A spatio-temporal network for video semantic segmentation in surgical videos

PURPOSE

Semantic segmentation in surgical videos has applications in intra-operative guidance, post-operative analytics and surgical education. Models need to provide accurate predictions since temporally inconsistent identification of anatomy can hinder patient safety. We propose a novel architecture for modelling temporal relationships in videos to address these issues.

METHODS

We developed a temporal segmentation model that includes a static encoder and a spatio-temporal decoder. The encoder processes individual frames whilst the decoder learns spatio-temporal relationships from frame sequences. The decoder can be used with any suitable encoder to improve temporal consistency.

RESULTS

Model performance was evaluated on the CholecSeg8k dataset and a private dataset of robotic Partial Nephrectomy procedures. Mean Intersection over Union improved by 1.30% and 4.27% respectively for each dataset when the temporal decoder was applied. Our model also displayed improvements in temporal consistency up to 7.23%.

CONCLUSIONS

This work demonstrates an advance in video segmentation of surgical scenes with potential applications in surgery with a view to improve patient outcomes. The proposed decoder can extend state-of-the-art static models, and it is shown that it can improve per-frame segmentation output and video temporal consistency.

Coronary artery segmentation in angiographic videos utilizing spatial-temporal information

Background

Coronary artery angiography is an indispensable assistive technique for cardiac interventional surgery. Segmentation and extraction of blood vessels from coronary angiographic images or videos are very essential prerequisites for physicians to locate, assess and diagnose the plaques and stenosis in blood vessels.

Methods

This article proposes a novel coronary artery segmentation framework that combines a three–dimensional (3D) convolutional input layer and a two–dimensional (2D) convolutional network. Instead of a single input image in the previous medical image segmentation applications, our framework accepts a sequence of coronary angiographic images as input, and outputs the clearest mask of segmentation result. The 3D input layer leverages the temporal information in the image sequence, and fuses the multiple images into more comprehensive 2D feature maps. The 2D convolutional network implements down–sampling encoders, up–sampling decoders, bottle–neck modules, and skip connections to accomplish the segmentation task.

Results

The spatial–temporal model of this article obtains good segmentation results despite the poor quality of coronary angiographic video sequences, and outperforms the state–of–the–art techniques.

Conclusions

The results justify that making full use of the spatial and temporal information in the image sequences will promote the analysis and understanding of the images in videos.

Temporal variability of surgical technical skill perception in real robotic surgery

PURPOSE

Summary score metrics, either from crowds of non-experts, faculty surgeons or from automated performance metrics, have been trusted as the prevailing method of reporting surgeon technical skill. The aim of this paper is to learn whether there exist significant fluctuations in the technical skill assessments of a surgeon throughout long durations of surgical footage.

METHODS

A set of 12 videos of robotic surgery cases from common human patient robotic surgeries were used to evaluate the perceived technical skill at each individual minute of the surgical videos, which were originally 12-15 min in length. A linear mixed-effects model for each video was used to compare the ratings of each minute to those from every other minute in order to learn whether a change in scores over time can be detected and reliably measured apart from inter- and intrarater variation.

RESULTS

Modeling the change over time of the global evaluative assessment of robotic skills scores significantly contributed to the prediction models for 11 of the 12 surgeons. This demonstrates that measurable changes in technical skill occur over time during robotic surgery.

CONCLUSION

The findings from this research raise questions about the optimal duration of footage needed to be evaluated to arrive at an accurate rating of surgical technical skill for longer procedures. This may imply non-negligible label noise for supervised machine learning approaches. In the future, it may be necessary to report a surgeon's skill variability in addition to their mean score to have proper knowledge of a surgeon's overall skill level.

Coronary angiography video segmentation method for assisting cardiovascular disease interventional treatment

Background

Coronary heart disease is one of the diseases with the highest mortality rate. Due to the important position of cardiovascular disease prevention and diagnosis in the medical field, the segmentation of cardiovascular images has gradually become a research hotspot. How to segment accurate blood vessels from coronary angiography videos to assist doctors in making accurate analysis has become the goal of our research.

