixed vertebral bodies is larger than 9.0, it is determined that “there is a risk”.

  If the node 111 determines that the age is higher than 77.0, the process moves to the terminal node 121. As shown in the terminal node 121, the percentage of the group without complications is 100% for patients whose age is higher than 77.0, which is larger than the no threshold. Therefore, when it is determined that the age is higher than 77.0 at the node 111, it is determined that there is no danger.

  As illustrated in FIG. 11D, when it is determined whether the LIV high level is equal to or higher than L5 at the node 110, the process proceeds to the node 122. At node 122, it is determined whether the number of fixed vertebral bodies is less than or equal to 9.0. If the number of fixed vertebral bodies is 9.0 or less, the process moves to the terminal node 123. As shown in the terminal node 123, when it is determined that the number of fixed vertebral bodies is 9.0 or less, it is determined that “the determination is impossible”.

  If it is determined at node 122 that the number of fixed vertebral bodies is greater than 9.0, the process moves to node 124. At the node 124, it is determined whether the PI-LL is equal to or less than 40.800. If the PI-LL is less than or equal to 40.800, the process moves to the node 125. At node 125, it is determined whether the operation time is less than or equal to 432.0. If the operative time is less than or equal to 432.0, the process transitions to node 126. At node 126, it is determined whether the age is less than or equal to 77.022. If the age is less than or equal to 77.022, the process moves to the terminal node 127. As shown in the terminal node 127, when it is determined that the age is 77.0022 or less, it is determined that "the determination is impossible". If it is determined in the node 126 that the age is higher than 77.022, the process moves to the terminal node 128. As shown in the terminal node 128, the percentage of the group with a complication is 100% for the patient older than 77.022, which is larger than the threshold value. Therefore, when it is determined in the node 126 that the age is higher than 77.022, it is determined that there is a risk.

  If it is determined at the node 125 that the operation time is longer than 432.0, the process shifts to the terminal node 129. As shown in the end node 129, the ratio of the group with a complication is 96.426% for a patient whose operation time is longer than 432.0, which is larger than the threshold value. Therefore, when it is determined in the node 125 that the operation time is longer than 432.0, it is determined that there is a risk.

  If it is determined in the node 124 that the PI-LL is greater than 40.800, the process proceeds to the node 130. At node 130, it is determined whether the age is 65.680 or less. If the age is equal to or less than 65.680, the process moves to the node 131. In the node 131, it is determined whether the BMD is equal to or less than -1.0. When it is determined that the BMD is equal to or less than −1.0, the process moves to the terminal node 132. As shown in the terminal node 132, when the BMD is determined to be −1.0 or less at the node 131, it is determined that “the determination is impossible”. When it is determined that the BMD is larger than -1.0, the process shifts to the terminal node 133. As shown in the terminal node 133, when the node 131 determines that the BMD is larger than −1.0, it is determined that “the determination is impossible”.

  If the node 130 determines that the age is above 65.680, the process moves to the terminal node 134. As shown in the terminal node 134, the percentage of the group with a complication is 100% for the patient over the age of 65.680, which is larger than the threshold. Therefore, when it is determined in the node 134 that the age is higher than 65.680, it is determined that there is a risk.

  As shown in FIG. 11E, when it is determined that the number of fixed vertebral bodies is larger than 10.0 at the node 109, the process proceeds to the node 135. At node 135, it is determined whether the UIV high is UT or LT. If the UIV high is UT, then transition to node 136. At node 136, it is determined whether the LIV high is the pelvis, L5 and above. If the LIV high is the pelvis, then transition is made to node 137. The node 137 determines whether the BMI is equal to or less than 21.641. If the BMI is less than or equal to 21.641, the process moves to the terminal node 138. As shown in the terminal node 138, for the patient whose BMI is 21.641 or less, the ratio of the uncomplicated group is 100%, which is larger than the no threshold. Therefore, when it is determined in the node 137 that the BMI is 21.641 or less, it is determined that there is no danger. If the node 137 determines that the BMI is greater than 21.641, the process moves to the terminal node 139. As shown in the terminal node 139, the ratio of the group with a complication is 91.635% for a patient with a BMI greater than 21.641, which is larger than the threshold value. Therefore, when it is determined in the node 139 that the BMI is greater than 21.641, it is determined that there is a risk.

