Surgical Workflow Recognition Research Articles

Temporal Memory Relation Network for Workflow Recognition From Surgical Video.

Automatic surgical workflow recognition is a key component for developing context-aware computer-assisted systems in the operating theatre. Previous works either jointly modeled the spatial features with short fixed-range temporal information, or separately learned visual and long temporal cues. In this paper, we propose a novel end-to-end temporal memory relation network (TMRNet) for relating long-range and multi-scale temporal patterns to augment the present features. We establish a long-range memory bank to serve as a memory cell storing the rich supportive information. Through our designed temporal variation layer, the supportive cues are further enhanced by multi-scale temporal-only convolutions. To effectively incorporate the two types of cues without disturbing the joint learning of spatio-temporal features, we introduce a non-local bank operator to attentively relate the past to the present. In this regard, our TMRNet enables the current feature to view the long-range temporal dependency, as well as tolerate complex temporal extents. We have extensively validated our approach on two benchmark surgical video datasets, M2CAI challenge dataset and Cholec80 dataset. Experimental results demonstrate the outstanding performance of our method, consistently exceeding the state-of-the-art methods by a large margin (e.g., 67.0% v.s. 78.9% Jaccard on Cholec80 dataset).

Impact of data on generalization of AI for surgical intelligence applications

AI is becoming ubiquitous, revolutionizing many aspects of our lives. In surgery, it is still a promise. AI has the potential to improve surgeon performance and impact patient care, from post-operative debrief to real-time decision support. But, how much data is needed by an AI-based system to learn surgical context with high fidelity? To answer this question, we leveraged a large-scale, diverse, cholecystectomy video dataset. We assessed surgical workflow recognition and report a deep learning system, that not only detects surgical phases, but does so with high accuracy and is able to generalize to new settings and unseen medical centers. Our findings provide a solid foundation for translating AI applications from research to practice, ushering in a new era of surgical intelligence.

LRTD: long-range temporal dependency based active learning for surgical workflow recognition.

Automatic surgical workflow recognition in video is an essentially fundamental yet challenging problem for developing computer-assisted and robotic-assisted surgery. Existing approaches with deep learning have achieved remarkable performance on analysis of surgical videos, however, heavily relying on large-scale labelled datasets. Unfortunately, the annotation is not often available in abundance, because it requires the domain knowledge of surgeons. Even for experts, it is very tedious and time-consuming to do a sufficient amount of annotations. In this paper, we propose a novel active learning method for cost-effective surgical video analysis. Specifically, we propose a non-local recurrent convolutional network, which introduces non-local block to capture the long-range temporal dependency (LRTD) among continuous frames. We then formulate an intra-clip dependency score to represent the overall dependency within this clip. By ranking scores among clips in unlabelled data pool, we select the clips with weak dependencies to annotate, which indicates the most informative ones to better benefit network training. We validate our approach on a large surgical video dataset (Cholec80) by performing surgical workflow recognition task. By using our LRTD based selection strategy, we can outperform other state-of-the-art active learning methods who only consider neighbor-frame information. Using only up to 50% of samples, our approach can exceed the performance of full-data training. By modeling the intra-clip dependency, our LRTD based strategy shows stronger capability to select informative video clips for annotation compared with other active learning methods, through the evaluation on a popular public surgical dataset. The results also show the promising potential of our framework for reducing annotation workload in the clinical practice.

Real-time automatic surgical phase recognition in laparoscopic sigmoidectomy using the convolutional neural network-based deep learning approach.

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. The dataset used contained 71 cases of Lap-S. The video data were divided into frame units every 1/30s 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. The average surgical time was 175min (± 43min 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. 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.

Presence Detection of Surgical Tool Via Densely Connected Convolutional Networks

Surgical tool detection is important to surgical workflow recognition. It is considered as an essential task in surgical phase recognition. Recently, Densely Connected Convolutional Networks have gained a huge success in computer vision applications, especially in object detection and image Classification. In this paper, we proposed a method to solve the surgical tool presence detection problem as a multi-label classification problem based on Densely Connected Convolutional Networks. The performance of the proposed method has been evaluated in the surgical tool presence detection challenge dataset held by Modeling and Monitoring of Computer Assisted Interventions workshop. The result shows that our proposed model has achieved significant success in detecting surgical tool and got a mean average precision of 62.9% on the testing data. The technology studied in this paper has broad application prospects in computer-aided surgical systems and is a core component of the artificial intelligence medical operating room in the future.

“Deep-Onto” network for surgical workflow and context recognition

Surgical workflow recognition and context-aware systems could allow better decision making and surgical planning by providing the focused information, which may eventually enhance surgical outcomes. While current developments in computer-assisted surgical systems are mostly focused on recognizing surgical phases, they lack recognition of surgical workflow sequence and other contextual element, e.g., "Instruments." Our study proposes a hybrid approach, i.e., using deep learning and knowledge representation, to facilitate recognition of the surgical workflow. We implemented "Deep-Onto" network, which is an ensemble of deep learning models and knowledge management tools, ontology and production rules. As a prototypical scenario, we chose robot-assisted partial nephrectomy (RAPN). We annotated RAPN videos with surgical entities, e.g., "Step" and so forth. We performed different experiments, including the inter-subject variability, to recognize surgical steps. The corresponding subsequent steps along with other surgical contexts, i.e., "Actions," "Phase" and "Instruments," were also recognized. The system was able to recognize 10 RAPN steps with the prevalence-weighted macro-average (PWMA) recall of 0.83, PWMA precision of 0.74, PWMA F1 score of 0.76, and the accuracy of 74.29% on 9 videos of RAPN. We found that the combined use of deep learning and knowledge representation techniques is a promising approach for the multi-level recognition of RAPN surgical workflow.

Semi-supervised spatio-temporal CNN for recognition of surgical workflow

Robust and automated surgical workflow detection in real time is a core component of the future intelligent operating room. Based on this technology, it can help medical staff to automate and intelligently complete many routine activities during surgery. Recognition of surgical workflow based on traditional pattern recognition methods requires a large number of labeled surgical video data. However, the labeled surgical video data requires expert knowledge and it is difficult and time consuming to collect a sufficient number of labeled surgical video data in the medical field. Therefore, this paper proposes a semi-supervised spatio-temporal convolutional network for the recognition of surgical workflow based on convolutional neural networks and temporal-recursive networks. Firstly, we build a spatial convolutional extraction feature network based on unsupervised generative adversarial learning. Then, we build a bridge between low-level surgical video features and high-level surgical workflow semantics based on an unsupervised temporal-ordered network learning approach. Finally, we use the semi-supervised learning method to integrate the spatial model and the temporal model to fine-tune the network, and realize the intelligent recognition of the surgical workflow at a low cost to efficiently determine the progress of the surgical workflow. We performed some experiments for validating the mode based on m2cai16-workflow dataset. It shows that the proposed model can effectively extract the surgical feature and determine the surgical workflow. The Jaccard score of the model reaches 71.3%, and the accuracy of the model reaches 85.8%.

EndoNet: A Deep Architecture for Recognition Tasks on Laparoscopic Videos.

Surgical workflow recognition has numerous potential medical applications, such as the automatic indexing of surgical video databases and the optimization of real-time operating room scheduling, among others. As a result, surgical phase recognition has been studied in the context of several kinds of surgeries, such as cataract, neurological, and laparoscopic surgeries. In the literature, two types of features are typically used to perform this task: visual features and tool usage signals. However, the used visual features are mostly handcrafted. Furthermore, the tool usage signals are usually collected via a manual annotation process or by using additional equipment. In this paper, we propose a novel method for phase recognition that uses a convolutional neural network (CNN) to automatically learn features from cholecystectomy videos and that relies uniquely on visual information. In previous studies, it has been shown that the tool usage signals can provide valuable information in performing the phase recognition task. Thus, we present a novel CNN architecture, called EndoNet, that is designed to carry out the phase recognition and tool presence detection tasks in a multi-task manner. To the best of our knowledge, this is the first work proposing to use a CNN for multiple recognition tasks on laparoscopic videos. Experimental comparisons to other methods show that EndoNet yields state-of-the-art results for both tasks.

Automatic data-driven real-time segmentation and recognition of surgical workflow.

With the intention of extending the perception and action of surgical staff inside the operating room, the medical community has expressed a growing interest towards context-aware systems. Requiring an accurate identification of the surgical workflow, such systems make use of data from a diverse set of available sensors. In this paper, we propose a fully data-driven and real-time method for segmentation and recognition of surgical phases using a combination of video data and instrument usage signals, exploiting no prior knowledge. We also introduce new validation metrics for assessment of workflow detection. The segmentation and recognition are based on a four-stage process. Firstly, during the learning time, a Surgical Process Model is automatically constructed from data annotations to guide the following process. Secondly, data samples are described using a combination of low-level visual cues and instrument information. Then, in the third stage, these descriptions are employed to train a set of AdaBoost classifiers capable of distinguishing one surgical phase from others. Finally, AdaBoost responses are used as input to a Hidden semi-Markov Model in order to obtain a final decision. On the MICCAI EndoVis challenge laparoscopic dataset we achieved a precision and a recall of 91% in classification of 7 phases. Compared to the analysis based on one data type only, a combination of visual features and instrument signals allows better segmentation, reduction of the detection delay and discovery of the correct phase order.

Real-time instrument detection in minimally invasive surgery using radiofrequency identification technology

BackgroundA key part of surgical workflow recording is recognition of the instrument in use. We present a radiofrequency identification (RFID)-based approach for real-time tracking of laparoscopic instruments. MethodsThe system consists of RFID-tagged instruments and an antenna unit positioned on the Mayo stand. For reliability analysis, RFID tracking data were compared with the assessment of the perioperative video data of instrument changes (the reference standard for instrument application detection) in 10 laparoscopic cholecystectomies. When the tagged instrument was on the Mayo stand, it was referred to as “not in use.” Once it was handed to the surgeon, it was considered to be “in use.” Temporal miscounts (incorrect number of instruments “in use”) were analyzed. The surgeons and scrub nurses completed a questionnaire after each operation for individual system evaluation. ResultsA total of 110 distinct instrument applications (“in use” versus “not in use”) were eligible for analysis. No RFID tag failure occurred. The RFID detection rates were consistent with the period of effective instrument application. The delay in instrument detection was 4.2 ± 1.7 s. The highest percentage of temporal miscounts occurred during phases with continuous application of coagulation current. Surgeons generally rated the system better than the scrub nurses (P = 0.54). ConclusionsThe feasibility of RFID-based real-time instrument detection was successfully proved in our study, with reliable detection results during laparoscopic cholecystectomy. Thus, RFID technology has the potential to be a valuable additional tool for surgical workflow recognition that could enable a situation dependent assistance of the surgeon in the future.

Statistical modeling and recognition of surgical workflow

In this paper, we contribute to the development of context-aware operating rooms by introducing a novel approach to modeling and monitoring the workflow of surgical interventions. We first propose a new representation of interventions in terms of multidimensional time-series formed by synchronized signals acquired over time. We then introduce methods based on Dynamic Time Warping and Hidden Markov Models to analyze and process this data. This results in workflow models combining low-level signals with high-level information such as predefined phases, which can be used to detect actions and trigger an event. Two methods are presented to train these models, using either fully or partially labeled training surgeries. Results are given based on tool usage recordings from sixteen laparoscopic cholecystectomies performed by several surgeons.