Establishment And Application Of Abdominal Wall Tension Assessment Method Based On Tissue Mechanical Properties

Background:The abdominal wall comprises soft tissues,such as muscles and fascia,divided into the anterior ventral lateral wall in the front and the posterior ventral wall in the rear by the posterior axillary line on both sides.There are three main functions for the ventral lateral:1)maintain the integrity of the abdominal cavity,protect abdominal organs,and maintain posture and physique;2)increase and maintain intra-abdominal pressure（IAP）,such as sneezing,coughing,vomiting,or defecating;and 3)aid breathing and trunk movements include hard breathing,flexion,extension,lateral flexion,and rotation.These physiological or pathological changes in the abdominal wall are primarily related to the structure’s mechanical characteristics,called abdominal wall tension（AWT）.When the abdominal wall is pressed,there is tenderness,rebound pain,and increased abdominal muscle tension（peritoneal irritation syndrome）.When inflammation of abdominal organs does not affect the parietal peritoneum,there is only tenderness.If the parietal peritoneum is involved,it can cause rebound pain.However,abdominal muscle tension is the slow and persistent contraction of muscles of the abdominal wall after stimulation of the peritoneum.If the examiner feels that the entire or part of the patient’s abdominal wall is tense or has a sense of resistance,and its resistance exceeds the tension of the normal abdominal wall,it is called increased AWT.The degree and extent of AWT are the basis for diagnosis and even peritonitis operation.Patients who contract abdominal muscles nervously or unconsciously may have false positives.At the same time,old age,weakness,coma,high spinal cord injury,poisoning,and sedative or analgesic drugs can cause false negatives.Clinically,especially in trauma patients,there is always uncertainty about signs of peritoneal irritation.The subjectivity of signs of peritoneal irritation is one of the important reasons why the abdomen remains the "last black box" for diagnosing severe trauma,even in highly developed imaging today.At present,a preliminary assessment of AWT is usually performed during a physical examination.Accurate AWT measurement methods include the evaluation of stress characteristics of isolated human abdominal wall muscles,shear wave abdominal wall ultrasound,the tensiometer,the ultrasonic assessment of isolated shear wave abdominal wall,and the three-dimensional simulation model of abdominal wall elasticity.The previous research methods are mainly the measurement research of AWT of isolated specimens and/or open abdominal cavities in the abdominal surgery specialty.There are few studies on AWT measurement in an intact abdominal wall and high abdominal pressure.Meanwhile,previous AWT evaluation methods are unreliable or inaccurate due to a lack of validation and cannot be used in clinics.Diagnostic experiments must compare or combine the measurement results with more accurate or recognized techniques to judge their diagnostic efficacy.The AWT evaluation method based on tissue mechanical properties refers to the AWT measurement method of an intact abdominal wall under stress.The tensiometer is a non-invasive IAP evaluation technique.The principle is to measure the degree of depression applied to the hard part during abdominal palpation,which can detect the passive tension of the abdominal wall.If the tension in the abdominal wall increases,it indicates peritonitis.Although,from the perspective of technology,the AWT evaluation method of a tensiometer has shown reliable results,the relationship between the AWT evaluation method and the IAP standard measurement method is still unclear.Millimeter waves（mm-waves）are electromagnetic waves with a frequency of 30-300GHz,which can penetrate clothes and other organic materials.They have been used in airport security inspections to detect dangerous goods,such as weapons.Millimeter wave radar is widely used in vital signs detection,gesture recognition,sleep monitoring,and driverless cars because of its excellent performance in detecting tiny vibrations.Among all millimeter-wave radar systems,the frequency-modulated continuous-wave（FMCW）millimeter-wave radar system is superior to others in detecting small displacement and vital signals for longer distances and at faster speeds.In addition,it can also