  If it is determined that the LIV high level is L5 or higher at the node 136, the process moves to the terminal node 140. As shown in the terminal node 140, when it is determined at the node 136 that the LIV high level is L5 or higher, it is determined that “the determination is impossible”.

  If the node 135 determines that the high level of the UIV is LT, the process moves to the terminal node 141. As shown in the terminal node 141, when it is determined at the node 135 that the UIV high level is LT, it is determined that “determination is impossible”.

  12A to 12D are diagrams illustrating a configuration of a fifth decision tree model. As shown in FIG. 12A, it is determined at the start node 142 whether the number of fixed vertebral bodies is equal to or less than 7.0. If the number of fixed vertebral bodies is less than or equal to 7.0, the process moves to the end node 143. As shown in the terminal node 143, when it is determined in the node 142 that the number of fixed vertebral bodies is equal to or less than 7.0, it is determined that “the determination is impossible”.

  If it is determined at the start node 142 that the number of fixed vertebral bodies is greater than 7.0, the process moves to the node 144. At the node 144, it is determined whether the BMD is equal to or less than -2.0. If the BMD is equal to or less than -2.0, the process moves to the terminal node 145. As shown in the terminal node 145, when the BMD is determined to be −2.0 or less at the node 144, it is determined that “the determination is impossible”.

  As illustrated in FIG. 12B, when it is determined that the BMD is greater than −2.0 at the node 144, the process proceeds to the node 146. At node 146, it is determined whether C7SVA is less than or equal to 145.0. If C7SVA is less than or equal to 145.0, the process moves to node 147. At node 147, it is determined whether the UIV high is UT or LT. If the UIV high is UT, then transition to node 148. Node 148 determines whether the BMI is less than or equal to 22.600. If the BMI is less than or equal to 22.600, the process moves to the terminal node 149. As shown in the end node 149, the proportion of the uncomplicated group is 91.159% for patients with a BMI of 22.600 or less, which is larger than the no threshold. Therefore, if the node 148 determines that the BMI is 22.600 or less, it is determined that there is no danger. When the node 148 determines that the BMI is greater than 22.600, the process moves to the terminal node 150. As shown in the terminal node 150, the ratio of the group with a complication is 100% for a patient with a BMI of greater than 22.600, which is larger than the threshold. Therefore, when it is determined in the node 148 that the BMI is larger than 22.600, it is determined that there is a risk.

  As illustrated in FIG. 12C, when it is determined that the high level of the UIV is LT at the node 147, the process proceeds to the node 151. At node 151, it is determined whether the number of fixed vertebral bodies is less than or equal to 10.0. If the number of fixed vertebral bodies is less than or equal to 10.0, the process moves to node 152. At node 152, it is determined whether PSO is applied. If there is an application of PSO, it is shifted to node 153. At the node 153, it is determined whether C7SVA is 114.0 or less. If the C7SVA is equal to or less than 114.0, the process moves to the terminal node 154. As shown in the terminal node 154, the ratio of the group having a complication is 100% with respect to the patient whose C7SVA is 114.0 or less, which is larger than the threshold value. Therefore, when it is determined in the node 153 that the C7SVA is equal to or less than 114.0, it is determined that there is a risk. If the node 153 determines that C7SVA is greater than 114.0, the process proceeds to the terminal node 155. As shown in the terminal node 155, it is determined that the patient has a C7SVA of greater than 114.0 as "indeterminable".

  As illustrated in FIG. 12D, when it is determined that the PSO is not applied in the node 152, the process proceeds to the node 156. At node 156, it is determined whether the LIV client is a pelvis, L5 and above. If the LIV high is the pelvis, then transition is made to node 157. At the node 157, it is determined whether the PI-LL is equal to or less than 7.0. If the PI-LL is less than or equal to 7.0, the process moves to the terminal node 158. As shown in the terminal node 158, the percentage of the group with a complication is 100% for patients with a PI-LL of 7.0 or less, which is larger than the threshold value. Therefore, when it is determined in the node 157 that the PI-LL is equal to or less than 7.0, it is determined that there is a risk. When it is determined in the node 157 that the PI-LL is larger than 7.0, the processing shifts to the terminal node 159. As shown in the terminal node 159, it is determined that the patient whose PI-LL is greater than 7.0 is "indeterminable".

