Machine learning prediction models for in-hospital postoperative functional outcome after moderate-to-severe traumatic brain injury

AIM

This study aims to utilize machine learning (ML) and logistic regression (LR) models to predict surgical outcomes among patients with traumatic brain injury (TBI) based on admission examination, assisting in making optimal surgical treatment decision for these patients.

METHOD

We conducted a retrospective review of patients hospitalized in our department for moderate-to-severe TBI. Patients admitted between October 2011 and October 2022 were assigned to the training set, while patients admitted between November 2022 and May 2023 were designated as the external validation set. Five ML algorithms and LR model were employed to predict the postoperative Glasgow Outcome Scale (GOS) status at discharge using clinical and routine blood data collected upon admission. The Shapley (SHAP) plot was utilized for interpreting the models.

RESULTS

A total of 416 patients were included in this study, and they were divided into the training set (n = 396) and the external validation set (n = 47). The ML models, using both clinical and routine blood data, were able to predict postoperative GOS outcomes with area under the curve (AUC) values ranging from 0.860 to 0.900 during the internal cross-validation and from 0.801 to 0.890 during the external validation. In contrast, the LR model had the lowest AUC values during the internal and external validation (0.844 and 0.567, respectively). When blood data was not available, the ML models achieved AUCs of 0.849 to 0.870 during the internal cross-validation and 0.714 to 0.861 during the external validation. Similarly, the LR model had the lowest AUC values (0.821 and 0.638, respectively). Through repeated cross-validation analysis, we found that routine blood data had a significant association with higher mean AUC values in all ML and LR models. The SHAP plot was used to visualize the contributions of all predictors and highlighted the significance of blood data in the lightGBM model.

CONCLUSION

The study concluded that ML models could provide rapid and accurate predictions for postoperative GOS outcomes at discharge following moderate-to-severe TBI. The study also highlighted the crucial role of routine blood tests in improving such predictions, and may contribute to the optimization of surgical treatment decision-making for patients with TBI.

Machine learning models for predicting in-hospital outcomes after non-surgical treatment among patients with moderate-to-severe traumatic brain injury

AIM

This study aims to develop prediction models for in-hospital outcomes after non-surgical treatment among patients with moderate-to-severe traumatic brain injury (TBI).

METHOD

We conducted a retrospective review of patients hospitalized for moderate-to-severe TBI in our department from 2011 to 2020. Five machine learning (ML) algorithms and the conventional logistic regression (LR) model were employed to predict in-hospital mortality and the Glasgow Outcome Scale (GOS) functional outcomes. These models utilized clinical and routine blood data collected upon admission.

RESULTS

This study included a total of 196 patients who received only non-surgical treatment after moderate-to-severe TBI. When predicting mortality, ML models achieved area under the curve (AUC) values of 0.921 to 0.994 using clinical and routine blood data, and 0.877 to 0.982 using only clinical data. In comparison, LR models yielded AUCs of 0.762 and 0.730 respectively. When predicting the GOS outcome, ML models achieved AUCs of 0.870 to 0.915 using clinical and routine blood data, and 0.858 to 0.927 using only clinical data. In comparison, the LR model yielded AUCs of 0.798 and 0.787 respectively. Repeated internal validation showed that the contributions of routine blood data for prediction models may depend on different prediction algorithms and different outcome measurements.

CONCLUSION

The study reported ML-based prediction models that provided rapid and accurate predictions on short-term outcomes after non-surgical treatment among patients with moderate-to-severe TBI. The study also highlighted the superiority of ML models over conventional LR models and proposed the complex contributions of routine blood data in such predictions.

The potential of high-order features of routine blood test in predicting the prognosis of non-small cell lung cancer

BACKGROUND

Numerous studies have demonstrated that the high-order features (HOFs) of blood test data can be used to predict the prognosis of patients with different types of cancer. Although the majority of blood HOFs can be divided into inflammatory or nutritional markers, there are still numerous that have not been classified correctly, with the same feature being named differently. It is an urgent need to reclassify the blood HOFs and comprehensively assess their potential for cancer prognosis.

METHODS

Initially, a review of existing literature was conducted to identify the high-order features (HOFs) and classify them based on their calculation method. Subsequently, a cohort of patients diagnosed with non-small cell lung cancer (NSCLC) was established, and their clinical information prior to treatment was collected, including low-order features (LOFs) obtained from routine blood tests. The HOFs were then computed and their associations with clinical features were examined. Using the LOF and HOF data sets, a deep learning algorithm called DeepSurv was utilized to predict the prognostic risk values. The effectiveness of each data set's prediction was evaluated using the decision curve analysis (DCA). Finally, a prognostic model in the form of a nomogram was developed, and its accuracy was assessed using the calibration curve.

RESULTS

From 1210 documents, over 160 blood HOFs were obtained, arranged into 110, and divided into three distinct categories: 76 proportional features, 6 composition features, and 28 scoring features. Correlation analysis did not reveal a strong association between blood features and clinical features; however, the risk value predicted by the DeepSurv LOF- and HOF-models is significantly linked to the stage. Results from DCA showed that the HOF model was superior to the LOF model in terms of prediction, and that the risk value predicted by the blood data model could be employed as a complementary factor to enhance the prognosis of patients. A nomograph was created with a C-index value of 0.74, which is capable of providing a reasonably accurate prediction of 1-year and 3-year overall survival for patients.

CONCLUSIONS

This research initially explored the categorization and nomenclature of blood HOF, and proved its potential in lung cancer prognosis.

Effects of storage duration of suspended red blood cells before intraoperative infusion on coagulation indexes, routine blood examination and immune function in patients with gastrointestinal tumors

Objective

To investigate the effect of storage duration of suspended red blood cells (SRBC) before intraoperative infusion on coagulation indexes, routine blood examination and immune function in patients with gastrointestinal (GI) tumors.

Methods

We divided clinical data of one hundred patients with GI tumors who underwent surgical treatment in our hospital into two different groups according to the storage duration of SRBC use for intraoperative infusion. The short-term group (n=50) had patients with SRBC storage durations shorter than two weeks, and the long-term group (n=50) had patients with storage durations longer than two weeks. We compared the coagulation, immune function, routine blood profile, electrolyte levels and adverse reactions assessment results between the two groups.

Results

Compared with before transfusions, the levels of fibrinogen (FIB) and activated partial prothrombin time (APTT) after blood transfusions were higher than those before transfusion (P<0.05). The levels of hemoglobin (Hb) and hematocrit (HCT) in the two groups after blood transfusions were also higher than those before transfusion (P<0.05). However, the levels of CD4+ decreased and those of CD8+ increased in both groups after the blood transfusions. In addition, the levels of CD4+ and CD4+/CD8+ in the short-term group were higher than those of the long-term group (P<0.05) while the CD8+ levels were lower than that of the long-term group (P<0.05). After the blood transfusions, the potassium ion (K+) levels in the two groups increased, and those in the long-term group were higher than in the short-term group (P<0.05). The sodium ion (Na+) levels in the two groups increased after the transfusions, and the short-term group had higher levels than the long-term group (P<0.05). Finally, the incidence of adverse reactions in the short-term group (4.00%) was lower than that in the long-term group (18.00%) (P<0.05).

Conclusion

Intraoperative infusion of SRBC with storage duration longer than two weeks increases the risk of perioperative adverse transfusion reactions, which implies that the storage duration of SRBC should be strictly controlled in clinical practice to reduce the risk of blood transfusion.

The differences of hemogram, myelogram, and driver gene mutations in classic myeloproliferative neoplasms

The aim of this study was to explore and compare routine blood features and pathological characteristics of bone marrow tissues in essential thrombocythemia (ET), polycythemia vera (PV), primary myelofibrosis, prefibrotic stage (prePMF) and overt fibrotic stage (overtPMF), and the correlation between common driver gene mutations and clinical manifestations of myeloproliferative neoplasms (MPN). Methods: We analyzed 259 MPN patients treated at Tongji Hospital of Huazhong University of Science and Technology from January 2016 to December 2020. Results: Among ET, PV, prePMF, and overtPMF, the median leukocyte counts of PV and prePMF were significantly higher than those of ET. The average hemoglobin level of overtPMF was significantly lower than that of ET, PV, and prePMF. ET and prePMF had higher platelet counts than PV and overtPMF, whereas ET had the lowest platelet distribution width. Regarding hematopoietic tissues in the bone marrow, enlarged megakaryocytes were easily found in ET, PV, and prePMF, whereas the average diameter of megakaryocytes in prePMF was smaller than in ET, and PV showed various sizes of megakaryocytes. An increased M/E ratio and dilation of sinus were seen more frequently in PMF. Additionally, JAK2-positive patients tended to have significantly higher leukocyte counts than CALR-positive patients in ET and PMF.

[Correlation analysis between blood routine-derived inflammatory markers and respiratory function in pneumoconiosis patients]

Objective: To analyze the correlation between blood routine-derived inflammation indicators and respiratory function in patients with pneumoconiosis. Methods: In January 2021, 492 male pneumoconiosis patients hospitalized in Hefei Institute of Occupational Disease Control and Prevention from 2012 to 2020 were randomly selected as the case group, 492 dust exposed non pneumoconiosis workers who underwent occupational health examination at the same time were taken as the control group. The occupational history and clinical examination data of the two groups of subjects were collected, the correlation between blood routine-derived inflammatory indexes and pulmonary function and blood gas analysis was analyzed retrospectively. Results: Compared with the control group, the lymphocyte monocyte ratio (LMR) in the case group was decreased, and the neutrophil lymphocyte ratio (NLR) was increased, and the difference was statistically significant (P<0.05) . There were significant differences in forced vital capacity as a percentage of the predicted value (FVC) , forced expiratory volume in the first second as a percentage of the predicted value (FEV(1)%) , one second rate (FEV(1)/FVC) , partial pressure of oxygen (PaO(2)) , partial pressure of carbon dioxide (PaCO(2)) , and pH among pneumoconiosis patients at different stages (P<0.05) . FVC%, FEV(1)%, FEV(1)/FVC, and PaO(2) decreased with the increase of the stage, the trend test was statistically significant (tau-b=-0.24, -0.34, -0.37, -0.17, P<0.05) , PaCO(2) and pH increased with the increase of the stage, and the trend test was statistically significant (tau-b=0.10, 0.08, P<0.05) . There were statistically significant differences in LYM, LMR, NLR, platelet lymphocyte ratio (PLR) in patients with pneumoconiosis at different stages (P<0.05) , and LYM and LMR decreased with the increase of stage, trend test showed that there was statistically significant (tau-b=-0.11, -0.13, P<0.05) . There were significant differences in FVC%, FEV(1)%, FEV(1)/FVC, PaO(2), pH, LMR, NLR, PLR among patients with different types of pneumoconiosis (P<0.05) . LMR in pneumoconiosis patients was significantly positively correlated with FVC%, FEV(1)%, FEV(1)/FVC and PaO(2) (r(s)=0.342, 0.324, 0.203, 0.207, P<0.05) , NLR was significantly negatively correlated with FVC%, FEV(1)%, FEV(1)/FVC and PaO(2) (r(s)=-0.193, -0.202, -0.164, -0.177, P<0.05) , PLR was significantly negatively correlated with FVC%, FEV(1)%, FEV(1)/FVC and PaO(2) (r(s)=-0.194, -0.193, -0.106, -0.113, P<0.05) . Multiple linear regression analysis showed that LMR in pneumoconiosis patients was positively related with FVC%, FEV(1)% and PaO(2) (P<0.05) . Conclusion: LMR in patients with pneumoconiosis has a certain correlation with lung function and blood gas analysis, LMR is expected to become a sensitive indicator for evaluating pneumoconiosis.

[Dynamics of routine blood tests in BALB/c mice with Babesia microti infection]

OBJECTIVE

To understand the changes in body weight, spleen weight and complete blood cells in BALB/c mice infected with Babesia microti.

METHODS

For the infection group, six weeks old BALB/c mice were injected intraperitoneally with a dose of 100 μL of B. microti infected blood (20% RBC infection rate, each mouse). For the determination of the progression of B. microti infection up to 28 days of the infection, the microscopic visualization of thin blood smears of tail blood stained with Giemsa staining was performed in the infection group. The experiment was carried out at different intervals on days 0, 7, 14, 21, and 28 after the infection, respectively. The mice were sacrificed, and spleens were collected and weighed, and the body weight of the mice was also determined. The blood cells of the mice were analyzed by using Mindray BC-5300 Vet animal automatic hematology analyzer.

RESULTS

On the first day after the infection, B. microti was visualized in RBC of the infection group. The significantly highest infection rate (55%) appeared on the seventh day of the infection, and then steadily decreased; the mice attained the latent infection phase on the 28th day post-infection, when the parasite could not be visualized in the peripheral blood. The mice in the infected group acquired a significantly lowest body weight on the 7th day of the infection, and then gradually returned to normal. The weight of the spleen was the significantly highest on the 14th day of the infection, and then consistently decreased. On the 28th day of infection, the spleen weight was still higher than that of the control group. There were no significant changes in the number of white blood cells (WBC), lymphocytes, and eosinophils in the infected mice; and altered levels were all within the normal mouse reference range. The number of red blood cells, hemoglobin, and platelet count in the infected mice were decreased to the lowest level when the B. microti infection rate achieved to the highest, and then gradually returned to the normal levels.

CONCLUSIONS

B. microti infection can cause body weight loss, splenic weight gain, and reduction in the number of erythrocytes and platelets in whole blood of the mice. Besides, the whole blood cell analyzer has a diagnostic significance in the identification of babesiosis.

Value of routine blood test in early diagnosis of gastric cancer and its relationship with Helicobacter pylori infection

AIM

To assess the value of each index of routine blood test in the early diagnosis of gastric cancer (GC), and investigate the effect of Helicobacter pylori (H. pylori) infection on routine blood indexes.

METHODS

A total of 223 inpatients or hospitalized patients who underwent gastroscopy from September 2016 to September 2017 were selected and divided into four groups according to gastroscopic and pathological results, including 58 patients with chronic superficial gastritis (CSG), 59 patients with chronic atrophic gastritis (CAG), 52 patients with high-grade intraepithelial neoplasia (HIN), and 57 patients with GC. Red blood cell distribution width (RDW), platelet count (PLT), mean platelet volume (MPV), platelet distribution width (PDW), white blood cell count (WBC), lymphocyte count (LYMP), neutrophil count (NR), neutrophil count to lymphocyte ratio (NLR), and platelet count to lymphocyte ratio (PLR) were compared in different groups, and the sensitivity and specificity of each index for diagnosis of GC were calculated. The effect of H. pylori infection on routine blood indexes was also analyzed.

RESULTS

LYMP and MPV gradually decreased and RDW, NLR, and PLR gradually increased from CSG to CAG, HIN, and GC. RDW, MPV, LYMP, NLR, and PLR were all significantly different between the CSG group and GC group (P < 0.05). In detecting GC, the sensitivities of RDW, MPV, LYMP, NLR, PLR, and NC were 82.5%, 80.7%, 84.2%, 84.2%, 71.9%, and 45.6%, respectively, and the specificities were 72.4%, 55.2%, 82.8%, 91.0%, 89.7%, and 89.7%, respectively. The positive rates of H. pylori in the CSG group, CAG group, HIN group, and GC group were 63.7%, 62.5%, 63.5%, and 66.7%, respectively. In the CSG group, PLT, WBC, and NR were significantly higher in H. pylori positive patients than in H. pylori negative patients (P < 0.05). There were no significant differences in blood routine indexes between H. pylori positive and H. pylori negative patients in the CAG, HIN, and GC groups (P > 0.05).

CONCLUSION

RDW, PLT, MPV, LYMP, NLR, and PLR can be used in the early diagnosis of GC. H. pylori infection can lead to systemic inflammatory reactions. Early eradication of H. pylori is beneficial to the prevention of GC.

Comparison of routine blood parameters and blood biochemical parameters between patients with alcoholic and non-alcoholic fatty liver disease

AIM: To compare routine blood parameters and blood biochemical parameters between patients with alcoholic fatty liver disease (AFLD) and non-alcoholic fatty liver disease (NAFLD).

METHODS: A total of 1 203 subjects, who underwent physical examination, including measurement of abdominal circumference and blood pressure, routine blood examination, biochemical examination and two-dimensional ultrasound, were included in this study. Of these subjects, 509 (415 men and 94 women) were diagnosed with fatty liver disease (FLD) and 694 (446 men and 248 women) without FLD by two-dimensional ultrasound. According to the Guidelines for the Diagnosis and Treatment of Alcoholic Liver Disease, subjects with FLD were divided into two groups: AFLD group (n = 106, 105 men and 1 woman) and NAFLD group (n = 403, 310 men and 93 women).

RESULTS: Subjects with FLD had a younger mean age and a higher BMI than subjects without FLD (P = 0.011 and 0.000, respectively). WBC, RBC, HGB, ALT, AST, GGT, ALB, TG, TC, LDL-C, FBG, UA (P < 0.001 for all), ALP (P = 0.015), TP (P = 0.026), AST/PLT ratio (P = 0.001), AST/ALT ratio (P < 0.001), and GGT/PLT ratio (P < 0.001) were higher and HDL-C, AST/GGT ratio, and TBIL/GGT ratio (P < 0.001 for all) were lower in subjects with FLD than in those without FLD. Subjects with AFLD had a younger mean age (P < 0.001), a higher HGB, GGT, ALB, UA (P < 0.001 for all), TG, AST, RBC, ALT (P = 0.007, 0.005, 0.011 and 0.027, respectively), AST/PLT ratio, AST/ALT ratio, and GGT/PLT ratio (P = 0.012, 0.014 and 0.002, respectively), and a lower UREA, AST/GGT ratio and TBIL/GGT ratio (P < 0.001 for all) than those with NAFLD.

CONCLUSION: Routine blood examination and blood biochemical examination are useful in distinguishing patients with and without FLD as well as those with AFLD and NAFLD.