The prediction of surgical complications using artificial intelligence in patients undergoing major abdominal surgery: A systematic review

Predicting Glaucoma Surgical Outcomes Using Neural Networks and Machine Learning on Electronic Health Records

Purpose

To develop machine learning (ML) and deep learning (DL) models to predict glaucoma surgical outcomes, including postoperative intraocular pressure, use of ocular antihypertensive medications, and need for repeat surgery.

Methods

We identified glaucoma surgeries performed at Stanford from 2013-2024, with two or more postoperative visits with intraocular pressure (IOP) measurement. Patient features were identified from the electronic health record (EHR), including demographics, prior diagnosis and procedure codes, medications and eye exam findings. Classical ML and DL models were developed to predict which glaucoma surgeries would result in surgical failure, defined as (1) IOP not reduced by more than 20% of preoperative baseline on two consecutive postoperative visits, (2) increased classes of glaucoma medications, and (3) need for additional glaucoma surgery or revision of original surgery.

Results

A total of 2398 glaucoma surgeries of 1571 patients were included, of which 1677 surgeries met failure criteria. Random forest performed best for prediction of overall surgical failure, with accuracy of 75.5% and area under the receiver operator curve (AUROC) of 76.7%, similar to the deep learning model (accuracy 75.5%, AUROC 76.6%). Across all models, prediction performance was better for IOP outcomes (AUROC 86%) than need for an additional surgery (AUROC 76%) or need for additional glaucoma medication (AUC 70%).

Conclusions

ML and DL algorithms can predict glaucoma surgery outcomes using structured data inputs from EHRs.

Translational Relevance

Models that predict outcomes of glaucoma surgery may one day provide the basis for clinical decision support tools supporting surgeons in personalizing glaucoma treatment plans.

A feasibility study assessing quantitative indocyanine green angiographic predictors of reconstructive complications following nipple-sparing mastectomy

Introduction

Immediate post-mastectomy breast reconstruction offers benefits; however, complications can compromise outcomes. Intraoperative indocyanine green fluorescence angiography (ICGFA) may mitigate perfusion-related complications (PRC); however, its interpretation remains subjective. Here, we examine and develop methods for ICGFA quantification, including machine learning (ML) algorithms for predicting complications.

Methods

ICGFA video recordings of flap perfusion from a previous study of patients undergoing nipple-sparing mastectomy (NSM) with either immediate or staged immediate (delayed by a week due to perfusion insufficiency) reconstructions were analysed. Fluorescence intensity time series data were extracted, and perfusion parameters were interrogated for overall/regional associations with postoperative PRC. A naïve Bayes ML model was subsequently trained on a balanced data subset to predict PRC from the extracted meta-data.

Results

The analysable video dataset of 157 ICGFA featured females (average age 48 years) having oncological/risk-reducing NSM with either immediate (n=90) or staged immediate (n=26) reconstruction. For those delayed, peak brightness at initial ICGFA was lower (p<0.001) and significantly improved (both quicker-onset and brighter p=0.001) one week later. The overall PRC rate in reconstructed patients (n=116) was 11.2%, with such patients demonstrating significantly dimmer (overall, p=0.018, centrally, p=0.03, and medially, p=0.04) and slower-onset (p=0.039) fluorescent peaks with shallower slopes (p=0.012) than uncomplicated patients with ICGFA. Importantly, such relevant parameters were converted into a whole field of view heatmap potentially suitable for intraoperative display. ML predicted PRC with 84.6% sensitivity and 76.9% specificity.

Conclusion

Whole breast quantitative ICGFA assessment reveals statistical associations with PRC that are potentially exploitable via ML.

Artificial intelligence algorithms for predicting post-operative ileus after laparoscopic surgery

Objective

By constructing a predictive model using machine learning and deep learning technologies, we aim to understand the risk factors for postoperative intestinal obstruction in laparoscopic colorectal cancer patients, and establish an effective artificial intelligence-based predictive model to guide individualized prevention and treatment, thus improving patient outcomes.

