TEsoNet: knowledge transfer in surgical phase recognition from laparoscopic sleeve gastrectomy to the laparoscopic part of Ivor-Lewis esophagectomy.

BACKGROUND: Surgical phase recognition using computer vision presents an essential requirement for artificial intelligence-assisted analysis of surgical workflow. Its performance is heavily dependent on large amounts of annotated video data, which remain a limited resource, especially concerning highly specialized procedures. Knowledge transfer from common to more complex procedures can promote data efficiency. Phase recognition models trained on large, readily available datasets may be extrapolated and transferred to smaller datasets of different procedures to improve generalizability. The conditions under which transfer learning is appropriate and feasible remain to be established. METHODS: We defined ten operative phases for the laparoscopic part of Ivor-Lewis Esophagectomy through expert consensus. A dataset of 40 videos was annotated accordingly. The knowledge transfer capability of an established model architecture for phase recognition (CNN + LSTM) was adapted to generate a "Transferal Esophagectomy Network" (TEsoNet) for co-training and transfer learning from laparoscopic Sleeve Gastrectomy to the laparoscopic part of Ivor-Lewis Esophagectomy, exploring different training set compositions and training weights. RESULTS: The explored model architecture is capable of accurate phase detection in complex procedures, such as Esophagectomy, even with low quantities of training data. Knowledge transfer between two upper gastrointestinal procedures is feasible and achieves reasonable accuracy with respect to operative phases with high procedural overlap. CONCLUSION: Robust phase recognition models can achieve reasonable yet phase-specific accuracy through transfer learning and co-training between two related procedures, even when exposed to small amounts of training data of the target procedure. Further exploration is required to determine appropriate data amounts, key characteristics of the training procedure and temporal annotation methods required for successful transferal phase recognition. Transfer learning across different procedures addressing small datasets may increase data efficiency. Finally, to enable the surgical application of AI for intraoperative risk mitigation, coverage of rare, specialized procedures needs to be explored.

Simultaneous laryngopharyngeal pH monitoring (Restech) and conventional esophageal pH monitoring-correlation using a large patient cohort of more than 100 patients with suspected gastroesophageal reflux disease.

24-hour esophageal pH-metry is not designed to detect laryngopharyngeal reflux (LPR). The new laryngopharyngeal pH-monitoring system (Restech) may detect LPR better. There is no established correlation between these two techniques as only small case series exist. The aim of this study is to examine the correlation between the two techniques with a large patient cohort. All patients received a complete diagnostic workup for gastroesophageal reflux including symptom evaluation, endoscopy, 24-hour pH-metry, high resolution manometry, and Restech. Consecutive patients with suspected gastroesophageal reflux and disease-related extra-esophageal symptoms were evaluated using 24-hour laryngopharyngeal and concomitant esophageal pH-monitoring. Subsequently, the relationship between the two techniques was evaluated subdividing the different reflux scenarios into four groups. A total of 101 patients from December 2013 to February 2017 were included. All patients presented extra-esophageal symptoms such as cough, hoarseness, asthma symptoms, and globus sensation. Classical reflux symptoms such as heartburn (71%), regurgitation (60%), retrosternal pain (54%), and dysphagia (32%) were also present. Esophageal 24-hour pH-metry was positive in 66 patients (65%) with a mean DeMeester Score of 66.7 [15-292]. Four different reflux scenarios were detected (group A-D): in 39% of patients with abnormal esophageal pH-metry, Restech evaluation was normal (group A, n = 26, mean DeMeester-score = 57.9 [15-255], mean Ryan score = 2.6 [2-8]). In 23% of patients with normal pH-metry (n = 8, group B), Restech evaluation was abnormal (mean DeMeester-score 10.5 [5-13], mean Ryan score 63.5 [27-84]). The remaining groups C and D showed corresponding results. Restech evaluation was positive in 48% of cases in this highly selective patient cohort. As demonstrated by four reflux scenarios, esophageal pH-metry and Restech do not necessarily need to correspond. Especially in patients with borderline abnormal 24-hour pH-metry, Restech may help to support the decision for or against laparoscopic anti-reflux surgery.