Israel Medical Association Journal

Over the last decade the use of artificial intelligence (AI) has reformed academic research. While clinical diagnosis of psoriasis and psoriatic arthritis is largely straightforward, the determining factors of a clinical response to therapy, and specifically to biologic agents, have not yet been found. AI may meaningfully impact attempts to unravel the prognostic factors that affect response to therapy, assist experimental techniques being used to investigate immune cell populations, examine whether these populations are associated with treatment responses, and incorporate immunophenotype data in prediction models. The aim of this mini review was to present the current state of the AI-mediated attempts in the field. We executed a Medline search in October 2023. Selection and presentation of studies were conducted following the principles of a narrative–review design. We present data regarding the impact AI can have on the management of psoriatic disease by predicting responses utilizing clinical or biological parameters. We also reviewed the ways AI has been implemented to assist development of models that revolutionize the investigation of peripheral immune cell subsets that can be used as biomarkers of response to biologic treatment. Last, we discussed future perspectives and ethical considerations regarding the use of machine learning models in the management of immune-mediated diseases.

Background: Acute mesenteric ischemia (AMI) is a medical condition with high levels of morbidity and mortality. However, most patients suspected of AMI will eventually have a different diagnosis. Nevertheless, these patients have a high risk for co-morbidities.

Objectives: To analyze patients with suspected AMI with an alternative final diagnosis, and to evaluate a machine learning algorithm for prognosis prediction in this population.

Methods: In a retrospective search, we retrieved patient charts of those who underwent computed tomography angiography (CTA) for suspected AMI between January 2012 and December 2015. Non-AMI patients were defined as patients with negative CTA and a final clinical diagnosis other than AMI. Correlation of past medical history, laboratory values, and mortality rates were evaluated. We evaluated gradient boosting (XGBoost) model for mortality prediction.

Results: The non-AMI group comprised 325 patients. The two most common groups of diseases included gastrointestinal (33%) and biliary-pancreatic diseases (27%). Mortality rate was 24.6% for the entire cohort. Medical history of chronic kidney disease (CKD) had higher risk for mortality (odds ratio 2.2). Laboratory studies revealed that lactate dehydrogenase (LDH) had the highest diagnostic ability for predicting mortality in the entire cohort (AUC 0.70). The gradient boosting model showed an area under the curve of 0.82 for predicting mortality.

Conclusions: Patients with suspected AMI with an alternative final diagnosis showed a 25% mortality rate. A past medical history of CKD and elevated LDH were associated with increased mortality. Non-linear machine learning algorithms can augment single variable inputs for predicting mortality.

The ongoing coronavirus disease 2019 (COVID-19) pandemic has led to more than 200 million infected cases and 4.6 million deaths worldwide, and the numbers continue to grow. The disease presentation varies, and while most patients will present with a mild disease course, 5% will eventually develop significant respiratory failure, some despite initially presenting with mild symptoms. Early detection of patients at risk for deterioration is crucial for decisions regarding hospitalization, monitoring, timing, and extent of treatment.

The coronavirus disease 2019 (COVID-19) pandemic has had a profound impact on our world and has cost millions their lives. It has disrupted economies and education systems and has taken away means of support from masses of people around the world. No wonder this pandemic is like a black hole, drawing in all resources and all expertise. In the scientific arena, the pandemic has created a tremendous opportunity for new and exciting synergies between different disciplines.

Background: It is important to predict acute cholecystitis (AC) before a laparoscopic cholecystectomy because inflammation of the gallbladder predicts the need for open conversion and subsequent morbidity after a laparoscopic cholecystectomy.

Objectives: To create an index based on clinical, laboratory, and ultrasound criteria on admission that will predict AC on pathological examination in patients presenting acutely.

Methods: We retrospectively reviewed consecutive cases of emergency laparoscopic cholecystectomies conducted by three experienced surgeons between 1 October 2014 and 31 January 2018. Independent variables were age, sex, presenting symptoms, admission laboratory tests, and ultrasound findings. The outcome variable was AC on histological examination. An index was created from all variables that added significantly to the logistic regression analysis.