extract and distinguish characteristic information from multiple targets so that a single radar system can measure the life signals of various targets at the same time.In this study,the project team’s invention patent,"a non-invasive multi-point abdominal wall tension measuring device"（ZL 201510799207.4）,and the FMCW radar were applied to the experimental study of severe patients and IAH pig models to accurately and quantitatively evaluate AWT.The correlation between IAP and intra-abdominal infection（IAI）was analyzed using statistical and artificial intelligence data processing methods.A mathematical model was established to provide a reference for the early diagnosis of IAH and IAI in critically ill patients.The research includes two experiments.Experiment 1:A new abdominal wall tension measuring device and its clinical research on the screening value of abdominal infection;Experiment 2:Experimental study on the correlation between the characteristics of the abdominal wall tension signal extracted by the FMCW radar and the intra-abdominal pressure grading.Methods:1.A new AWT measuring device and its screening value for abdominal infectionCritically ill patients were included in the Department of Intensive Care Unit（ICU）,Daping Hospital,Chinese Army Medical University,from August 30,2018,to June 30,2020.Inclusion criteria were patients who 1)were aged≥18 years,2)were hospitalized for>7 days,3)had high-risk factors of abdominal hypertension,4)had a urinary catheter,5)were without contraindication to IAP measurement through the bladder,and 6)signed the informed consent form.Patients were divided into intra-abdominal hypertension（IAH）group（IAP≥12mmHg）,non-IAH group（IAP<12mmHg）,IAI group（abdominal infection,diffuse peritonitis,localized peritonitis,retroperitoneal abscess,abdominal abscess）,and non-IAI group（no infection,as mentioned）.Basic data,clinical data,and prognosis were collected.The basic data included the gender,age,body mass index（BMI）of the patients,and reasons for admission to the ICU.Clinical data included the Injury Severity Score（ISS）,the Acute Physiology and Chronic Health Evaluation（APACHE）Ⅱ score on the first day in the ICU,the Sequential Organ Failure Assessment（SOFA）score,procalcitonin（PCT）,lactic acid（LAC）,C-reactive protein（CRP）,high-risk factors for IAH（increased abdominal wall tension,intestinal contents,abdominal contents,capillary leakage,or fluid resuscitation）.The prognosis of the patients included the date admitted to the ICU,the length of stay,and ICU mortality.On the first day after admission to the ICU,intra-bladder pressure（IVP）and AWT were measured at 9 points of the abdominal wall.The correlation between AWT and I VP was analyzed to explore the role of AWT in the diagnosis of IAI.2.Experimental study on the correlation between the characteristics of the abdominal wall tension signal extracted by FMCW millimeter-wave radar and the intra-abdominal pressure gradingThe project team proposed a non-contact and non-invasive measurement method of intraabdominal pressure.The relationship between AWT data measured by FMCW millimeter-wave radar and IAP was analyzed by deep learning,and IAP was associated with abdominal wall mobility.The Pearson-coefficient-guided Domain Adversarial Neural Network（PCG-DANN）was proposed to learn the mapping relationship between them.A discrete wavelet transform（DWT）was added to the feature extractor,which could explicitly extract random migration features.To verify the effectiveness of the proposed method,four pigs were used to construct IAP of different grades by intraperitoneal injection,and IAP was measured using the bladder method.The FMCW radar system was directed to the abdominal wall of animals approximately 1m away to obtain the displacement signal of the abdominal wall.The deep learning method was adopted to deal with the relationship between different IAP data and abdominal wall mobility.The AWT artificial intelligence model was established to predict IAP.Through the verification of a new test data set collected by a pig not included in the training set,the proposed PCG-DANN was compared with the Convolutional Recurrent Neural Network（CRNN）,the Short-Time Fourier Transformation Convolutional Neural Networks（STFT-CNN）,the Domain Adversarial Neural Network（DANN）with Squeeze-and-Excitation（SE）block,the DANN,and the PCG-DANN without Discrete Wavelet Transformation（DWT）layer.The performance