Methods

We constructed a model of the artificial intelligence algorithm in Python. Subjects were randomly assigned to either a training set for variable identification and model construction, or a test set for testing model performance, at a ratio of 7:3. The model was trained with ten algorithms. We used the AUC values of the ROC curves, as well as accuracy, precision, recall rate and F1 scores.

Results

The results of feature engineering composited with the GBDT algorithm showed that opioid use, anesthesia duration, and body weight were the top three factors in the development of POI. We used ten machine learning and deep learning algorithms to validate the model, and the results were as follows: the three algorithms with best accuracy were XGB (0.807), Decision Tree (0.807) and Neural DecisionTree (0.807); the two algorithms with best precision were XGB (0.500) and Decision Tree (0.500); the two algorithms with best recall rate were adab (0.243) and Decision Tree (0.135); the two algorithms with highest F1 score were adab (0.290) and Decision Tree (0.213); and the three algorithms with best AUC were Gradient Boosting (0.678), XGB (0.638) and LinearSVC (0.633).

Conclusion

This study shows that XGB and Decision Tree are the two best algorithms for predicting the risk of developing ileus after laparoscopic colon cancer surgery. It provides new insight and approaches to the field of postoperative intestinal obstruction in colorectal cancer through the application of machine learning techniques, thereby improving our understanding of the disease and offering strong support for clinical decision-making.

Machine learning models to predict success of endoscopic sleeve gastroplasty using total and excess weight loss percent achievement: a multicentre study

BACKGROUND

The large amount of heterogeneous data collected in surgical/endoscopic practice calls for data-driven approaches as machine learning (ML) models. The aim of this study was to develop ML models to predict endoscopic sleeve gastroplasty (ESG) efficacy at 12 months defined by total weight loss (TWL) % and excess weight loss (EWL) % achievement. Multicentre data were used to enhance generalizability: evaluate consistency among different center of ESG practice and assess reproducibility of the models and possible clinical application. Models were designed to be dynamic and integrate follow-up clinical data into more accurate predictions, possibly assisting management and decision-making.

METHODS

ML models were developed using data of 404 ESG procedures performed at 12 centers across Europe. Collected data included clinical and demographic variables at the time of ESG and at follow-up. Multicentre/external and single center/internal and temporal validation were performed. Training and evaluation of the models were performed on Python's scikit-learn library. Performance of models was quantified as receiver operator curve (ROC-AUC), sensitivity, specificity, and calibration plots.

RESULTS

Multicenter external validation: ML models using preoperative data show poor performance. Best performances were reached by linear regression (LR) and support vector machine models for TWL% and EWL%, respectively, (ROC-AUC: TWL% 0.87, EWL% 0.86) with the addition of 6-month follow-up data. Single-center internal validation: Preoperative data only ML models show suboptimal performance. Early, i.e., 3-month follow-up data addition lead to ROC-AUC of 0.79 (random forest classifiers model) and 0.81 (LR models) for TWL% and EWL% achievement prediction, respectively. Single-center temporal validation shows similar results.

CONCLUSIONS

Although preoperative data only may not be sufficient for accurate postoperative predictions, the ability of ML models to adapt and evolve with the patients changes could assist in providing an effective and personalized postoperative care. ML models predictive capacity improvement with follow-up data is encouraging and may become a valuable support in patient management and decision-making.

Artificial Intelligence: Singularity Approaches

SUMMARY

Artificial intelligence (AI) has been a disruptive technology within health care, from the development of simple care algorithms to complex deep-learning models. AI has the potential to reduce the burden of administrative tasks, advance clinical decision-making, and improve patient outcomes. Unlocking the full potential of AI requires the analysis of vast quantities of clinical information. Although AI holds tremendous promise, widespread adoption within plastic surgery remains limited. Understanding the basics is essential for plastic surgeons to evaluate the potential uses of AI. This review provides an introduction of AI, including the history of AI, key concepts, applications of AI in plastic surgery, and future implications.