Results: Eight variables that contributed significantly to the model, included age, male sex, vomiting on admission, an increased proportion of neutrophils, a normal aspartate aminotransferase test, a normal serum amylase test result, a thick gall bladder wall, and pericholecystic fluid. An index of ≤ 2 to ≥ 8 created from those variables had a graded risk for AC of 1.8% to 92.0% with a c-statistic of 0.86 (95% confidence interval 0.81–0.91). Operating time and bleeding increased in those with a higher index.

Conclusions: An index including age, sex, symptoms, and selected laboratory results as well as ultrasound characteristics had an excellent graded risk in the prediction of histological AC that was associated with operating time and an increased risk of bleeding during the operation.

Background: For locally advanced rectal cancer patients a watch-and-wait strategy is an acceptable treatment option in cases of complete tumor response. Clinicians need robust methods of patient selection after neoadjuvant chemoradiation.

Objectives: To predict pathologic complete response (pCR) using computer vision. To analyze radiomic wavelet transform to predict pCR.

Methods: Neoadjuvant chemoradiation for patients with locally advanced rectal adenocarcinoma who passed computed tomography (CT)-based simulation procedures were examined. Gross tumor volume was examind on the set of CT simulation images. The volume has been analyzed using radiomics software package with wavelets feature extraction module. Statistical analysis using descriptive statistics and logistic regression was performed was used. For prediction evaluation a multilayer perceptron algorithm and Random Forest model were used.

Results: In the study 140 patients with II–III stage cancer were included. After a long course of chemoradiation and further surgery the pathology examination showed pCR in 38 (27.1%) of the patients. CT-simulation images of tumor volume were extracted with 850 parameters (119,000 total features). Logistic regression showed high value of wavelet contribution to model. A multilayer perceptron model showed high predictive importance of wavelet. We applied random forest analysis for classifying the texture and predominant features of wavelet parameters. Importance was assigned to wavelets.

Conclusions: We evaluated the feasibility of using non-diagnostic CT images as a data source for texture analysis combined with wavelets feature analysis for predicting pCR in locally advanced rectal cancer patients. The model performance showed the importance of including wavelets features in radiomics analysis.

Background: Multiple myeloma (MM) affects the long bones in 25% of patients. The advent of positron-emission tomography/computed tomography (PET/CT) scanners offers the possibility of both metabolic and radiographic information and may help determine fracture risk. To the best of our knowledge, no published study correlates these two factors with long bone fractures.

Objective: To evaluate the impact of PET/CT on fracture risk assessment in multiple myeloma patients.

Methods: We identified all bone marrow biopsy proven multiple myeloma patients from 1 January 2010 to 31 January 2015 at a single institution. We prospectively followed patients with long bone lesions using PET/CT scan images.

Results: We identified 119 patients (59 males/60 females) with 256 long bone lesions. Mean age at diagnosis was 58 years. The majority of lesions were in the femur (n=150, 59%) and humerus (n=84, 33%); 13 lesions in 10 patients (8%) required surgery for impending (n=4) or actual fracture (n=9). Higher median SUVmax was measured for those with cortical involvement (8.05, range 0–50.8) vs. no involvement (5.0, range 2.1–18.1). SUVmax was found to be a predictor of cortical involvement (odds ratio = 1.17, P = 0.026). No significant correlation was found between SUVmax and pain or fracture (P = 0.43).

Conclusions: Improved medical treatment resulted improvement in 8% of patients with an actual or impending fracture. The orthopedic surgeons commonly use the Mirels classification for long bone fracture prediction. Adding PET/CT imaging to study in myeloma long bone lesions did not predict fracture risk directly but suggested it indirectly by cortical erosion.

Background: Recently, Israel established the first national-level adult cardiac surgery database, which was linked to the Society of Thoracic Surgeons (STS).

Objectives: To validate and compare the STS predicted risk of mortality (PROM) to logistic EuroSCORE I (LESI) and EuroSCORE II (ESII) in Israeli patients undergoing cardiac surgery.

Methods: We retrospectively studied 1279 consecutive patients who underwent cardiac surgeries with a calculable PROM. Data were prospectively entered into our database and used to calculate PROM, LESI, and ESII. Scores were normalized and correlated using linear regression and Pearson's test. To examine model calibration, we plotted the total observed versus expected mortality for each score and across five risk-score subgroups. Model discrimination was assessed by measuring the area under the receiver operating curves.

