Search

Objective: To develop a deep learning algorithm for anatomy recognition in thoracoscopic video frames from robot-assisted minimally invasive esophagectomy (RAMIE) procedures using deep learning. Background: RAMIE is a complex operation with substantial perioperative morbidity and a considerable learning curve. Automatic anatomy recognition may improve surgical orientation and recognition of anatomical structures and might contribute to reducing morbidity or learning curves. Studies regarding anatomy recognition in complex surgical procedures are currently lacking. Methods: Eighty-three videos of consecutive RAMIE procedures between 2018 and 2022 were retrospectively collected at University Medical Center Utrecht. A surgical PhD candidate and an expert surgeon annotated the azygos vein and vena cava, aorta, and right lung on 1050 thoracoscopic frames. 850 frames were used for training of a convolutional neural network (CNN) to segment the anatomical structures. The remaining 200 frames of the dataset were used for testing the CNN. The Dice and 95% Hausdorff distance (95HD) were calculated to assess algorithm accuracy. Results: The median Dice of the algorithm was 0.79 (IQR = 0.20) for segmentation of the azygos vein and/or vena cava. A median Dice coefficient of 0.74 (IQR = 0.86) and 0.89 (IQR = 0.30) were obtained for segmentation of the aorta and lung, respectively. Inference time was 0.026 s (39 Hz). The prediction of the deep learning algorithm was compared with the expert surgeon annotations, showing an accuracy measured in median Dice of 0.70 (IQR = 0.19), 0.88 (IQR = 0.07), and 0.90 (0.10) for the vena cava and/or azygos vein, aorta, and lung, respectively. Conclusion: This study shows that deep learning-based semantic segmentation has potential for anatomy recognition in RAMIE video frames. The inference time of the algorithm facilitated real-time anatomy recognition. Clinical applicability should be assessed in prospective clinical studies.

Indocyanine green fluorescence angiography (ICG-FA) allows for real-time intraoperative assessment of the perfusion of the gastric conduit during esophagectomy. The aim of this study was to investigate the effect of the implementation of ICG-FA during robot-assisted minimally invasive esophagectomy (RAMIE) with an intrathoracic anastomosis. In this prospective cohort study, a standardized protocol for ICG-FA was implemented in a high-volume center in December 2018. All consecutive patients who underwent RAMIE with an intrathoracic anastomosis were included. The primary outcome was whether the initial chosen site for the anastomosis on the gastric conduit was changed based on ICG-FA findings. In addition, ICG-FA was quantified based on the procedural videos. Out of the 63 included patients, the planned location of the anastomosis was changed in 9 (14%) patients, based on ICG-FA. The median time to maximum intensity at the base of the gastric conduit was shorter (25 s; range 13–49) compared to tip (34 s; range 12–83). In patients with anastomotic leakage, the median time to reach the FImax at the tip was 56 s (range 30–83) compared to 34 s (range 12–66) in patients without anastomotic leakage (p= 0.320). The use of ICG-FA resulted in an adaptation of the anastomotic site in nine (14%) patients during RAMIE with intrathoracic anastomosis. The quantification of ICG-FA showed that the gastric conduit reaches it maximum intensity in a base-to-tip direction. Perfusion of the entire gastric conduit was worse for patients with anastomotic leakage, although not statistically different.

Background: Evidence on the added value of robotic-assistance in the abdominal phase during esophagectomy is scarce. In 2003, our center implemented the robotic thoracic phase for esophagectomy. In November 2018 the robot was also implemented in the abdominal phase. The aim of this study was to evaluate the implementation of the abdominal phase during robot-assisted minimally invasive esophagectomy (RAMIE). Methods: Consecutive patients who underwent full RAMIE with intrathoracic anastomosis for esophageal cancer were included. Patients were extracted from a prospectively maintained institutional database. A cumulative sum (CUSUM) analysis was performed for abdominal operation time and abdominal lymph node yield. Intraoperative, postoperative and oncological outcomes including collected lymph nodes per abdominal lymph node station were reported. Results: Between 2018 and 2021, 70 consecutive patients were included. The majority of the patients had an adenocarcinoma (n= 55, 77%) and underwent neoadjuvant chemo(radio)therapy (n= 65, 95%). The median operative time for the abdominal phase was 180 min (range 110–233). The CUSUM analysis for abdominal operation time showed a plateau at case 22. There were no intraoperative complications or conversions during the abdominal phase. The most common postoperative complications were pneumonia (n= 18, 26%) and anastomotic leakage (n= 14, 20%). Radical resection margins were achieved in 69 (99%) patients. The median total lymph node yield was 42 (range 23–83) and the median abdominal lymph node yield was 16 (range 2–43). The CUSUM analysis for abdominal lymph node yield showed a plateau at case 21. Most abdominal lymph nodes were collected from the left gastric artery (median 4, range 0–20). Conclusions: This study shows that a robotic abdominal phase was safely implemented for RAMIE without compromising intraoperative, postoperative and oncological outcomes. The learning curve is estimated to be 22 cases in a high-volume center with experienced upper GI robotic surgeons.