  If it is determined at the node 156 that the LIV high level is L5 or higher, the process moves to the node 160. At node 160, it is determined whether the number of fixed vertebral bodies is less than or equal to 9.0. If the number of fixed vertebral bodies is 9.0 or less, the process moves to the end node 161. As shown in the end node 161, it is determined that “the determination is impossible” for a patient whose fixed vertebral body number is 9.0 or less. If it is determined at node 160 that the number of fixed vertebral bodies is greater than 9.0, the process moves to terminal node 162. As shown in the end node 162, it is determined that the patient has a fixed vertebral body number of more than 9.0 as "indeterminable".

  When it is determined in the node 151 that the number of fixed vertebral bodies is larger than 10.0, the process moves to the terminal node 163. As shown in the terminal node 163, it is determined that the determination is “undetermined” for a patient having a fixed vertebral body number greater than 10.0.

  If the node 146 determines that C7SVA is greater than 145.0, the process moves to the terminal node 164. As shown in the end node 164, a patient whose C7SVA is greater than 145.0 is determined to be "indeterminable".

  Thus, the description of the configuration of the five decision tree models is completed. Note that the decision tree model according to the present embodiment is not limited to the above five decision tree models. The types of variables used in the construction of the decision tree, the thresholds of the determination conditions, and the order of the determination conditions can be changed as appropriate. For example, the above five decision tree models are constructed based on both the background factor and the surgical invasive factor, and the determination conditions include the determination condition regarding the background factor and the determination condition regarding the surgical invasive factor. . However, the decision tree model may be constructed based only on the background factor, and may include only the determination condition regarding the background factor as the determination condition. This makes it possible to determine the risk of postoperative complications using only factors that can be measured and collected before surgery.

  In addition, the above five decision tree models determine the risk of complications occurring two years after the operation. However, in the decision tree model according to the present embodiment, the risk of complications may be determined for any period as long as the period is one month or more after the operation. In this case, a decision tree model may be constructed based on patient factors of a sample of a patient having a complication during a predetermined period after the operation and a sample of a patient having no complication during the predetermined period. Furthermore, the decision tree model according to the present embodiment is not limited to the determination of the risk of occurrence of general complications after ASD surgery. For example, the risk of a specific complication occurring may be determined. In this case, a decision tree model is constructed based on patient factors of a sample of a patient having a specific complication occurring in a predetermined period after the operation and a sample of a patient having no specific complication occurring in the predetermined period. I just need. In the decision tree model according to the present embodiment, the risk of occurrence of postoperative complications of any surgical operation may be determined. For example, the risk of postoperative complications of surgery for cancer may be determined. In this case, a decision tree model may be constructed based on patient factors of a sample of a patient having a complication during a predetermined period after the cancer operation and a sample of a patient having no complication during the predetermined period. As the image measurement factor, a factor according to the surgical operation may be appropriately selected.

  Next, a risk determination process for postoperative complications using these decision tree models will be described.

  FIG. 13 is a diagram showing a typical flow of the first danger determination process by the processing circuit 11. As shown in FIG. 13, first, the patient factor input unit 111 of the processing circuit 11 inputs a variable of the patient factor of the target patient (step SA1). The variable of the input patient factor corresponds to the variable of the patient factor input to the decision tree model selected in step SA2. The patient factor input in step SA1 includes at least a background factor. The value of each variable of the patient factor is input automatically or manually by the user via the input device 14. Since background factors are measured or collected before ASD surgery, they can be input before ASD surgery. It is also possible to include a surgical invasive factor as a patient factor input in step SA1. Typically, surgical invasive factors are measured or collected during or after ASD surgery. When performing the first risk determination process after the ASD operation, the patient factor input unit 111 inputs the actual measured value of the variable of the surgical invasive factor. When performing the first risk determination process before the ASD operation, the patient factor input unit 111 inputs a predicted value, a standard value, or an arbitrary value of the surgical invasive factor. The predicted value, the standard value, or the arbitrary value may be input by the user via the input device 14.