Results: The observed 30-day operative mortality was 1.95%. The median (IQ1; IQ3) PROM, LESI, and the ESII scores were 1.45% (0.69; 3.22), 4.54% (2.28; 9.27), and 1.88% (1.18; 3.54), respectively, with observed over expected ratios of 0.63 (95% confidence interval [95%CI] 0.42–0.93), 0.59 (95%CI 0.40–0.87), and 0.24 (95%CI 0.17–0.36), respectively, (STS vs. ESII P = 0.36, STS vs. LESI P = 0.0001). There was good correlation among all scores. All models overestimated mortality. Model discrimination was high and similar for all three scores. Model calibration of the STS, PROM, and ESII were more accurate than the LESI, particularly in higher risk subgroups.

Conclusions: All scores overestimated mortality. In Israeli patients, the STS, PROM, and ESII risk-scores were more reliable metrics than LESI, particularly in higher risk patients.

Objectives: To define preoperative predictors for MPC in prostatectomy specimens and to examine the accuracy of such prediction.

Methods: Data collected on 1526 consecutive radical prostatectomy patients operated in a single center between 2003 and 2008 included: age, body mass index, preoperative prostate-specific antigen level, biopsy Gleason score, clinical stage, percentage of positive biopsy cores, and maximal core length (MCL) involvement. MPC was defined as < 5% of prostate volume involvement with organ-confined Gleason score ≤ 6. Univariate and multivariate logistic regression analyses were used to define independent predictors of minimal disease. Classification and Regression Tree (CART) analysis was used to define cutoff values for the predictors and measure the accuracy of prediction.

Results: MPC was found in 241 patients (15.8%). Clinical stage, biopsy Gleason`s score, percent of positive biopsy cores, and maximal involved core length were associated with minimal disease (OR 0.42, 0.1, 0.92, and 0.9, respectively). Independent predictors of MPC included: biopsy Gleason score, percent of positive cores and MCL (OR 0.21, 095 and 0.95, respectively). CART showed that when the MCL exceeded 11.5%, the likelihood of MPC was 3.8%. ;Conversely, when applying the most favorable preoperative conditions (Gleason ≤ 6, < 20% positive cores, MCL ≤ 11.5%) the chance of minimal disease was 41%.

Conclusions: Biopsy Gleason score, the percent of positive cores and MCL are independently associated with MPC. While preoperative prediction of significant prostate cancer was accurate, clinical prediction of MPC was incorrect 59% of the time. Caution is necessary when implementing clinical data as selection criteria for active surveillance. 

 Background: Burn injuries are extremely common and may impose a serious load on public health around the world.

 Objectives: To compare mortality rates and length of hospitalization according to the identified risk factors, extent of burn, gender and age.

Methods: In this retrospective study, data from 558 archive files of hospitalization due to burns as the diagnosis in patients of all ages, between the years 2001 and 2002, were analyzed to identify the risk factors for mortality and length of hospitalization.

Results: Males comprised 62.4% of the hospitalized burn patients. The mortality rate was 3.2% (n=18) and among them 55.6% were women. Fifty percent of the fatality cases were over 48 years old, with statistically significant correlation of mortality rate and age. Most of the fatality cases (66.7%) had burns with total burn surface area (TBSA) larger than 40%. The multiple logistic regression model showed that leukocyte count on admission, TBSA, and age are the most important predictors of mortality. Smoke inhalation was not found to be an independent risk factor.

Conclusions: Using a statistical model for estimating the mortality rate, this study found that white blood cell count at admission, TBSA, and age were the most significant predictors of mortality. 

Background: Frequent readmissions significantly contribute to health care costs as well as work load in internal medicine wards.

Objective: To develop a simple scoring method that includes basic demographic and medical characteristics of  elderly patients in internal medicine wards, which would allow prediction of readmission within 3 months of discharge.

Methods: We conducted a retrospective observational study of 496 hospitalized patients using data collected from discharge letters in the computerized archives. Univariate and multivariate logistic regression analyses were performed and factors that were significantly associated with readmission were selected to construct a scoring tool. Validity was assessed in a cohort of 200 patients.