Background: Minimally invasive surgery is complex and associated with substantial learning curves. Computer-aided anatomy recognition, such as artificial intelligence-based algorithms, may improve anatomical orientation, prevent tissue injury, and improve learning curves. The study objective was to provide a comprehensive overview of current literature on the accuracy of anatomy recognition algorithms in intrathoracic and -abdominal surgery. Methods: This systematic review is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline. Pubmed, Embase, and IEEE Xplore were searched for original studies up until January 2022 on computer-aided anatomy recognition, without requiring intraoperative imaging or calibration equipment. Extracted features included surgical procedure, study population and design, algorithm type, pre-training methods, pre- and post-processing methods, data augmentation, anatomy annotation, training data, testing data, model validation strategy, goal of the algorithm, target anatomical structure, accuracy, and inference time. Results: After full-text screening, 23 out of 7124 articles were included. Included studies showed a wide diversity, with six possible recognition tasks in 15 different surgical procedures, and 14 different accuracy measures used. Risk of bias in the included studies was high, especially regarding patient selection and annotation of the reference standard. Dice and intersection over union (IoU) scores of the algorithms ranged from 0.50 to 0.98 and from 74 to 98%, respectively, for various anatomy recognition tasks. High-accuracy algorithms were typically trained using larger datasets annotated by expert surgeons and focused on less-complex anatomy. Some of the high-accuracy algorithms were developed using pre-training and data augmentation. Conclusions: The accuracy of included anatomy recognition algorithms varied substantially, ranging from moderate to good. Solid comparison between algorithms was complicated by the wide variety of applied methodology, target anatomical structures, and reported accuracy measures. Computer-aided intraoperative anatomy recognition is an upcoming research discipline, but still at its infancy. Larger datasets and methodological guidelines are required to improve accuracy and clinical applicability in future research.

Extra-regional lymph node metastases strongly determine treatment options in patients with esophageal cancer. Staging modalities such as (FDG-PET) CT scanning frequently show activity in retroperitoneal and lung hilar lymph nodes. This study evaluated the incidence of histologically confirmed metastases, treatment approach and recurrence patterns in patients with (FDG-PET) CT positivity in these regions.

Abstract: Background: Although studies have shown that EUS has a high sensitivity and specificity for T and N staging, the value of EUS for staging tumors as resectable or nonresectable after CT of the chest and abdomen and US neck assessment, is largely unknown. This study was designed to assess the diagnostic value of EUS for determining resectability of esophageal cancer. Methods: A retrospective analysis of all consecutive patients with esophageal carcinoma who underwent staging EUS, CT, and US. Tumors were considered resectable when there was no evidence of metastases or ingrowth in adjacent structures. Sensitivity, specificity, positive (PPV), and negative (NPV) predictive value of CT/US neck and CT/US neck + EUS for predicting surgical resectability were calculated. PPVs of CT/US alone and CT/US + EUS together were compared for assessing the diagnostic value of EUS. Results: In total, 211 patients (155 men; mean age of 64 ± 9.4 years) were included, of which 176 (83 %) underwent all three staging investigations. Based on preoperative staging, 173 (82 %) patients were considered resectable and 38 (18 %) nonresectable. Of all 173 initially resectable patients, 145 were operated on. Of these patients, five (3.4 %) tumors were found nonresectable during surgery. Postoperative sensitivity, specificity, PPV, and NPV of CT/US and CT/US + EUS for predicting surgical resectability were 88 versus 87 %, 20 versus 40 %, 97 versus 98 %, and 6 versus 10 %, respectively. Conclusions: Although EUS adds to the specificity of preoperative esophageal cancer staging after CT chest and abdomen, and US of the neck have been performed, the overall added value of EUS is limited.