Harvesting the Power of Artificial Intelligence for Surgery: Uses, Implications, and Ethical Considerations

Artificial intelligence is rapidly advancing, especially with the advent of ChatGPT technology, and its role in the world of medicine is expanding. Within surgery, AI has the capacity to improve efficiency and results in surgical treatments; however, it similarly has the potential to impose harm onto patients and undermine the role of medical providers. Its benefits may include improvements in surgical outcomes, spanning from enhanced pre-operative diagnostic capabilities to more refined intra-operative techniques, and long term patient experiences, by identifying and reducing complications. Nevertheless apprehensions revolve around laymen use potentially resulting in inappropriate therapeutic interventions, in addition to safety and ethical risks surrounding the use of patient data. Various strategies towards mitigating these harms must be considered, such as patient disclaimers and secondary review policies. While artificial intelligence brings exciting advancements to surgery, its integration must be cautiously monitored.

Machine learning models in clinical practice for the prediction of postoperative complications after major abdominal surgery

Complications after surgery have a major impact on short- and long-term outcomes, and decades of technological advancement have not yet led to the eradication of their risk. The accurate prediction of complications, recently enhanced by the development of machine learning algorithms, has the potential to completely reshape surgical patient management. In this paper, we reflect on multiple issues facing the implementation of machine learning, from the development to the actual implementation of machine learning models in daily clinical practice, providing suggestions on the use of machine learning models for predicting postoperative complications after major abdominal surgery.

Prehabilitation for general surgery: a systematic review of randomized controlled trials

BACKGROUND

Prehabilitation seeks to optimize patient health before surgery to improve outcomes. Randomized controlled trials (RCTs) have been conducted on prehabilitation, however an updated synthesis of this evidence is required across General Surgery to inform potential Supplementary discipline-level protocols. Accordingly, this systematic review of RCTs aimed to evaluate the use of prehabilitation interventions across the discipline of General Surgery.

METHODS

This study was registered with PROSPERO (CRD42023403289), and adhered to PRISMA 2020 and SWiM guidelines. PubMed/MEDLINE and Ovid Embase were searched to 4 March 2023 for RCTs evaluating prehabilitation interventions within the discipline of General Surgery. After data extraction, risk of bias was assessed using the Cochrane RoB 2 tool. Quantitative and qualitative data were synthesized and analysed. However, meta-analysis was precluded due to heterogeneity across included studies.

RESULTS

From 929 records, 36 RCTs of mostly low risk of bias were included. 17 (47.2%) were from Europe, and 14 (38.9%) North America. 30 (83.3%) investigated cancer populations. 31 (86.1%) investigated physical interventions, finding no significant difference in 16 (51.6%) and significant improvement in 14 (45.2%). Nine (25%) investigated psychological interventions: six (66.7%) found significant improvement, three (33.3%) found no significant difference. Five (13.9%) investigated nutritional interventions, finding no significant difference in three (60%), and significant improvement in two (40%).

CONCLUSIONS

Prehabilitation interventions showed mixed levels of effectiveness, and there is insufficient RCT evidence to suggest system-level delivery across General Surgery within standardized protocols. However, given potential benefits and non-inferiority to standard care, they should be considered on a case-by-case basis.

Machine learning versus logistic regression for the prediction of complications after pancreatoduodenectomy

BACKGROUND

Machine learning is increasingly advocated to develop prediction models for postoperative complications. It is, however, unclear if machine learning is superior to logistic regression when using structured clinical data. Postoperative pancreatic fistula and delayed gastric emptying are the two most common complications with the biggest impact on patient condition and length of hospital stay after pancreatoduodenectomy. This study aimed to compare the performance of machine learning and logistic regression in predicting pancreatic fistula and delayed gastric emptying after pancreatoduodenectomy.

METHODS

This retrospective observational study used nationwide data from 16 centers in the Dutch Pancreatic Cancer Audit between January 2014 and January 2021. The area under the curve of a machine learning and logistic regression model for clinically relevant postoperative pancreatic fistula and delayed gastric emptying were compared.