  In step SA1, a radiation image of the target patient may be input instead of the variable of the image measurement factor. In this case, the image measurement unit 112 of the processing circuit 11 may perform image processing on the input radiation image to measure each variable of the image measurement factor.

  When step SA1 is performed, the model selection unit 113 of the processing circuit 11 selects a decision tree model to be used (step SA2). In step SA2, the model selection unit 113 may select a preset decision tree model from a plurality of decision tree models stored in the storage device 12, or may select any of the decision tree models via the input device 14 by the user. May be selected. In addition, the model selection unit 113 may select a decision tree model that can execute the determination process based on the patient factor input in step SA1. The selection candidate is set, for example, in the first, second, third, fourth or fifth decision tree model.

  When step SA2 is performed, the risk determination unit 114 of the processing circuit 11 determines the risk of complication occurring two years after the ASD operation of the target patient according to the decision tree model selected in step SA2 (step SA3). ). In step SA3, the risk determination unit 114 inputs the patient factor input in step SA1 to the decision tree model, and sequentially calculates the determination conditions of each node from the start node to the terminal node according to the decision tree model, Output the judgment result. The determination result is, for example, “has danger” at which a complication occurs, “no danger” at which a complication occurs, or “cannot be determined”.

  When step SA3 is performed, the display control unit 115 of the processing circuit 11 displays the determination result output in step SA4 on the display device 13.

  FIG. 14 is a diagram illustrating an example of the display screen I1 of the determination result. As shown in FIG. 14, the display screen I1 is provided with a display column R1 of the target patient and a display column R2 of the determination result. The display column R1 displays the name of the target patient, such as “〇 山 △ 男”. Note that the information displayed in the display column R1 is not limited to the name of the target patient, and may be any information that can identify the target patient, such as a patient number. The determination result is displayed in the display column R2. For example, if the determination result is “at risk”, a message such as “there is a risk of complication occurring within two years after ASD operation” is displayed.

  As described above, according to the first risk determination process, it is possible to determine the risk of occurrence of postoperative complications based on background factors that can be collected before surgery. This makes it possible to predict the presence or absence of postoperative complications with high probability. This has advantages for both physicians and patients. As a physician, you will be able to take precautionary measures against complications. As a patient, in addition to understanding the complications of general surgery, it is possible to understand the risks associated with their own complications and to understand the surgery more deeply. In addition, by taking preventive measures in advance for a patient who is expected to have a complication, treatment costs associated with an increased reoperation can be suppressed. If it is determined that there is no risk of occurrence of post-operative complications, the doctor or patient can start the operation with confidence.

  Note that the information displayed as the determination result is not limited to a message corresponding to “has danger”, “has no danger”, or “impossible to determine”. For example, the display control unit 115 may display the ratio between the group with complications and the group without complications or the number of patients at the terminal node reached by the calculation in step SA3. By displaying the ratio and the number of patients, it is possible to estimate the danger even when it is determined that “determination is impossible”. Further, the display control unit 115 may display a threshold value and a non-threshold value.

  When step SA4 ends, the first danger determination process by the processing circuit 11 ends.

  Note that the processing flow shown in FIG. 13 can be changed as appropriate. For example, the order of step SA1 and step SA2 may be reversed. When there is only one type of prediction model, or when a decision tree model to be used is set in advance, step SA2 can be omitted.

  With respect to the selection of the decision tree model in step SA2, the following modes can be considered. There are cases where any human characteristics such as race, birth area, residence area, ethnicity, lifestyle (lifestyle), age, etc. of the patient sample used for construction of the decision tree model are limited to specific characteristics. is there. In this case, the model selection unit 113 may select a decision tree model corresponding to the personal characteristics of the target patient according to the personal characteristics of the target patient. For example, a birth area and a residence area are based on decision tree models constructed based on patient data in Asian countries, decision tree models constructed based on data on patients in Europe, and data on patients in North America. The decision tree model constructed based on the data of the patient in South America, the decision tree model constructed based on the data of the patient in Africa, and the decision tree model constructed based on the data of the patient in Africa are stored in the storage device 12. The model selection unit 113 selects a decision tree model corresponding to the birth area and the residence area of the target patient from the plurality of decision tree models. As a result, it is possible to reduce the reduction in the accuracy of predicting complications caused by differences in human characteristics.