Background: The need for routine axillary lymph node dissection (ALND) in patients with invasive breast cancer and low-volume sentinel node (SN) involvement is questionable. Accurate prediction of second echelon lymph node involvement could identify those patients most likely to benefit from ALND.Methods: A consecutive series of 317 patients with invasive breast cancer and a tumor positive axillary SN followed by ALND was reviewed. Clinicopathologic features of the primary tumor and the SN were assessed as possible predictors of second echelon lymph node involvement.Results: Second echelon metastases were found in 116317 cases (36.6%). Frequency of second echelon lymph node involvement in patients with isolated tumor cells (ITC, N23), micro- (N101) and macrometastases (N193) was 13%, 20% and 48%, respectively (p<0.001). Based on the area % of SN occupied by tumor no subgroup of patients could be selected with less than 20% second echelon lymph node involvement. However, none of the patients with SN ITC or micrometastases and a primary tumor size ≤1 cm (N12, 3.8%) had second echelon lymph node involvement.Conclusions: Accurately measured SN tumor load predicts second echelon lymph node involvement. However, even in patients with ITC, the second echelon lymph nodes are involved in 13% justifying ALND.

Transthoracic esophagectomy with extended lymph node dissection is associated with higher morbidity rates than transhiatal esophagectomy. This morbidity rate could be reduced by the use of minimally invasive techniques. The feasibility of robot-assisted thoracoscopic esophagectomy (RTE) with mediastinal lymphadenectomy was assessed prospectively.This study investigated 21 consecutive patients with esophageal cancer who underwent RTE using the Da Vinci™ robotic system. Continuity was restored with a gastric conduit and a cervical anastomosis.A total of 18 (86%) procedures were completed thoracoscopically. The operating time for the thoracoscopic phase was 180 min (range, 120–240 min), and the median blood loss was 400 ml (range, 150–700 ml). A median of 20 (range, 9–30) lymph nodes were retrieved. The median intensive care unit stay was 4 days (range, 1–129 days), and the hospital stay was 18 days (range, 11–182 days). Pulmonary complications occurred in 10 patients (48%), and one patient (5%) died of a tracheoneoesophageal fistula.In this initial experience, robot-assisted thoracoscopic esophagectomy was found to be feasible, providing an effective lymphadenectomy with low blood loss. Standardization of the technique and increased experience should reduce the complication rate, which is in the range of the rate for open transthoracic dissection.

Abstract: . The effect of fractionating the 633 nm illumination of 5-aminolaevulinic (ALA)-based photodynamic therapy (PDT) of the normal rat oesophagus was studied. Fractionation of the illumination could enhance the PDT effect in two ways: (a) delay of the vascular shutdown or relaxation of the vasoconstriction induced by ALA-PDT and (b) use of newly formed protoporphyrin IX (PpIX), produced during the dark interval. Forty rats were randomly allocated to two groups of 20 animals each. To study vascular effects, in group 1 illumination with 633 nm (100 mW/cm) was performed at 3 h after oral ALA administration (200 mg/kg) either continuously with 20 J/cm diffuser length (n=5) or fractionated 2×10 J/cm with a 150 s interval (n=5), five animals served as controls. Blood flow was measured with a laser Doppler flowmeter. To study the effect of renewed PpIX forming, animals in group 2 were illuminated continuously at 3 h after ALA with 20 J/cm (n=5) or 40 J/cm (n=5) or fractionated 2×20 J/cm with a 3 h interval (n=5), five animals served as controls. In all animals the in vivo fluence rate and PpIX fluorescence were measured during illuminations and animals were killed at 48 h after PDT. ALA-PDT did not cause any significant vasoconstriction. Fluorescence measurements and dosimetric results in group 1 did not differ between animals illuminated continuously or fractionated with a 150 s interval. In group 2, during a 3 h dark interval, PpIX fluorescence increased and was bleached during the second illumination. The tissue optical properties changed during the 3 h dark interval, resulting in a lower in vivo fluence rate (p≤0.001). Fractionation did not result in more oesophageal damage. It was concluded that a 150 s interval during illumination in ALA-PDT does not increase oesophageal blood flow. During an interval of 3 h new PpIX is formed. In the present study, fractionated illumination using short or long time intervals did not result in more damage. Thus, this study shows no evidence for improved PDT effect with fractionated light delivery.