RESULTS

Overall, 799 (16.3%) patients developed a postoperative pancreatic fistula, and 943 developed (19.2%) delayed gastric emptying. For postoperative pancreatic fistula, the area under the curve of the machine learning model was 0.74, and the area under the curve of the logistic regression model was 0.73. For delayed gastric emptying, the area under the curve of the machine learning model and logistic regression was 0.59.

CONCLUSION

Machine learning did not outperform logistic regression modeling in predicting postoperative complications after pancreatoduodenectomy.

Network prediction of surgical complication clusters: a prospective multicenter cohort study

Complicated relationships exist in both occurrence and progression of surgical complications, which are difficult to account for using a separate quantitative method such as prediction or grading. Data of 51,030 surgical inpatients were collected from four academic/teaching hospitals in a prospective cohort study in China. The relationship between preoperative factors, 22 common complications, and death was analyzed. With input from 54 senior clinicians and following a Bayesian network approach, a complication grading, cluster-visualization, and prediction (GCP) system was designed to model pathways between grades of complication and preoperative risk factor clusters. In the GCP system, there were 11 nodes representing six grades of complication and five preoperative risk factor clusters, and 32 arcs representing a direct association. Several critical targets were pinpointed on the pathway. Malnourished status was a fundamental cause widely associated (7/32 arcs) with other risk factor clusters and complications. American Society of Anesthesiologists (ASA) score ⩾3 was directly dependent on all other risk factor clusters and influenced all severe complications. Grade III complications (mainly pneumonia) were directly dependent on 4/5 risk factor clusters and affected all other grades of complication. Irrespective of grade, complication occurrence was more likely to increase the risk of other grades of complication than risk factor clusters.

One Decade of Declining Use of Defunctioning Stomas After Rectal Cancer Surgery in the Netherlands: Are We on the Right Track?

BACKGROUND

The beneficial effect of a defunctioning stoma in mitigating the consequences of anastomotic leakage after rectal cancer surgery is still debated.

OBJECTIVE

This study aims to reflect on a decade of rectal cancer surgery in terms of stoma construction and anastomotic leakage.

DESIGN

Retrospective observational study.

SETTING

This study used data from the Dutch Colorectal Audit from 2011 to 2020.

PATIENTS

Patients undergoing rectal cancer surgery with a primary anastomosis.

MAIN OUTCOME MEASURES

Primary outcome was anastomotic leakage. Secondary outcomes were minor complications, admission to intensive care, length of stay, readmission, and patient death.

RESULTS

A total of 13,263 patients were included in this study. A defunctioning stoma was constructed in 7106 patients (53.6%). Patients with a defunctioning stoma were less likely to develop anastomotic leakage (7.9% vs 13.0%), and if anastomotic leakage occurred, fewer patients needed surgical reintervention (37.7% vs 81.1%). An annual decrease in the construction of a defunctioning stoma was seen (69.8% in 2011 vs 51.8% in 2015 vs 29.7% in 2020), accompanied by a 5% increase in anastomotic leakage (9.1% in 2011 vs 14.1% in 2020). A defunctioning stoma was associated with a higher occurrence of minor complications, increased admissions to the intensive care unit, longer length of stay, and more readmissions within 90 days.

LIMITATION

This retrospective study is susceptible to confounders by indications, and there could be risk factors for anastomotic leakage and the use of a stoma that were not regarded.

CONCLUSIONS

The reduction in defunctioning stomas is paralleled with an increase in anastomotic leakage. However, patients with a defunctioning stoma also showed more minor complications, a prolonged length of stay, more intensive care admissions, and more readmissions. In our opinion, the trade-offs of selective use should be individually considered. See Video Abstract at http://links.lww.com/DCR/C137 .

UNA DCADA DISMINUYENDO EL USO DE ESTOMAS DISFUNCIONANTES EN LOS CASOS DE CNCER DE RECTO EN HOLANDA ESTAMOS HACIENDO LO CORRECTO

ANTECEDENTES:Aún se debate el efecto benéfico de la confección de un estoma disfuncionante para limitar las consecuencias de la fuga anastomótica en los casos de cirugía por cáncer de recto.OBJETIVO:Reflexiones sobre una década de cirugía por cáncer de recto en términos de confección de estomas y de fugas anastomóticas.DISEÑO:Estudio retrospectivo y observacional.AJUSTE:El presente estudio utilizó datos de la Auditoría Colorectal Holandesa entre 2011 y 2020.PACIENTES:Todos aquellos intervenidos por cáncer de recto con anastomosis primaria.PRINCIPALES MEDIDAS DE RESULTADO:El resultado primario fue evaluar la fuga anastomótica. Los resultados secundarios fueron las complicaciones menores, la permanencia en cuidados intensivos, la duración de la hospitalización, las rehospitalizaciones y las causas de muerte en los pacientes.RESULTADOS:Un total de 13.263 pacientes fueron incluidos en el presente estudio. Se confeccionó un estoma disfuncionante en 7.106 (53,6%) pacientes. Aquellos portadores de un estoma disfuncionante tenían menos probabilidades de desarrollar una fuga anastomótica (7,9 % frente a 13,0 %) y, si ocurría una fuga anastomótica, menos pacientes necesitaban reintervención quirúrgica (37,7 % frente a 81,1 %). Se observó una disminución anual en la confección de un estoma disfuncionante (69,8 % en 2011 frente a 51,8 % en 2015 frente a 29,7 % en 2020), acompañada de un aumento del 5 % en la fuga anastomótica (9,1 % en 2011 frente a 14,1 % en 2020). Un estoma disfuncionante se asoció con una mayor incidencia de complicaciones menores, permanencia en la unidad de cuidados intensivos, una estadía más prolongada y más rehospitalizaciones dentro de los 90 días.LIMITACIÓN:Estudio retrospectivo susceptible de factores de confusión según las indicaciones, donde podrían no haber sido considerados ciertos factores de riesgo con relación a la fuga anastomótica y a la confección de un estoma disfuncionante.CONCLUSIÓN:La reducción de estomas disfuncionantes es paralela con el aumento de la fuga anastomótica. Sin embargo, los pacientes con un estoma disfuncionante también mostraron más complicaciones menores, una estadía prolongada, más admisiones a cuidados intensivos y más rehospitalizaciones. En nuestra opinión, las ventajas y desventajas del uso selectivo de estomas disfuncionantes deben ser consideradas caso por caso. Consulte Video Resumen en https://links.lww.com/DCR/C137 . (Traducción-Dr. Xavier Delgadillo ).

Automated Capture of Intraoperative Adverse Events Using Artificial Intelligence: A Systematic Review and Meta-Analysis

Intraoperative adverse events (iAEs) impact the outcomes of surgery, and yet are not routinely collected, graded, and reported. Advancements in artificial intelligence (AI) have the potential to power real-time, automatic detection of these events and disrupt the landscape of surgical safety through the prediction and mitigation of iAEs. We sought to understand the current implementation of AI in this space. A literature review was performed to PRISMA-DTA standards. Included articles were from all surgical specialties and reported the automatic identification of iAEs in real-time. Details on surgical specialty, adverse events, technology used for detecting iAEs, AI algorithm/validation, and reference standards/conventional parameters were extracted. A meta-analysis of algorithms with available data was conducted using a hierarchical summary receiver operating characteristic curve (ROC). The QUADAS-2 tool was used to assess the article risk of bias and clinical applicability. A total of 2982 studies were identified by searching PubMed, Scopus, Web of Science, and IEEE Xplore, with 13 articles included for data extraction. The AI algorithms detected bleeding (n = 7), vessel injury (n = 1), perfusion deficiencies (n = 1), thermal damage (n = 1), and EMG abnormalities (n = 1), among other iAEs. Nine of the thirteen articles described at least one validation method for the detection system; five explained using cross-validation and seven divided the dataset into training and validation cohorts. Meta-analysis showed the algorithms were both sensitive and specific across included iAEs (detection OR 14.74, CI 4.7-46.2). There was heterogeneity in reported outcome statistics and article bias risk. There is a need for standardization of iAE definitions, detection, and reporting to enhance surgical care for all patients. The heterogeneous applications of AI in the literature highlights the pluripotent nature of this technology. Applications of these algorithms across a breadth of urologic procedures should be investigated to assess the generalizability of these data.