Method

Based on the U-net architecture, we use a context-based convolutional network for capturing more information of the vessel in the video. The proposed method includes three modules: the sequence encoder module, the sequence decoder module, and the sequence filter module. The high-level information of the feature is extracted in the encoder module. Multi-kernel pooling layers suitable for the extraction of blood vessels are added before the decoder module. In the filter block, we add a simple temporal filter to reducing inter-frame flickers.

Results

The performance comparison with other method shows that our work can achieve 0.8739 in Sen, 0.9895 in Acc. From the performance of the results, the accuracy of our method is significantly improved. The performance benefit from the algorithm architecture and our enlarged dataset.

Conclusion

Compared with previous methods that only focus on single image analysis, our method can obtain more coronary information through image sequences. In future work, we will extend the network to 3D networks.

FetNet: a recurrent convolutional network for occlusion identification in fetoscopic videos

PURPOSE

Fetoscopic laser photocoagulation is a minimally invasive surgery for the treatment of twin-to-twin transfusion syndrome (TTTS). By using a lens/fibre-optic scope, inserted into the amniotic cavity, the abnormal placental vascular anastomoses are identified and ablated to regulate blood flow to both fetuses. Limited field-of-view, occlusions due to fetus presence and low visibility make it difficult to identify all vascular anastomoses. Automatic computer-assisted techniques may provide better understanding of the anatomical structure during surgery for risk-free laser photocoagulation and may facilitate in improving mosaics from fetoscopic videos.

METHODS

We propose FetNet, a combined convolutional neural network (CNN) and long short-term memory (LSTM) recurrent neural network architecture for the spatio-temporal identification of fetoscopic events. We adapt an existing CNN architecture for spatial feature extraction and integrated it with the LSTM network for end-to-end spatio-temporal inference. We introduce differential learning rates during the model training to effectively utilising the pre-trained CNN weights. This may support computer-assisted interventions (CAI) during fetoscopic laser photocoagulation.

RESULTS

We perform quantitative evaluation of our method using 7 in vivo fetoscopic videos captured from different human TTTS cases. The total duration of these videos was 5551 s (138,780 frames). To test the robustness of the proposed approach, we perform 7-fold cross-validation where each video is treated as a hold-out or test set and training is performed using the remaining videos.

CONCLUSION

FetNet achieved superior performance compared to the existing CNN-based methods and provided improved inference because of the spatio-temporal information modelling. Online testing of FetNet, using a Tesla V100-DGXS-32GB GPU, achieved a frame rate of 114 fps. These results show that our method could potentially provide a real-time solution for CAI and automating occlusion and photocoagulation identification during fetoscopic procedures.

Incremental multi-class semi-supervised clustering regularized by Kalman filtering

This paper introduces an on-line semi-supervised learning algorithm formulated as a regularized kernel spectral clustering (KSC) approach. We consider the case where new data arrive sequentially but only a small fraction of it is labeled. The available labeled data act as prototypes and help to improve the performance of the algorithm to estimate the labels of the unlabeled data points. We adopt a recently proposed multi-class semi-supervised KSC based algorithm (MSS-KSC) and make it applicable for on-line data clustering. Given a few user-labeled data points the initial model is learned and then the class membership of the remaining data points in the current and subsequent time instants are estimated and propagated in an on-line fashion. The update of the memberships is carried out mainly using the out-of-sample extension property of the model. Initially the algorithm is tested on computer-generated data sets, then we show that video segmentation can be cast as a semi-supervised learning problem. Furthermore we show how the tracking capabilities of the Kalman filter can be used to provide the labels of objects in motion and thus regularizing the solution obtained by the MSS-KSC algorithm. In the experiments, we demonstrate the performance of the proposed method on synthetic data sets and real-life videos where the clusters evolve in a smooth fashion over time.