Abstract: . Simultaneous application of multiple fibres could increase the volume of coagulation produced with interstitial laser photocoagulation (ILP) for solid tumours. To take full advantage of the presumed synergistic thermal effect between the fibres, the optimal combination of laser power and distance between the fibres was investigated. Four fibres with a cylindrical diffusing tip of 2 cm length were used, coupled to an optical beamsplitter for Nd:YAG light (four channels, maximal variation 9.5%, transmission >85%). The distance between the fibres was 1, 1.5, 2, 2.5 , 3 or 4 cm with a power output of either 4, 5, 6 or 7 W/fibre; energy per fibre was constant at 1800 J by adjusting exposure time. After laser application, dimensions of the coagulated lesions were measured. The optimal mutual fibre distance was 2 cm ( p<0.01) at all power levels. This resulted in lesions with a mean (SD) volume of 44.5 (2.1) cm3 and a largest diameter of 5.1 (0.4) cm at 7 W/fibre. Smaller distances between the fibres resulted in smaller lesions with central carbonisation, whereas larger distances resulted in four separate zones of coagulation. It was concluded that simultaneous application of four interstitial fibres may result in a considerable increase of volume of coagulation. Fibre position and mutual fibre distance determines whether synergism of the coagulative effect occurs.

Background Interstitial laser coagulation is a local treatment modality for solid tumours, such as liver metastases. Perfusion by the vascular system can transfer heat from the treatment site, thereby reducing the volume of destroyed tissue.
Methods This effect was investigated in a pig model (n=10). Lesions were produced in the liver with four interstitial laser fibres (5 W per fibre for 6 min): (1) with flow through the liver (n=9); (2) with the portal vein clamped (n=8); and (3) with both hepatic artery and portal vein clamped (n=6). The pigs were killed and the liver was dissected; lesions were assessed macroscopically and microscopically.
Results The volume of the lesion was significantly larger when portal flow was eliminated (30·6 versus 6·4 cm3, P&lt;0·001).
Conclusion Hepatic blood flow substantially reduces the size of the lesion produced by laser coagulation. Portal flow should therefore be occluded during laser treatment to produce lesions of clinically relevant dimensions.

Photodynamic therapy is based on the administration of a compound that is preferentially accumulated by a tumor and which causes tumor destruction after exposure to light of a specific wavelength. The photosensitizers most commonly used in treating tumors of the gastrointestinal tract are porphyrins - hematoporphyrin derivative and dihematoporphyrin ether. These compounds have been used with success to produce reduction in tumor size of esophageal, gastric, and colorectal cancers. In some instances long-lasting complete remissions have been observed after photodynamic therapy. New developments include photosensitizers that react to light of a longer wavelength, which is able to penetrate tissue to a greater depth, the use of 5-aminolevulinic acid, which is preferentially converted to porphyrin in malignant cells, and combination of photodynamic therapy with thermic laser, radiotherapy, or chemotherapy.

Administration of 5-aminolaevulinic acid (ALA) leads to porphyrin accumulation in malignant and premalignant tissues, and ALA is used as a prodrug in photodynamic therapy (PDT). To understand the mechanism of porphyrin accumulation after the administration of ALA and to investigate whether ALA-induced protoporphyrin IX might be a suitable photosensitizer in Barrett's oesophagus and adenocarcinoma, we determined the activities of porphobilinogen deaminase (PBG-D) and ferrochelatase (FC) in various malignant and premalignant as well as in normal tissues of the human oesophagus. A PDT power index for ALA-induced porphyrin accumulation, the ratio of PBG-D to FC normalized for the normal squamous epithelium of the oesophagus, was calculated to evaluate intertissue variation in the ability to accumulate porphyrins. In malignant and premalignant tissue a twofold increased PBG-D activity and a marginally increased FC activity was seen compared with normal squamous epithelium. A significantly increased PDT power index in Barrett's epithelium and adenocarcinoma was found. Our results suggest that, after the administration of ALA, porphyrins will accumulate in a greater amount in Barrett's epithelium and adenocarcinoma of the oesophagus because of an imbalance between PBG-D and FC activities. The PDT power index here defined might be a useful indicative parameter for predicting the susceptibility of these tissues to ALA-PDT.