Artificial Intelligence in Surgical Learning

(1) Background: Artificial Intelligence (AI) is transforming healthcare on all levels. While AI shows immense potential, the clinical implementation is lagging. We present a concise review of AI in surgical learning; (2) Methods: A non-systematic review of AI in surgical learning of the literature in English is provided; (3) Results: AI shows utility for all components of surgical competence within surgical learning. AI presents with great potential within robotic surgery specifically (4) Conclusions: Technology will evolve in ways currently unimaginable, presenting us with novel applications of AI and derivatives thereof. Surgeons must be open to new modes of learning to be able to implement all evidence-based applications of AI in the future. Systematic analyses of AI in surgical learning are needed.

An orthopaedic intelligence application successfully integrates data from a smartphone-based care management platform and a robotic knee system using a commercial database

PURPOSE

To evaluate the feasibility of using a smartphone-based care management platform (sbCMP) and robotic-assisted total knee arthroplasty (raTKA) to collect data throughout the episode-of-care and assess if intra-operative measures of soft tissue laxity in raTKA were associated with post-operative outcomes.

METHODS

A secondary data analysis of 131 patients in a commercial database who underwent raTKA was performed. Pre-operative through six week post-operative step counts and KOOS JR scores were collected and cross-referenced with intra-operative laxity measures. A Kruskal-Wallis test or a Wilcoxon sign-rank was used to assess outcomes.

RESULTS

There were higher step counts at six weeks post-operatively in knees with increased laxity in both the lateral compartment in extension and medial compartment in flexion (p < 0.05). Knees balanced in flexion within < 0.5 mm had higher KOOS JR scores at six weeks post-operative (p = 0.034) compared to knees balanced within 0.5-1.5 mm.

CONCLUSION

A smartphone-based care management platform can be integrated with raTKA to passively collect data throughout the episode-of-care. Associations between intra-operative decisions regarding laxity and post-operative outcomes were identified. However, more robust analysis is needed to evaluate these associations and ensure clinical relevance to guide machine learning algorithms.

Development and Validation of an Artificial Intelligence-Based Model to Predict Gastroesophageal Reflux Disease After Sleeve Gastrectomy

Purpose

Prediction of the onset of de novo gastroesophageal reflux disease (GERD) after sleeve gastrectomy (SG) would be helpful in decision-making and selection of the optimal bariatric procedure for every patient. The present study aimed to develop an artificial intelligence (AI)-based model to predict the onset of GERD after SG to help clinicians and surgeons in decision-making.

Materials and Methods

A prospectively maintained database of patients with severe obesity who underwent SG was used for the development of the AI model using all the available data points. The dataset was arbitrarily split into two parts: 70% for training and 30% for testing. Then ranking of the variables was performed in two steps. Different learning algorithms were used, and the best model that showed maximum performance was selected for the further steps of machine learning. A multitask AI platform was used to determine the cutoff points for the top numerical predictors of GERD.

Results

In total, 441 patients (76.2% female) of a mean age of 43.7 ± 10 years were included. The ensemble model outperformed the other models. The model achieved an AUC of 0.93 (95%CI 0.88–0.99), sensitivity of 79.2% (95% CI 57.9–92.9%), and specificity of 86.1% (95%CI 70.5–95.3%). The top five ranked predictors were age, weight, preoperative GERD, size of orogastric tube, and distance of first stapler firing from the pylorus.

Conclusion

An AI-based model for the prediction of GERD after SG was developed. The model had excellent accuracy, yet a moderate sensitivity and specificity. Further prospective multicenter trials are needed to externally validate the model developed.

Graphical Abstract