  In the first risk determination process, only one decision tree model is used. However, the present embodiment is not limited to this. The processing circuit 11 according to the present embodiment may execute a second risk determination process of predicting a risk of complication occurring two years after the ASD operation using a plurality of decision tree models.

  FIG. 15 is a diagram showing a typical flow of the second danger determining process by the processing circuit 11. As shown in FIG. 15, first, the patient factor input unit 111 of the processing circuit 11 inputs data of the patient factor of the target patient (step SB1). The processing content of step SB1 is the same as that of step SA1.

  When step SB1 is performed, the model selection unit 113 of the processing circuit 11 selects a prediction model to be used (step SB2). In step SB2, the model selection unit 113 selects a prediction model 121 set to be used first from a plurality of decision tree models stored in the storage device. Alternatively, the model selection unit 113 may select a decision tree model arbitrarily specified by the user via the input device 14. A code to that effect is assigned to the selected decision tree model.

  When step SB2 is performed, the risk determination unit 114 of the processing circuit 11 determines the risk of complication occurring two years after the ASD operation of the target patient according to the decision tree model selected in step SB2 (step SB3) ). The processing content of step SB3 is the same as that of step SA3.

  When step SB3 is performed, the model selection unit 113 of the processing circuit 11 determines whether or not the result of the determination in step SB3 is "impossible to determine" (step SB4). When it is determined in step SB4 that the determination result is “not possible to determine” (step SB4: YES), the model selection unit 113 determines whether there is an unused decision tree model (step SB5). The unused decision tree model refers to, for example, a plurality of prediction models set as a used decision tree model among the decision tree models stored in the storage device 12. The model selection unit 113 determines whether there is a decision tree model that has not been used in the current risk determination process for the target patient among such unused decision tree models. Specifically, an unused decision tree model may be determined based on the presence or absence of the code assigned in step SB2.

  When it is determined in step SB5 that there is an unused decision tree model (step SB5: YES), the model selection unit 113 selects a next decision tree model to be used from the unused decision tree models (step SB5). SB2). Then, the risk determination unit 114 determines the risk of occurrence of a postoperative complication of the target patient according to the selected decision tree model (step SB3), and again the model selection unit 113 determines that the output result is “impossible to determine”. It is determined whether or not there is any data (step SB4). If the determination result is “not possible to determine” (step SB4: YES), the model selection unit 113 determines whether or not there is an unused decision tree model (step SB4). Step SB5). That is, in the second danger determination process, when the processing circuit 11 determines that the determination result is “impossible to determine” in the decision tree model in use, the processing circuit 11 determines whether or not there is a danger by using another decision tree model. I do.

  In this way, Step SB2 to Step SB5 are repeated, and when the determination result is determined to be “having danger” or “having no danger” in Step SB4 (Step SB4: NO), or an unused decision tree model If it is not determined that there is (Step SB5: NO), the display control unit 115 of the processing circuit 11 displays the determination result on the display device 13 (Step SB6). When the determination result is “dangerous” or “no danger” in step SB4, a message corresponding to “dangerous” or “no danger” is displayed on the display device 13. When the result of the determination is “dangerous” or “no danger” in Step SB4 and no unused decision tree model is determined to be present in Step SB5, the display device 13 sets “Unable to determine”. A corresponding message is displayed.

  When step SB6 ends, the second danger determination processing by the processing circuit 11 ends.

  According to the second risk determination process, when a result of “undetermined” is output in one decision tree model, the risk of a postoperative complication is determined using another decision tree model. As a result, the number of cases in which the result of “undetermined” is output can be reduced.

  The postoperative complication predicting method according to some of the above embodiments has at least a determining step and a displaying step. The determination step is based on background factors of a first sample in which a post-operative complication has occurred within a predetermined period and a second sample in which the post-operative complication has not occurred, and the post-operative complication is constructed. In accordance with a prediction model for predicting the risk of occurrence of postoperative, the risk of postoperative complications occurring in the subject is determined from background factors of the subject measured before surgery. The display step displays the result of the risk determination on a display device.

  According to the above method, the occurrence of postoperative complications can be predicted with high accuracy.

