Category: Digital Surgery and Telemedicine

A machine learning tool predicts musculoskeletal injury risk in surgeons

A study highlighted alarming findings regarding musculoskeletal strain among surgeons, with 90.6% reporting related pain. Key risk factors include long surgical interventions without breaks, obesity, and improper laparoscopic screen positioning. A novel online calculator was developed to predict musculoskeletal injury risk based on these factors, aiming to facilitate personalized prevention strategies. Enhanced ergonomics training was […]

Econsent via Patient Portal Reduces Surgical Delays

Econsent delivery through a patient portal significantly improves operational efficiency by allowing patients to complete their consent forms prior to surgery. In a cohort of 7,672 patients, those who signed their consent via the portal experienced fewer first-start delays compared to those who did so on the day of surgery (odds ratio, 1.59). Patients expressed […]

Early endoscopic treatment reduces hospital stay in acute biliary pancreatitis

A meta-analysis of 8,801 patients indicated that early endoscopic retrograde cholangiopancreatography (ERCP) for acute biliary pancreatitis is as safe as delayed ERCP. Notably, early ERCP significantly reduces hospital stay without increasing complications or mortality rates. These findings suggest that early intervention could enhance patient outcomes and warrant wider clinical adoption, offering substantial relief from the […]

Deep learning enhances ureter identification in laparoscopic surgery

A novel deep learning model, Ureternet, has demonstrated promising results in real-time ureter identification during laparoscopic colorectal surgery. Using a convolutional neural network-based approach, researchers achieved precision, recall, and dice coefficients of approximately 0.712, 0.722, and 0.716, respectively, on test data. The model operates swiftly, processing images in an average of 71 milliseconds. By potentially […]

Machine learning models enhance triage for abdominal pain surgery

Machine learning models significantly outperform traditional logistic regression in predicting emergency surgery for abdominal pain patients. An analysis of 38,214 individuals identified 4,208 requiring emergency surgical intervention. Key results showed that the Light GBM model achieved an area under the curve (AUC) of 0.899, improving clinical decision-making through enhanced sensitivity and specificity. Overall, these advancements […]

Triage-bot enhances emergency department triage efficiency in Canada.

Triage-bot, an AI-driven system based on the Canadian Triage and Acuity Scale, supports emergency department nurses by automating patient assessments. Key findings demonstrate that Triage-bot effectively measures vital signs and interprets patient expressions and tone for more accurate triage decisions. The systematic review indicates significant improvements in patient care through personalized instructions and remote monitoring. […]

Machine learning model surpasses clinical risk scores for GI bleeding

In a comparative analysis of an electronic health record-based machine learning model and established clinical risk scores for gastrointestinal bleeding, researchers found the model significantly outperformed the Glasgow-Blatchford Score and Oakland Score. With an area under the receiver-operating-characteristic curve (AUC) of 0.92 versus 0.89 (p

AI Model Effectively Quantifies Residual Pancreatic Cancer Post-Treatment

Development and validation of the ISGPP model marks a significant advancement in automating residual pancreatic cancer (RPC) quantification. The model demonstrated robust performance across diverse scanners, achieving mean F1 scores of 0.81-0.71 upon validation. A comprehensive dataset of 528 unique H&E slides from 528 patients facilitated training the model, which is now publicly available. This […]

New neural network enhances drug discovery for surgical treatments

A novel attention-based convolution transpositional interfusion network (ACTIN) has shown promising results for drug discovery with limited data. Utilizing just 393 training instances, ACTIN achieved state-of-the-art performance by leveraging graph convolution and transformer mechanisms to analyze drug and transcriptome data. It identifies pharmacophores that may benefit surgical patients, aiming to reduce complications and expedite recovery. […]