Challenges in multi-centric generalization: phase and step recognition in Roux-en-Y gastric bypass surgery

PURPOSE

Most studies on surgical activity recognition utilizing artificial intelligence (AI) have focused mainly on recognizing one type of activity from small and mono-centric surgical video datasets. It remains speculative whether those models would generalize to other centers.

METHODS

In this work, we introduce a large multi-centric multi-activity dataset consisting of 140 surgical videos (MultiBypass140) of laparoscopic Roux-en-Y gastric bypass (LRYGB) surgeries performed at two medical centers, i.e., the University Hospital of Strasbourg, France (StrasBypass70) and Inselspital, Bern University Hospital, Switzerland (BernBypass70). The dataset has been fully annotated with phases and steps by two board-certified surgeons. Furthermore, we assess the generalizability and benchmark different deep learning models for the task of phase and step recognition in 7 experimental studies: (1) Training and evaluation on BernBypass70; (2) Training and evaluation on StrasBypass70; (3) Training and evaluation on the joint MultiBypass140 dataset; (4) Training on BernBypass70, evaluation on StrasBypass70; (5) Training on StrasBypass70, evaluation on BernBypass70; Training on MultiBypass140, (6) evaluation on BernBypass70 and (7) evaluation on StrasBypass70.

RESULTS

The model's performance is markedly influenced by the training data. The worst results were obtained in experiments (4) and (5) confirming the limited generalization capabilities of models trained on mono-centric data. The use of multi-centric training data, experiments (6) and (7), improves the generalization capabilities of the models, bringing them beyond the level of independent mono-centric training and validation (experiments (1) and (2)).

CONCLUSION

MultiBypass140 shows considerable variation in surgical technique and workflow of LRYGB procedures between centers. Therefore, generalization experiments demonstrate a remarkable difference in model performance. These results highlight the importance of multi-centric datasets for AI model generalization to account for variance in surgical technique and workflows. The dataset and code are publicly available at https://github.com/CAMMA-public/MultiBypass140.

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.

Artificial intelligence software available for medical devices: surgical phase recognition in laparoscopic cholecystectomy

Background

Surgical process modeling automatically identifies surgical phases, and further improvement in recognition accuracy is expected with deep learning. Surgical tool or time series information has been used to improve the recognition accuracy of a model. However, it is difficult to collect this information continuously intraoperatively. The present study aimed to develop a deep convolution neural network (CNN) model that correctly identifies the surgical phase during laparoscopic cholecystectomy (LC).

Methods

We divided LC into six surgical phases (P1–P6) and one redundant phase (P0). We prepared 115 LC videos and converted them to image frames at 3 fps. Three experienced doctors labeled the surgical phases in all image frames. Our deep CNN model was trained with 106 of the 115 annotation datasets and was evaluated with the remaining datasets. By depending on both the prediction probability and frequency for a certain period, we aimed for highly accurate surgical phase recognition in the operation room.

Results

Nine full LC videos were converted into image frames and were fed to our deep CNN model. The average accuracy, precision, and recall were 0.970, 0.855, and 0.863, respectively.

Conclusion

The deep learning CNN model in this study successfully identified both the six surgical phases and the redundant phase, P0, which may increase the versatility of the surgical process recognition model for clinical use. We believe that this model can be used in artificial intelligence for medical devices. The degree of recognition accuracy is expected to improve with developments in advanced deep learning algorithms.

Real-time medical phase recognition using long-term video understanding and progress gate method

We introduce a real-time system for recognizing five phases of the trauma resuscitation process, the initial management of injured patients in the emergency department. We used depth videos as input to preserve the privacy of the patients and providers. The depth videos were recorded using a Kinect-v2 mounted on the sidewall of the room. Our dataset consisted of 183 depth videos of trauma resuscitations. The model was trained on 150 cases with more than 30 minutes each and tested on the remaining 33 cases. We introduced a reduced long-term operation (RLO) method for extracting features from long segments of video and combined it with the regular model having short-term information only. The model with RLO outperformed the regular short-term model by 5% using the accuracy score. We also introduced a progress gate (PG) method to distinguish visually similar phases using video progress. The final system achieved 91% accuracy and significantly outperformed previous systems for phase recognition in this setting.

Assisted phase and step annotation for surgical videos

PURPOSE

Annotation of surgical videos is a time-consuming task which requires specific knowledge. In this paper, we present and evaluate a deep learning-based method that includes pre-annotation of the phases and steps in surgical videos and user assistance in the annotation process.

METHODS

We propose a classification function that automatically detects errors and infers temporal coherence in predictions made by a convolutional neural network. First, we trained three different architectures of neural networks to assess the method on two surgical procedures: cholecystectomy and cataract surgery. The proposed method was then implemented in an annotation software to test its ability to assist surgical video annotation. A user study was conducted to validate our approach, in which participants had to annotate the phases and the steps of a cataract surgery video. The annotation and the completion time were recorded.

RESULTS

The participants who used the assistance system were 7% more accurate on the step annotation and 10 min faster than the participants who used the manual system. The results of the questionnaire showed that the assistance system did not disturb the participants and did not complicate the task.

CONCLUSION

The annotation process is a difficult and time-consuming task essential to train deep learning algorithms. In this publication, we propose a method to assist the annotation of surgical workflows which was validated through a user study. The proposed assistance system significantly improved annotation duration and accuracy.

Real-time automatic surgical phase recognition in laparoscopic sigmoidectomy using the convolutional neural network-based deep learning approach. Surg Endosc. 2020;34:4924-4931. [PMID: 31797047 DOI: 10.1007/s00464-019-07281-0]

BACKGROUND

Automatic surgical workflow recognition is a key component for developing the context-aware computer-assisted surgery (CA-CAS) systems. However, automatic surgical phase recognition focused on colorectal surgery has not been reported. We aimed to develop a deep learning model for automatic surgical phase recognition based on laparoscopic sigmoidectomy (Lap-S) videos, which could be used for real-time phase recognition, and to clarify the accuracies of the automatic surgical phase and action recognitions using visual information.

METHODS

The dataset used contained 71 cases of Lap-S. The video data were divided into frame units every 1/30 s as static images. Every Lap-S video was manually divided into 11 surgical phases (Phases 0-10) and manually annotated for each surgical action on every frame. The model was generated based on the training data. Validation of the model was performed on a set of unseen test data. Convolutional neural network (CNN)-based deep learning was also used.

RESULTS

The average surgical time was 175 min (± 43 min SD), with the individual surgical phases also showing high variations in the duration between cases. Each surgery started in the first phase (Phase 0) and ended in the last phase (Phase 10), and phase transitions occurred 14 (± 2 SD) times per procedure on an average. The accuracy of the automatic surgical phase recognition was 91.9% and those for the automatic surgical action recognition of extracorporeal action and irrigation were 89.4% and 82.5%, respectively. Moreover, this system could perform real-time automatic surgical phase recognition at 32 fps.

CONCLUSIONS

The CNN-based deep learning approach enabled the recognition of surgical phases and actions in 71 Lap-S cases based on manually annotated data. This system could perform automatic surgical phase recognition and automatic target surgical action recognition with high accuracy. Moreover, this study showed the feasibility of real-time automatic surgical phase recognition with high frame rate.

Surgical phase modelling in minimal invasive surgery

Background

Surgical Process Modelling (SPM) offers the possibility to automatically gain insight in the surgical workflow, with the potential to improve OR logistics and surgical care. Most studies have focussed on phase recognition modelling of the laparoscopic cholecystectomy, because of its standard and frequent execution. To demonstrate the broad applicability of SPM, more diverse and complex procedures need to be studied. The aim of this study is to investigate the accuracy in which we can recognise and extract surgical phases in laparoscopic hysterectomies (LHs) with inherent variability in procedure time. To show the applicability of the approach, the model was used to automatically predict surgical end-times.

Methods

A dataset of 40 video-recorded LHs was manually annotated for instrument use and divided into ten surgical phases. The use of instruments provided the feature input for building a Random Forest surgical phase recognition model that was trained to automatically recognise surgical phases. Tenfold cross-validation was performed to optimise the model for predicting the surgical end-time throughout the procedure.

Results

Average surgery time is 128 ± 27 min. Large variability within specific phases is seen. Overall, the Random Forest model reaches an accuracy of 77% recognising the current phase in the procedure. Six of the phases are predicted accurately over 80% of their duration. When predicting the surgical end-time, on average an error of 16 ± 13 min is reached throughout the procedure.

Conclusions

This study demonstrates an intra-operative approach to recognise surgical phases in 40 laparoscopic hysterectomy cases based on instrument usage data. The model is capable of automatic detection of surgical phases for generation of a solid prediction of the surgical end-time.

Toward cognitive pipelines of medical assistance algorithms

PURPOSE

Assistance algorithms for medical tasks have great potential to support physicians with their daily work. However, medicine is also one of the most demanding domains for computer-based support systems, since medical assistance tasks are complex and the practical experience of the physician is crucial. Recent developments in the area of cognitive computing appear to be well suited to tackle medicine as an application domain.

METHODS

We propose a system based on the idea of cognitive computing and consisting of auto-configurable medical assistance algorithms and their self-adapting combination. The system enables automatic execution of new algorithms, given they are made available as Medical Cognitive Apps and are registered in a central semantic repository. Learning components can be added to the system to optimize the results in the cases when numerous Medical Cognitive Apps are available for the same task. Our prototypical implementation is applied to the areas of surgical phase recognition based on sensor data and image progressing for tumor progression mappings.

RESULTS

Our results suggest that such assistance algorithms can be automatically configured in execution pipelines, candidate results can be automatically scored and combined, and the system can learn from experience. Furthermore, our evaluation shows that the Medical Cognitive Apps are providing the correct results as they did for local execution and run in a reasonable amount of time.

CONCLUSION

The proposed solution is applicable to a variety of medical use cases and effectively supports the automated and self-adaptive configuration of cognitive pipelines based on medical interpretation algorithms.