import gradio as gr import pandas as pd import numpy as np # 导入pickle模块 import pickle import matplotlib.pyplot as plt # 导入C:\MyProject\MIMIC\dvt_diabetes\savemodel\c-curve\00calibrated.pickle with open(r'./model/00calibrated_clf.pickle', 'rb') as f: modelcalibration00 = pickle.load(f) with open(r'./model/28calibrated_clf.pickle', 'rb') as f: modelcalibration28 = pickle.load(f) with open(r'./model/60calibrated_clf.pickle', 'rb') as f: modelcalibration60 = pickle.load(f) with open(r'./model/90calibrated_clf.pickle', 'rb') as f: modelcalibration90 = pickle.load(f) def plot_probabilities(probabilities): plt.figure(figsize=(18, 7)) plt.xticks([0, 28, 60, 90]) plt.plot([0, 28, 60, 90], probabilities, marker='o') plt.xlabel('Days') plt.ylabel('Probability') plt.title('Probability Line Chart') plt.grid(True) plt.tight_layout() plt.show() return plt.gcf() def get_risk_category(calibrated_class): if calibrated_class == 1: return 'Risks that need attention' else: return 'Lower risk' def process_input(age_input, height_input, weight_input, lods_input, apache_input, cci_input, oasis_input, saps_input, sofa_input, alp_max, alp_min, alt_max, alt_min, ast_max, ast_min, bilirubin_max, bilirubin_min, bun_max, bun_min, creatinine_max, creatinine_min, glucose_mean, glucose_min, inr_max, inr_min, pt_max, pt_min, ptt_max, ptt_min, wbc_max, wbc_min, platelet_max, platelet_min): # 对lods_input执行标准化，均值为4.99065420560747，标准差为2.92882867084736 LODS = lods_input # 对age_input执行标准化，均值为68.6822429906542，标准差为12.561298298873 admission_age = age_input alp_max = alp_max alp_min = alp_min alt_max = alt_max alt_min = alt_min apsiii = apache_input ast_max = ast_max ast_min = ast_min bilirubin_total_max = bilirubin_max bilirubin_total_min = bilirubin_min bun_max = bun_max bun_min = bun_min charlson_comorbidity_index = cci_input creatinine_max = creatinine_max creatinine_min = creatinine_min glucose_mean = glucose_mean glucose_min = glucose_min height = height_input inr_max = inr_max inr_min = inr_min oasis = oasis_input platelets_max = platelet_max platelets_min = platelet_min pt_max = pt_max pt_min = pt_min ptt_max = ptt_max ptt_min = ptt_min sapsii = saps_input sofa_24hours = sofa_input wbc_max = wbc_max wbc_min = wbc_min weight = weight_input data_28 = np.array([ [charlson_comorbidity_index, oasis, apsiii, weight, sapsii, ast_max, ast_min, alt_min, admission_age, sofa_24hours, pt_min, LODS, alt_max, wbc_min, bilirubin_total_max, height, wbc_max, inr_min, ptt_min, pt_max, inr_max, creatinine_max] ]) data_60 = np.array([ [charlson_comorbidity_index, sofa_24hours, wbc_min, apsiii, sapsii, admission_age, wbc_max, alp_min, weight, height, ast_max, oasis, ptt_max, alt_max, LODS, alt_min, ptt_min, alp_max, pt_min, platelets_max, glucose_min, bilirubin_total_max, bun_max, bun_min, bilirubin_total_min, pt_max] ]) data_90 = np.array([ [charlson_comorbidity_index, apsiii, admission_age, sapsii, LODS, sofa_24hours, oasis, wbc_max, height, wbc_min, weight, bun_min, creatinine_max, ptt_min, ptt_max, creatinine_min, glucose_min, glucose_mean, bilirubin_total_max, alt_min, platelets_min, pt_min, alp_min, bun_max, inr_min, alp_max, platelets_max] ]) data_00 = np.array([ [oasis, charlson_comorbidity_index, LODS, sofa_24hours, wbc_min, apsiii, sapsii, weight, platelets_min, wbc_max, admission_age, pt_min, alp_min, bun_min, ptt_min, alp_max, inr_min, ast_min] ]) # 将数据直接输入概率校准模型 calibrated_prob_28 = modelcalibration28.predict_proba(data_28)[:, 1] calibrated_class_28 = modelcalibration28.predict(data_28)[0] calibrated_prob_60 = modelcalibration60.predict_proba(data_60)[:, 1] calibrated_class_60 = modelcalibration60.predict(data_60)[0] calibrated_prob_90 = modelcalibration90.predict_proba(data_90)[:, 1] calibrated_class_90 = modelcalibration90.predict(data_90)[0] calibrated_prob_00 = modelcalibration00.predict_proba(data_00)[:, 1] calibrated_class_00 = modelcalibration00.predict(data_00)[0] probabilities = [calibrated_prob_00, calibrated_prob_28, calibrated_prob_60, calibrated_prob_90] output_plot = plot_probabilities(probabilities) risk_category_00 = get_risk_category(calibrated_class_00) risk_category_28 = get_risk_category(calibrated_class_28) risk_category_60 = get_risk_category(calibrated_class_60) risk_category_90 = get_risk_category(calibrated_class_90) # 返回校准后的概率和类别 