of this artificial intelligence method was verified.Results:1.Establish AWT and IAP polynomial fitting regression modelA total of 127 patients were enrolled:68 men（53.54%）,34 medical patients（26.77%）,and 93 surgical patients（73.23%）.The average age of patients was 64.10±11.63 years,with a BMI of 24.93±2.53.The AWT and IVP of the patients on the first day of admission were 2.77±0.38 N/mm and 12.31±7.01 mmHg,respectively.There was a positive correlation between AWT and IVP（correlation coefficient R=0.706,P<0.05）.The PCT,AWT and IVP of IAH group were 6.82±3.10 ug/L,3.00±0.32 N/mm and 19.06±2.43 mmHg,respectively,which were higher than those of non-IAH group（4.65±2.16 ug/L,2.44±0.22 N/mm and 6.26±3.16 mmHg）,P<0.05 was statistically significant.The ICU length of stay,total length of stay and ICU mortality of IAH group were 8.52±3.84 days,17.25±5.32 days and 31.67%（19/60）,respectively,which were higher than those of non-IAH group（7.06±2.97 days,14.18±3.99 days and 14.93%（10/67）,P<0.05）.The fitting model of multiple equation regression quadratic functions is AWT=-1.616×10^-3IVP²+8.323×10^-2IVP²+2.094.There was a correlation between high AWT and IAI.The cutoff value of AWT for the sensitivity and specificity of diagnosing IAI was 2.57 N/mm.The area under ROC curve of AWT in diagnosing IAI was 0.677,followed by AWT+IVP 0.659,IVP 0.549,and lactic acid 0.490.AWT had the best diagnostic efficiency,superior to IAP and lactic acid.2.A non-contact and non-invasive IAP measurement method based on FMCW radar is establishedA total of 158 groups of.bin data obtained using millimeter wave radar were segmented to simulate clinical environment and avoid spectral leakage.The collected time series were split into 8 segments using a Hamming window with time span of 25 seconds and moving peed of 1 movement in every 5 seconds to expand the training dataset and improve the performance of the model.A total of 1264（158*8）data files were obtained.To train and test deep learning,the data collected from 1 pig were used as test set to verify the invariant feature in learning domain in this model,and the data from the other 4 pigs were randomly divided into training set and verification set in a ratio of 9:1.The data size was 64 for test set,1080 for training set,and 120 for the verification set.To establish the correlation of IAP with abdominal wall mobility measurement,the mapping relation of the two was determined using PCG-DANN.The addition of DWT to the feature extractor enhanced neural network’s capture on the features of the whole time series,demonstrating the success of our FMCW millimeter wave radar-based artificial intelligence model for capturing the AWT signal feature and correlating AWT signal feature with IAP grading.When comparing the predicated IAP value obtained using the PCG-DANN neural network model with those obtained using other similar neural network structures,the mean absolute error（MAE）of the test set using PCG-DANN model was the lowest at 0.64,as compared of 1.6 for CRNN,5.75 for STFT-CNN,1.46 for SE-DANN,1.45 for DANN,and 1.07 for PCG-DANN in the absence of DWT layer.The variance in output uncertainty at each level of IAP in PCG-DANN model was nonsignificant.The results demonstrate that our artificial intelligence model has better performance than other available approaches and it offers new option for evaluating nonlinear IAP indicators.Conclusion:1.AWT was successfully measured for the first time with the non-invasive abdominal wall tension tensiometer developed by our project team.It was found that there was a nonlinear positive correlation between AWT and IAP in severe patients,and the regression model of polynomial fitting of AWT-IAP was constructed,which further verified the potential value of AWT in diagnosing abdominal infection.2.The project team proposed a new non-invasive and non-contact IAP measurement method based on the FMCW radar system.This method can measure IAP from a long distance,which is convenient for continuous monitoring.The PCG-DANN neural network model with DWT is innovatively proposed and used to fit the relationship between the AWT data and different IAP values.PCG-DANN reduces the differences between individuals,so it is easier to extract domain-invariant features,thus establishing the relationship between AWT and IAP.Our study provides a new strategy for clinical non-contact,non-invasive,and continuous monitoring of IAP.