The use of photodynamic therapy (PDT) as an adjunct to curative tumour resection was investigated in a tumour recurrence model, using rat mammary adenocarcinoma BN472. Tumours were inoculated subcutaneously in 60 animals and resected after 21 days of growth. Immediately after removal, the operation site was exposed to 320-450 nm light of 0.1 W cm-2 and 60 J cm-2 after photosensitisation with either Photofrin (5 mg kg-1 i.v. 48 h before illumination) or 5-aminolaevulinic acid (ALA) (2 mg ml-1 in drinking water for 9 days). Porphyrin concentrations were measured in tissue samples. After 28 days, animals treated with adjunctive PDT had a significantly longer tumour-free interval than controls (P < 0.01); median 25 days (Photofrin), 18 days (ALA), 14 days (controls). Moreover, in the PDT groups significantly fewer rats had lymph node metastasis. A prophyrin concentration ratio between tumour and mammary tissue of 2:1 was found after Photofrin and 4:1 after ALA. The results indicate that adjuvant intraoperative PDT may be a safe and effective method of destroying residual tumour, thereby preventing locoregional tumour recurrence.

Photodynamic therapy (PDT) of hepatic tumours has been restricted owing to the preferential retention of photosensitizers in liver tissue. We therefore investigated interstitial tumour illumination as a means of selective PDT. A piece of colon carcinoma CC531 was implanted in the liver of Wag/Rij rats. Photofrin was administered (5 mg kg-1 i.v.) 2 days before laser illumination. Tumours with a mean (+/- s.e.) diameter of 5.7 +/- 0.1 mm (n = 106, 20 days after implantation) were illuminated with 625 nm light, at 200 mW cm-1 from a 0.5 cm cylindrical diffuser and either 100, 200, 400, 800 or 1600 J cm-1. Control groups received either laser illumination only, Photofrin only or diffuser insertion only. Short-term effects were studied on the second day after illumination by light microscopy and computer-assisted integration of the circumference of damaged areas. Long-term effects were studied on day 36. To determine the biochemistry of liver damage and function, serum ASAT and ALAT levels were measured on day 1 and 2, and antipyrine clearance on day 1. Tumour and surrounding liver necrosis increased with light dose delivered (P < 0.001). Best long-term results were obtained at 800 J cm-1 with complete tumour remission in 4 out of 6 animals. No deterioration in liver function was found. The results of this study show the ability of interstitial PDT to cause major destruction of tumour tissue in the liver combined with minimal liver damage.

Background: The need for routine axillary lymph node dissection (ALND) in patients with invasive breast cancer and low-volume sentinel node (SN) involvement is questionable. Accurate prediction of second echelon lymph node involvement could identify those patients most likely to benefit from ALND.Methods: A consecutive series of 317 patients with invasive breast cancer and a tumor positive axillary SN followed by ALND was reviewed. Clinicopathologic features of the primary tumor and the SN were assessed as possible predictors of second echelon lymph node involvement.Results: Second echelon metastases were found in 116/317 cases (36.6%). Frequency of second echelon lymph node involvement in patients with isolated tumor cells (ITC, N = 23), micro- (N = 101) and macrometastases (N = 193) was 13%, 20% and 48%, respectively (p < 0.001). Based on the area % of SN occupied by tumor no subgroup of patients could be selected with less than 20% second echelon lymph node involvement. However, none of the patients with SN ITC or micrometastases and a primary tumor size ≤1 cm (N = 12, 3.8%) had second echelon lymph node involvement.Conclusions: Accurately measured SN tumor load predicts second echelon lymph node involvement. However, even in patients with ITC, the second echelon lymph nodes are involved in 13% justifying ALND.