Trans-SVNet: hybrid embedding aggregation Transformer for surgical workflow analysis

PURPOSE

Real-time surgical workflow analysis has been a key component for computer-assisted intervention system to improve cognitive assistance. Most existing methods solely rely on conventional temporal models and encode features with a successive spatial-temporal arrangement. Supportive benefits of intermediate features are partially lost from both visual and temporal aspects. In this paper, we rethink feature encoding to attend and preserve the critical information for accurate workflow recognition and anticipation.

METHODS

We introduce Transformer in surgical workflow analysis, to reconsider complementary effects of spatial and temporal representations. We propose a hybrid embedding aggregation Transformer, named Trans-SVNet, to effectively interact with the designed spatial and temporal embeddings, by employing spatial embedding to query temporal embedding sequence. We jointly optimized by loss objectives from both analysis tasks to leverage their high correlation.

RESULTS

We extensively evaluate our method on three large surgical video datasets. Our method consistently outperforms the state-of-the-arts across three datasets on workflow recognition task. Jointly learning with anticipation, recognition results can gain a large improvement. Our approach also shows its effectiveness on anticipation with promising performance achieved. Our model achieves a real-time inference speed of 0.0134 second per frame.

CONCLUSION

Experimental results demonstrate the efficacy of our hybrid embeddings integration by rediscovering the crucial cues from complementary spatial-temporal embeddings. The better performance by multi-task learning indicates that anticipation task brings the additional knowledge to recognition task. Promising effectiveness and efficiency of our method also show its promising potential to be used in operating room.

Deep learning-based fetoscopic mosaicking for field-of-view expansion

PURPOSE

Fetoscopic laser photocoagulation is a minimally invasive surgical procedure used to treat twin-to-twin transfusion syndrome (TTTS), which involves localization and ablation of abnormal vascular connections on the placenta to regulate the blood flow in both fetuses. This procedure is particularly challenging due to the limited field of view, poor visibility, occasional bleeding, and poor image quality. Fetoscopic mosaicking can help in creating an image with the expanded field of view which could facilitate the clinicians during the TTTS procedure.

METHODS

We propose a deep learning-based mosaicking framework for diverse fetoscopic videos captured from different settings such as simulation, phantoms, ex vivo, and in vivo environments. The proposed mosaicking framework extends an existing deep image homography model to handle video data by introducing the controlled data generation and consistent homography estimation modules. Training is performed on a small subset of fetoscopic images which are independent of the testing videos.

RESULTS

We perform both quantitative and qualitative evaluations on 5 diverse fetoscopic videos (2400 frames) that captured different environments. To demonstrate the robustness of the proposed framework, a comparison is performed with the existing feature-based and deep image homography methods.

CONCLUSION

The proposed mosaicking framework outperformed existing methods and generated meaningful mosaic, while reducing the accumulated drift, even in the presence of visual challenges such as specular highlights, reflection, texture paucity, and low video resolution.

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.

Combined 2D and 3D tracking of surgical instruments for minimally invasive and robotic-assisted surgery

PURPOSE

Computer-assisted interventions for enhanced minimally invasive surgery (MIS) require tracking of the surgical instruments. Instrument tracking is a challenging problem in both conventional and robotic-assisted MIS, but vision-based approaches are a promising solution with minimal hardware integration requirements. However, vision-based methods suffer from drift, and in the case of occlusions, shadows and fast motion, they can be subject to complete tracking failure.

METHODS

In this paper, we develop a 2D tracker based on a Generalized Hough Transform using SIFT features which can both handle complex environmental changes and recover from tracking failure. We use this to initialize a 3D tracker at each frame which enables us to recover 3D instrument pose over long sequences and even during occlusions.

RESULTS

We quantitatively validate our method in 2D and 3D with ex vivo data collected from a DVRK controller as well as providing qualitative validation on robotic-assisted in vivo data.

CONCLUSIONS

We demonstrate from our extended sequences that our method provides drift-free robust and accurate tracking. Our occlusion-based sequences additionally demonstrate that our method can recover from occlusion-based failure. In both cases, we show an improvement over using 3D tracking alone suggesting that combining 2D and 3D tracking is a promising solution to challenges in surgical instrument tracking.

Comparative study on surface reconstruction accuracy of stereo imaging devices for microsurgery

PURPOSE

Processing stereoscopic image data is an emerging field. Especially in microsurgery that requires sub-millimeter accuracy, application of stereo-based methods on endoscopic and microscopic scenarios is of major interest. In this context, direct comparison of stereo-based surface reconstruction applied to several camera settings is presented.

METHODS

A method for stereo matching is proposed and validated on in-vitro data. Demonstrating suitability for surgical scenarios, this method is applied to two custom-made stereo cameras, a miniaturized, bendable surgical endoscope and an operating microscope. Reconstruction accuracy is assessed on a custom-made reference sample. Subsequent to its fabrication, a coordinate measuring arm is used to acquire ground truth. Next, the sample is positioned by a robot at varying distances to each camera. Surface estimation is performed, while the specimen is localized based on. markers. Finally, the error between estimated surface and ground truth is computed.

RESULTS

Sample measurement with the coordinate measuring arm yields reliable ground truth data with a root-mean-square error of 11.2 μm. Overall surface reconstruction with analyzed cameras is quantified by a root-mean-square error of less than 0.18 mm. Microscope setting with the highest magnification yields the most accurate measurement, while the maximum deviation does not exceed 0.5 mm. Custom-made stereo cameras perform similar but with outliers of increased magnitude. Miniaturized, bendable surgical endoscope produces the maximum error of approximately 1.2 mm.

CONCLUSIONS

Reconstruction results reveal that microscopic imaging outperforms investigated chip-on-the-tip solutions, i.e., at higher magnification. Nonetheless, custom-made cameras are suitable for application in microsurgery. Although reconstruction with the miniaturized endoscope is more inaccurate, it provides a good trade-off between accuracy, outer dimensions and accessibility to hard-to-reach surgical sites.