  The present invention is not limited to the above-described embodiment, and can be variously modified in an implementation stage without departing from the gist of the invention. In addition, the embodiments may be combined as appropriate, and in that case, the combined effect is obtained. Furthermore, the above-described embodiment includes various inventions, and various inventions can be extracted by combinations selected from a plurality of disclosed constituent features. For example, even if some components are deleted from all the components shown in the embodiment, if the problem can be solved and an effect can be obtained, a configuration from which the components are deleted can be extracted as an invention.

1 Postoperative Complication Prediction Device 11 Processing Circuit 12 Storage Device 13 Display Device 14 Input Device 15 Communication Device 111 Patient Factor Input Unit 112 Image Measurement Unit 113 Model Selection Unit 114 Risk Determination Unit 115 Display Control Unit 121 Prediction Model

Claims (11)

A computer configured based on background factors of the first sample having a post-operative complication within a predetermined period and the second sample having no post-operative complication within a predetermined period; According to a prediction model for predicting the risk of occurrence, a determination step of determining the risk of the post-operative complication occurring in the subject from the background factors of the subject measured before surgery,
A display step in which the computer displays the risk determination result on a display device;
A method for predicting postoperative complications, comprising: The prediction model is a decision tree model including a plurality of nodes provided in a tree shape,
Each of the plurality of nodes is defined by a determination condition regarding each of a plurality of variables included in the background factor,
The method for predicting postoperative complications according to claim 1. In the determining step, at a terminal node of the plurality of nodes, when a predetermined ratio of the first sample is true and a predetermined ratio of the second sample is false, the post-operative complication is determined. When it is determined that the risk of occurrence is high, and when the predetermined ratio of the first sample is false and the predetermined ratio of the second sample is true, the risk of developing the postoperative complication is low. Judge, otherwise, judge that it is impossible to judge,
The method for predicting postoperative complications according to claim 2. The prediction model includes a plurality of decision tree models in which the type of the variable, the threshold of the determination condition, and the order of the determination condition are different from each other,
The determining step includes determining, when using the first decision tree model among the plurality of decision tree models, that the determination result indicates that the determination is not possible, the other decision tree model among the plurality of decision tree models. Determining the risk using a tree model,
The method for predicting postoperative complications according to claim 3. The post-operative complication is a post-operative complication resulting from a surgical operation,
The background factors include at least one variable of factors measured by age, gender, BMI, BMD, and radiographic image,
The method for predicting postoperative complications according to claim 2. The post-operative complication is a complication resulting from surgery for adult spinal deformity,
The background factor includes at least one variable of age, gender, BMI, BMD, frailty, C7SVA, PI-LL, fixed vertebral body number, TPA, high UIV and high LIV.
The method for predicting postoperative complications according to claim 2.   The postoperative complication predicting method according to claim 1, wherein the determining step determines the risk from a background factor of the subject and a surgical invasive factor of the subject.   The method for predicting a postoperative complication according to claim 7, wherein the surgical invasive factor includes at least one variable of whether or not PSO is applied, operation time, and EBL. The prediction model has a plurality of prediction models according to human characteristics,
The determining step, from among the plurality of prediction models, selects a prediction model according to the human characteristics of the subject, and determines the risk using the selected prediction model,
The method for predicting postoperative complications according to claim 1. On the computer,
The risk of the occurrence of the postoperative complication, which is constructed based on the background factors of the first sample in which the postoperative complication has occurred within a predetermined period and the second sample in which the postoperative complication has not occurred. According to a prediction model for predicting sex, a function of determining the risk of the post-operative complication occurring in the subject from the background factors of the subject measured before surgery,
A function of displaying the judgment result of the risk on a display device,
Postoperative complication prediction program that realizes. The risk of the occurrence of the postoperative complication, which is constructed based on the background factors of the first sample in which the postoperative complication has occurred within a predetermined period and the second sample in which the postoperative complication has not occurred. A storage unit for storing a prediction model for predicting the gender,
According to the prediction model, a determination unit that determines the risk of the post-operative complication occurring in the subject from the background factors of the subject measured before surgery,
A display unit that displays the risk determination result on a display device,
A postoperative complication predicting device comprising: