Automated operative phase identification in peroral endoscopic myotomy.

BACKGROUND: Artificial intelligence (AI) and computer vision (CV) have revolutionized image analysis. In surgery, CV applications have focused on surgical phase identification in laparoscopic videos. We proposed to apply CV techniques to identify phases in an endoscopic procedure, peroral endoscopic myotomy (POEM). METHODS: POEM videos were collected from Massachusetts General and Showa University Koto Toyosu Hospitals. Videos were labeled by surgeons with the following ground truth phases: (1) Submucosal injection, (2) Mucosotomy, (3) Submucosal tunnel, (4) Myotomy, and (5) Mucosotomy closure. The deep-learning CV model-Convolutional Neural Network (CNN) plus Long Short-Term Memory (LSTM)-was trained on 30 videos to create POEMNet. We then used POEMNet to identify operative phases in the remaining 20 videos. The model's performance was compared to surgeon annotated ground truth. RESULTS: POEMNet's overall phase identification accuracy was 87.6% (95% CI 87.4-87.9%). When evaluated on a per-phase basis, the model performed well, with mean unweighted and prevalence-weighted F1 scores of 0.766 and 0.875, respectively. The model performed best with longer phases, with 70.6% accuracy for phases that had a duration under 5 min and 88.3% accuracy for longer phases. DISCUSSION: A deep-learning-based approach to CV, previously successful in laparoscopic video phase identification, translates well to endoscopic procedures. With continued refinements, AI could contribute to intra-operative decision-support systems and post-operative risk prediction.

Outcomes of Laparoscopic Heller Myotomy for Achalasia: 22-Year Experience.

INTRODUCTION: Laparoscopic Heller Myotomy is the most effective treatment of achalasia. We examined the durability of symptomatic relief, with and without fundoplication. METHODS: A single institution database between 1995 and 2017 was reviewed. Achalasia symptom severity was assessed by Eckardt scores (ES) obtained at 3-time points via patient questionnaire. Primary outcome was treatment success defined as ES of < 3. RESULTS: Completed surveys were returned by 130 patients (median follow-up of 6.6 years). A partial fundoplication was performed in 86%. At both 1-year and late follow-up, patients reported a significant improvement in ES compared to baseline (p < 0.05). Of those followed for ≥ 10 years (n = 44), 82% reported ES < 3 at 1-year (p < 0.001), and 78% at last follow-up (p < .001). Of patients who reported treatment success 1-year postoperatively (103/130), 85% continued to report symptomatic relief at last follow-up. Five-year cohort analysis did not show deterioration of dysphagia relief over time. The presence or absence of fundoplication had no impact on long-term outcome (p > 0.05). CONCLUSIONS: LHM provides immediate and durable symptomatic relief, with very few patients requiring further therapeutic intervention. Fundoplication does not appear to influence the durability of symptom relief. Treatment success at 1-year is predictive of long-lasting symptomatic relief.

Colorectal Surgical Site Infection Prevention Kits Prior to Elective Colectomy Improve Outcomes.

INTRODUCTION: Patient compliance with preoperative mechanical and antibiotic bowel preparation, skin washes, carbohydrate loading, and avoidance of fasting are key components of successful colorectal ERAS and surgical site infection (SSI)-reduction programs. In July 2016, we began a quality improvement project distributing a free SSI Prevention Kit (SSIPK) containing patient instructions, mechanical and oral bowel preparation, chlorhexidine washes, and carbohydrate drink to all patients scheduled for elective colectomy, with the goal of improving patient compliance and rates of SSI. METHODS: This was a prospective data audit of our first 221 SSIPK+ patients, who were compared to historical controls (SSIPK-) of 1760 patients undergoing elective colectomy from January 2013 to March 2017. A 1:1 propensity score system accounted for nonrandom treatment assignment. Matched patients' complications, particularly postoperative infection and ileus, were compared. RESULTS: SSIPK+ (n = 219) and SSIPK- (n = 219) matched patients were statistically identical on demographics, comorbidities, BMI, surgical indication, and procedure. SSIPK+ patients had higher compliance with mechanical (95% vs 71%, P < 0.001) and oral antibiotic (94% vs 27%, P < 0.001) bowel preparation. This translated into lower overall SSI rates (5.9% vs 11.4%, P = 0.04). SSIPK+ patients also had lower rates of anastomotic leak (2.7% vs 6.8%, P = 0.04), prolonged postoperative ileus (5.9% vs 14.2%, P < 0.01), and unplanned intubation (0% vs 2.3%, P = 0.02). Furthermore, SSIPK+ patients had shorter mean hospital length of stay (3.1 vs 5.4 d, P < 0.01) and had fewer unplanned readmissions (5.9% vs 14.6%, P < 0.001). There were no differences in rates of postoperative pneumonia, urinary tract infection, Clostridium difficile colitis, sepsis, or death. CONCLUSION: Provision of a free-of-charge SSIPK is associated with higher patient compliance with preoperative instructions and significantly lower rates of surgical site infections, lower rates of prolonged postoperative ileus, and shorter hospital stays with fewer readmissions. Widespread utilization of such a bundle could therefore lead to significantly improved outcomes.

Third-time ("redo-redo") anti-reflux surgery: patient-reported outcomes after a thoracoabdominal approach.

BACKGROUND: Approximately 3-6% of patients undergoing anti-reflux surgery require "redo" surgery for persistent gastroesophageal reflux disease (GERD). Further surgery for patients with two failed prior anti-reflux operations is controversial due to the morbidity of reoperation and poor outcomes. We examined our experience with surgical revision of patients with at least two failed anti-reflux operations. METHODS: Adults undergoing at least a second-time revision anti-reflux surgery between 1999 and 2017 were eligible. The primary outcomes were general and disease-specific quality-of-life (QoL) scores determined by Short-Form-36 (SF36) and GERD-Health-Related QoL (GERD-HRQL) instruments, respectively. Secondary outcomes included perioperative morbidity and mortality. RESULTS: Eighteen patients undergoing redo-redo surgery (13 with 2 prior operations, 5 with 3 prior operations) were followed for a median of 6 years [IQR 3, 12]. Sixteen patients (89%) underwent open revisions (14 thoracoabdominal, 2 laparotomy) and two patients had laparoscopic revisions. Indications for surgery included reflux (10 patients), regurgitation (5 patients), and dysphagia (3 patients). Intraoperative findings were mediastinal wrap herniation (9 patients), misplaced wrap (2 patients), mesh erosion (1 patient), or scarring/stricture (6 patients). Procedures performed included Collis gastroplasty + fundoplication (6 patients), redo fundoplication (5 patients), esophagogastrectomy (4 patients), and primary hiatal closure (3 patients). There were no deaths and 13/18 patients (72%) had no postoperative complications. Ten patients completed QoL surveys; 8 reported resolution of reflux, 6 reported resolution of regurgitation, while 4 remained on proton-pump inhibitors (PPI). Mean SF36 scores (± standard deviation) in the study cohort in the eight QoL domains were as follows: physical functioning (79.5 [± 19.9]), physical role limitations (52.5 [± 46.3]), emotional role limitations (83.3 [± 36.1]), vitality (60.0 [± 22.7]), emotional well-being (88.4 [± 8.7]), social functioning (75.2 [± 31.0]), pain (66.2 [± 30.9]), and general health (55.0 [± 39.0]). CONCLUSION: An open thoracoabdominal approach in appropriately selected patients needing third-time anti-reflux surgery carries low morbidity and provides excellent results as reflected in QoL scores.

Computer Vision Analysis of Intraoperative Video: Automated Recognition of Operative Steps in Laparoscopic Sleeve Gastrectomy.

OBJECTIVE(S): To develop and assess AI algorithms to identify operative steps in laparoscopic sleeve gastrectomy (LSG). BACKGROUND: Computer vision, a form of artificial intelligence (AI), allows for quantitative analysis of video by computers for identification of objects and patterns, such as in autonomous driving. METHODS: Intraoperative video from LSG from an academic institution was annotated by 2 fellowship-trained, board-certified bariatric surgeons. Videos were segmented into the following steps: 1) port placement, 2) liver retraction, 3) liver biopsy, 4) gastrocolic ligament dissection, 5) stapling of the stomach, 6) bagging specimen, and 7) final inspection of staple line. Deep neural networks were used to analyze videos. Accuracy of operative step identification by the AI was determined by comparing to surgeon annotations. RESULTS: Eighty-eight cases of LSG were analyzed. A random 70% sample of these clips was used to train the AI and 30% to test the AI's performance. Mean concordance correlation coefficient for human annotators was 0.862, suggesting excellent agreement. Mean (±SD) accuracy of the AI in identifying operative steps in the test set was 82% ±â€Š4% with a maximum of 85.6%. CONCLUSIONS: AI can extract quantitative surgical data from video with 85.6% accuracy. This suggests operative video could be used as a quantitative data source for research in intraoperative clinical decision support, risk prediction, or outcomes studies.

Outcomes of Extended Lymphadenectomy for Gastroesophageal Carcinoma: A Large Western Series.

BACKGROUND: The extent of lymph node dissection for patients with gastroesophageal carcinoma remains controversial. We sought to examine the perioperative risk and survival outcomes in a large Western series of patients undergoing limited (D0/D1) vs extended (D1+/D2) lymphadenectomy (LAD) for gastroesophageal carcinoma. STUDY DESIGN: Clinicopathologic and treatment factors for 520 patients with gastroesophageal carcinoma undergoing potentially curative resection at a single institution from 1995 to 2017 were analyzed for their impact on perioperative morbidity and mortality, lymph node yield, and overall survival. RESULTS: A total of 362 (70%) patients underwent D0/D1 LAD and 158 (30%) underwent D1+/D2 LAD. Median follow-up was 3.1 years. Patients undergoing D1+/D2 LAD were more likely to have distal tumors, to undergo distal/subtotal/total gastrectomy, and to undergo operation at a more contemporary time than patients undergoing D0/D1 LAD. The median number of lymph nodes examined and the percentage of patients with 16 or more lymph nodes examined was 16 and 53%, respectively, in the D0/D1 group vs 27 and 89%, respectively, in the D1+/D2 group. There were no differences in the rates of major complications (16.6% vs 14.6%) or operative mortality (2.8% vs 0.6%) between the D0/D1 and D1+/D2 groups, respectively. Patients undergoing D1+/D2 LAD had significantly improved overall survival (hazard ratio 0.74; p = 0.035) compared with those undergoing D0/D1 LAD on univariate analysis, but this survival benefit disappeared when controlling for the time period of operation. CONCLUSIONS: Gastrectomy with extended (D1+/D2) LAD can be performed safely at a high-volume Western center, and it improves nodal yield significantly and ensures accurate pathologic staging.

History of Minimally Invasive Surgical Oncology.

Introduction of the fiberoptic light-source and CCD chip camera resulted in the rapid growth of minimally invasive surgical procedures. In surgical oncology, the change came slowly owing to concerns about adhering to oncological principals while learning to use new technology. Pioneers in minimally invasive colorectal surgery proved that minimally invasive resection for cancer was oncologically noninferior to traditional surgery. Early adopters treating esophageal and gastric cancer established that a minimally invasive approach was feasible with lower morbidity and equivalent oncologic outcomes. These results provide a basis for the extension of minimally invasive surgical techniques to other types of cancer surgery.

Timing of Carotid Endarterectomy After Stroke: Retrospective Review of Prospectively Collected National Database.

OBJECTIVE: Our objective was to identify the postoperative risk associated with different timing intervals of repair. BACKGROUND: Timing of carotid intervention in poststroke patients is widely debated with the scales balanced between increased periprocedural risk and recurrent neurologic event. National database reviews show increased risk to patients treated within the first 2 days of a neurologic event compared to those treated after 6 days. METHODS: Utilizing Vascular Quality Initiative data, all carotid interventions performed on stroke patients between the years 2012 and 2017 were queried. Patients were then stratified based on the timing of surgery from their stroke (<48 hours, 3-7 days, 8-14 days, >15 days). Major outcomes included postoperative stroke, death, and myocardial infarction. RESULTS: A total of 8404 patients were included being predominantly men (5281, 62.8%), with an average age of 69 (±10). Patients treated at greater than 8 days showed significantly less risk of postoperative combined stroke/death and postoperative stroke. There were no significant differences in postoperative stroke or death between the 8 to 14 and greater than 15 days groups.Multivariate regression analysis showed that delayed timing of surgery between 3 and 7 days was protective for postoperative stroke/death (P = 0.003) and any postoperative complication (P = 0.028). Delaying surgery to more than 8 days after stroke was protective for postoperative stroke/death (P < 0.001), postoperative stroke (P < 0.001), and any postoperative complication (P < 0.001). CONCLUSIONS: Carotid revascularization should occur no sooner than 48 hours after index stroke event. Surgeons should strive to operate between 8 and 14 days to protect against postoperative stroke/death.

Addition of a scripted pre-operative patient education module to an existing ERAS pathway further reduces length of stay.

BACKGROUND: While enhanced recovery pathways (ERAS) appear to be beneficial for post-operative outcomes, there have been no studies evaluating the specific role of patient education within an ERAS pathway. METHODS: We identified all colectomies performed at our institution since initiation of an ERAS protocol, excluding for mortality and length of stay >30 days. Patients who received preoperative education by a nurse practitioner via a scripted telephone call were compared to patients who did not receive education using the NSQIP database. We then evaluated differences in surgical complications and length of stay among these cohorts. RESULTS: Patients who received scripted education phone calls had a significantly shorter mean length of stay when compared to patients that receiving usual care (3.0 ±â€¯2.2 vs 3.7 ±â€¯3.2 days; p = 0.005). Subgroup analysis demonstrates strongest benefit in patients undergoing left colectomy and laparoscopic surgery. CONCLUSIONS: Scripted patient education modules may shorten length of stays and postoperative complications, even when added to an already existing ERAS bundle, which may translate into significant hospital cost savings.

Lower Esophageal Sphincter Augmentation for Gastroesophageal Reflux Disease: The Safety of a Modern Implant.

INTRODUCTION: Use of the magnetic sphincter augmentation device (MSAD) for gastroesophageal reflux disease (GERD) is increasing. As this innovative treatment for GERD gains widespread use and adoption, an assessment of its safety since U.S. market introduction is presented. METHODS: Events were collected from the Manufacturer and User Facility Device Experience (MAUDE) database, which reports events submitted to the Food and Drug Administration (FDA) of suspected device-associated deaths, serious injuries, and malfunctions. The reporting period was from March 22, 2012 (FDA approval) through May 31, 2016, and included only events occurring in the United States. Additional information was provided by the manufacturer, allowing calculation of implant rates and durations. RESULTS: An estimated 3283 patients underwent magnetic sphincter augmentation (165 surgeons at 191 institutions). The median implant duration was 1.4 years, with 1016 patients implanted for at least 2 years. No deaths, life-threatening events, or device malfunctions were reported. The overall rate of device removal was 2.7% (89/3283). The most common reasons for device removal were dysphagia (52/89) and persistent reflux symptoms (19/89). Removal for erosion and migration was 0.15% (5/3283) and 0% (0/3283), respectively. There were no perforations. Of the device removals, 57.3% (51/89) occurred <1 year after implant, 30.3% (27/89) between 1 and 2 years, and 12.4% (11/89) >2 years after implant. The rate of device removal and erosion with an implant duration >2 years were 1.1% (11/1016) and 0.1% (1/1016), respectively. All device removals and erosions were managed nonemergently, with no complications or long-term consequences. CONCLUSIONS: During a 4-year period in more than 3000 patients, no unanticipated MSAD complications have emerged, and there is no data to suggest a trend of increased events over time. The presentation and management of device-related issues have been less complicated than revisions for laparoscopic fundoplication or other interventions for GERD. MSAD is considered safe for the widespread treatment of GERD.

The impact of neoadjuvant therapy for gastroesophageal adenocarcinoma on postoperative morbidity and mortality.

BACKGROUND AND OBJECTIVES: We sought to study the impact of neoadjuvant therapy (NAT) on postoperative complications following surgical resection of adenocarcinomas of the stomach and gastroesophageal junction (GEJ). METHODS: We compared the postoperative outcomes of 308 patients undergoing a surgery-first approach and 145 patients undergoing NAT followed by curative-intent surgery for adenocarcinomas of the stomach and GEJ from 1995-2014. RESULTS: Patients receiving NAT were more likely to be younger, have tumors of the GEJ, to undergo esophagogastrectomy and D2 lymphadenectomy, and to have more advanced stage disease than patients undergoing surgery first. There were no differences in overall 30-day morbidity or mortality rates between the groups, yet patients undergoing surgery first were more likely to have higher-grade complications than those undergoing NAT. Age >65 years, higher ASA score, concomitant splenectomy, more advanced tumor stage, and year of surgery were independent risk factors for postoperative morbidity, but receipt of NAT was not an independent predictor of postoperative morbidity. CONCLUSIONS: Despite having more advanced disease and undergoing higher-risk surgical procedures, patients with adenocarcinomas of the stomach or GEJ who receive NAT prior to surgery are no more likely to suffer postoperative complications than patients treated with a surgery-first approach. J. Surg. Oncol. 2016;113:560-564. © 2016 Wiley Periodicals, Inc.

Adjuvant Therapy Completion Rates in Patients with Gastric Cancer Undergoing Perioperative Chemotherapy Versus a Surgery-First Approach.

Delayed recovery after gastrectomy may preclude the administration of adjuvant therapy in a significant percentage of patients who undergo elective gastrectomy as the initial therapy for gastric cancer. Clinicopathologic and treatment variables of 155 patients undergoing potentially curative gastrectomy for stages Ib-IIIc gastric adenocarcinoma from 2001 to 2014 were analyzed, and rates of receipt of chemotherapy and radiotherapy in patients treated with either a surgery-first approach (SURG) or neoadjuvant therapy followed by surgery followed by postoperative therapy (PERIOP) were compared. SURG patients (n = 93) were older and more likely to have distal tumors and to undergo distal gastrectomy and D1 lymphadenectomy than PERIOP patients (n = 62). The distribution of ASA scores was similar between groups. SURG patients were less likely than PERIOP patients to complete at least one cycle of chemotherapy (56 vs 100%, P = 0.001) and all recommended chemotherapy and radiation therapy (44 vs 66%, P = 0.013). These findings were consistent for SURG patients treated during different time periods throughout the study and for patients of poorer performance status. A significantly higher percentage of gastric cancer patients treated with perioperative chemotherapy receive some or all of the recommended components of multimodality therapy than patients treated with a surgery-first approach.

Predictors of Lymph Node Metastasis in Western Early Gastric Cancer.

BACKGROUND: The application of endoscopic and local resection for early gastric cancer (EGC) is limited by the risk of regional lymph node (LN) metastasis. We sought to determine the incidence and predictors of LN metastasis in a contemporary cohort of Western patients with early gastric cancer. METHODS: Sixty-seven patients with pT1 gastric adenocarcinoma underwent radical surgery without neoadjuvant therapy at our institution between 1995 and 2011, and clinicopathologic factors predicting LN metastasis were analyzed. RESULTS: LN metastases were present in 15/67 (22 %) pT1 tumors, including 1/23 (4 %) T1a tumors and 14/44 (32 %) T1b tumors. Tumor size, site, degree of differentiation, macroscopic tumor sub-classification, perineural invasion status, and depth of submucosal tumor penetration did not predict LN metastasis. The presence of lymphovascular invasion (LVI) and positive nodal status by endoscopic ultrasound (EUS) were the only factors that predicted LN metastasis on multivariate analysis. T1a tumors without LVI had a 0 % rate of positive LN, whereas T1b tumors with LVI had a 64.3 % rate of positive LN. CONCLUSIONS: EGC limited to the mucosa, without evidence of LVI, and N0 on EUS, may be considered for limited resection. However, any EGC with submucosal invasion, LVI, or positive nodes on EUS should undergo radical resection with lymphadenectomy.

Evaluation of the LINX antireflux procedure.

PURPOSE OF REVIEW: To evaluate the current data on the safety, efficacy, and indications for magnetic sphincter augmentation (MSA) using the LINX device to treat gastroesophageal reflux disease (GERD). RECENT FINDINGS: The LINX device has demonstrated excellent safety and GERD efficacy in several recent nonblinded, single arm studies with strict inclusion criteria and up to 3 years follow-up. Dysphagia has been the most common adverse effect occurring after LINX. Other gastrointestinal side-effects seen after laparoscopic fundoplication (bloating, gas, and inability to belch) may be less common after LINX. SUMMARY: The LINX device is a safe, well tolerated, and effective therapy for GERD in the short term. MSA should be considered for selected GERD patients without significant anatomic or motility defects. However, the long-term safety and efficacy of LINX - both alone and in comparison to current GERD therapies - remains to be determined.

The influence of anesthesia on heart rate complexity during elective and urgent surgery in 128 patients.

BACKGROUND: As an emerging "new vital sign," heart rate complexity (by sample entropy [SampEn]) has been shown to be a useful trauma triage tool by predicting occult physiologic compromise and need for life-saving interventions. Sample entropy may be confounded by anesthesia possibly limiting its value intraoperatively. We investigated the effects of anesthesia on SampEn during elective and urgent surgical procedures. We hypothesized that SampEn is reduced by general anesthesia. METHODS: With institutional review board-approved waiver of informed consent, 128 patients undergoing elective or urgent general surgery were prospectively enrolled. Real-time heart rate complexity was calculated using SampEn through electrocardiogram recordings of 200 consecutive beats in a continuous sliding-window fashion. We recorded SampEn starting 10 minutes before induction until 10 minutes after emergence from anesthesia. The time before induction of anesthesia was categorized as period 1, the time after induction and before emergence as period 2 (intraoperative), and the time after emergence as period 3. We analyzed SampEn changes as patients moved between the different periods and made 3 comparisons: from period 1 with period 2 (comparison A), from period 2 with period 3 (comparison B). We also compared period 1 with period 3 SampEn (comparison C). RESULTS: The mean SampEn value for all patients before induction of anesthesia was 1.55 ± 0.58. In each 1 of the 3, comparisons there was a decline in SampEn. Comparison A had a mean decrease of 0.53 ± 0.55 (P < .0001), comparison B had a decrease of 0.13 ± 0.52 (P < .0051), and the mean SampEn difference for comparison C was 0.66 ± 0.53 (P < .0001). Certain pharmacologics had significant effect on SampEn as did need for urgent surgery and American Society of Anesthesiologists class. CONCLUSION: Sample entropy decreases after induction of anesthesia and continues to decrease even immediately after emergence in patients without any immediately life-threatening conditions. This finding may complicate interpretation low complexity as a predictor of life-saving interventions in patients in the perioperative period.