return risk_category_00, risk_category_28, risk_category_60, risk_category_90, output_plot with gr.Blocks() as demo: with gr.Row(): gr.Markdown('# Prediction of Death Events in ICU Patients with DVT &Diabetes') with gr.Row(): with gr.Column(): gr.Markdown('## Basic Information') # 输入控件 age_input = gr.Number(label="admission_age", step=1) height_input = gr.Number(label="Height (cm)", step=0.1) # 身高输入框，单位为厘米 weight_input = gr.Number(label="Weight (kg)", step=0.1) # 体重输入框，单位为千克 # 临床评分部分 gr.Markdown("## Clinical Scoring Section") lods_input = gr.Number(label="LODS (Logistic Organ Dysfunction Score)", step=0.1) apache_input = gr.Number(label="APACHE III (Acute Physiology and Chronic Health Evaluation III)", step=0.1) cci_input = gr.Number(label="CCI (Charlson Comorbidity Index)", step=0.1) oasis_input = gr.Number(label="OASIS (Oxford Acute Severity of Illness Score)", step=0.1) saps_input = gr.Number(label="SAPS II (Simplified Acute Physiology Score II)", step=0.1) sofa_input = gr.Number(label="SOFA (Sequential Organ Failure Assessment, 24hr average)", step=0.1) gr.Markdown("## Laboratory indicators on the first day") # 血液学检验部分 gr.Markdown("### Hematologic Tests") # 白细胞计数 gr.Markdown("#### White Blood Cell Count (WBC) (10^9/L)") wbc_max = gr.Number(label="Max") wbc_min = gr.Number(label="Min") # 血小板计数 gr.Markdown("#### Platelet Count (10^9/L)") platelet_max = gr.Number(label="Max") platelet_min = gr.Number(label="Min") with gr.Column(): # 肝功能测试部分 gr.Markdown("### Liver Function Tests") # 碱性磷酸酶 gr.Markdown("#### Alkaline Phosphatase (ALP) (U/L)") alp_max = gr.Number(label="Max") alp_min = gr.Number(label="Min") # 丙氨酸氨基转移酶 gr.Markdown("#### Alanine Aminotransferase (ALT) (U/L)") alt_max = gr.Number(label="Max") alt_min = gr.Number(label="Min") # 阿斯巴甜氨基转移酶 gr.Markdown("#### Aspartate Aminotransferase (AST) (U/L)") ast_max = gr.Number(label="Max") ast_min = gr.Number(label="Min") # 总胆红素 gr.Markdown("#### Total Bilirubin (mg/dL)") bilirubin_max = gr.Number(label="Max") bilirubin_min = gr.Number(label="Min") # 肾功能测试部分 gr.Markdown("### Renal Function Tests") # 尿素氮 gr.Markdown("#### Blood Urea Nitrogen (BUN) (mg/dL)") bun_max = gr.Number(label="Max") bun_min = gr.Number(label="Min") with gr.Column(): # 肌酐 gr.Markdown("#### Creatinine (mg/dL)") creatinine_max = gr.Number(label="Max") creatinine_min = gr.Number(label="Min") # 血糖水平 gr.Markdown("### Glucose Levels") # 血糖 gr.Markdown("#### Glucose (mg/dL)") glucose_mean = gr.Number(label="Mean") glucose_min = gr.Number(label="Min") # 凝血测试 gr.Markdown("### Coagulation Tests") # 国际标准化比率 gr.Markdown("#### International Normalized Ratio (INR) (ratio)") inr_max = gr.Number(label="Max") inr_min = gr.Number(label="Min") # 凝血酶原时间 gr.Markdown("#### Prothrombin Time (PT) (seconds)") pt_max = gr.Number(label="Max") pt_min = gr.Number(label="Min") # 部分凝血活酶时间 gr.Markdown("#### Partial Thromboplastin Time (PTT) (seconds)") ptt_max = gr.Number(label="Max") ptt_min = gr.Number(label="Min") with gr.Row(): submit_button = gr.Button("Submit") with gr.Row(): gr.Markdown("## Risk Prediction Results") with gr.Row(): risk_category_00 = gr.Label(label="00-day Prediction Category") #output_prob_28 = gr.Textbox(label="28-day Calibrated Probability", interactive=False) risk_category_28 = gr.Label(label="28-day Prediction Category") #output_prob_60 = gr.Textbox(label="60-day Calibrated Probability", interactive=False) risk_category_60 = gr.Label(label="60-day Prediction Category") #output_prob_90 = gr.Textbox(label="90-day Calibrated Probability", interactive=False) risk_category_90 = gr.Label(label="90-day Prediction Category") #output_prob_00 = gr.Textbox(label="00-day Calibrated Probability", interactive=False) with gr.Row(): output_plot = gr.Plot(label="Probability Line Chart") submit_button.click( process_input, inputs=[ age_input, height_input, weight_input, lods_input, apache_input, cci_input, oasis_input, saps_input, sofa_input, wbc_max, wbc_min, platelet_max, platelet_min, alp_max, alp_min, alt_max, alt_min, ast_max, ast_min, bilirubin_max, bilirubin_min, bun_max, bun_min, creatinine_max, creatinine_min, glucose_mean, glucose_min, inr_max, inr_min, pt_max, pt_min, ptt_max, ptt_min ], outputs=[ risk_category_00, risk_category_28, risk_category_60, risk_category_90, output_plot ] ) demo.launch(share=True)