Results: During a 2 year follow-up 292 patients were readmitted at least once within 3 months of discharge. Age 80 or older, any degree of impaired cognition, nursing home residence, congestive heart failure, and creatinine level > 1.5 mg/dl were found to be strong predictors of readmission. The presence of each variable was scored as 1. A score of 3 or higher in the derivation and validation cohorts corresponded with a positive predictive value of 80% and 67%, respectively, when evaluating the risk of rehospitalization.

Conclusions: We propose a practical, readily available five-item scoring tool that allows prediction of most unplanned readmissions within 3 months. The strength of this scoring tool, as compared with previously published scores, is its simplicity and straightforwardness.
 

Background: Myeloperoxidase levels were shown to reflect endothelial dysfunction, inflammation, atherosclerosis and oxidative stress.

Objectives: To examine the role of circulating myeloperoxidase, a leukocyte-derived enzyme, as a predictor of mortality in patients with congestive heart failure.

Methods: Baseline serum MPO[1] levels were measured in 285 consecutive CHF[2] patients and 35 healthy volunteers. N-terminal pro-brain natriuretic peptide and high sensitivity C-reactive protein concentrations were also measured. The primary outcome endpoint was overall mortality.

Results: MPO levels were significantly elevated in patients with CHF compared to healthy volunteers (P = 0.01). During a mean follow-up of 40.9 ± 11.3 months there were 106 deaths. On a univariate Cox regression analysis MPO levels were of marginal value (P = 0.07) whereas NT-proBNP[3] was of considerable value (P < 0.0001) in predicting all-cause mortality. By dividing our cohort according to NT-proBNP levels into high, intermediate and low risk groups a clear difference in mortality was shown. By further dividing the patient cohort according to MPO levels above or below the median (122.5 ng/ml), mortality prediction improved in the patients with intermediate NT-proBNP values.

[1] MPO = myeloperoxidase

[2] CHF = congestive heart failure

[3] NT-proBNP = N-terminal pro-brain natriuretic peptide

Medical screening is not a tangible existent tool in autoimmune disorders as it is in other illnesses. Numerous attempts are made to identify individuals destined to develop an autoimmune disease, including analysis of the genetic background, which along with the immunological profile, may assist in identifying those individuals. If these efforts turn out to be successful they may lead to the possibility of proactive measures that might prevent the emergence of such disorders. This review will summarize the attempts made to pursue autoantibodies specific for the central nervous system as potential predictors of autoimmune neurological disorders.

Background: Left ventricular outflow gradient is associated with increased morbidity and mortality in hypertrophic cardiomyopathy. Alcohol septal ablation is the alternative to surgery in cases refractory to drug therapy. The implication of LVOG[1] measured 1 week post-ASA[2] for prediction of outcome is unknown.

Objective: To observe the pattern of LVOG course and prediction of long-term clinical and hemodynamic outcome of ASA.

Methods: Baseline clinical and echocardiographic parameters were prospectively recorded in 14 consecutive patients with a first ASA, at the time of ASA, 3 and 7 days after ASA (in-hospital) and 3 and ≥ 12 months after ASA (last follow-up).

Results: There was improvement in NYHA class, exercise parameters and LVOG in 11 of 14 patients (P < 0.005 in all). Maximal creatine kinase level was lower than 500 U/L in those without such improvement and 850 U/L or higher in successful cases. LVOG dropped from 79 ± 30 to 19 ± 6 mmHg after the ASA. LVOG was 50 ± 21 mmHg on day 3, 39 ± 26 on day 7, 32 ± 26 at 3 months and 24 ± 20 mmHg at last follow-up. LVOG identified 27% sustained procedural successes on day 3 and 73% on day 7. The overall predictive accuracy of the test for sustained success and failure was 36% on day 3 and 71% on day 7. Combination of maximal CK[3] and LVOG on day 7 showed four distinct outcome patterns: "early success" with low LVOG and high CK (73% of successful cases), "late success" with high LVOG and high CK, and "early failure" and "late failure" with both low CK and high or low LVOG, respectively
Conclusion: LVOG measurement 7 days post-ASA combined with maximal CK levels predicts late procedural outcome in the majority of patients