Duygu Sarikaya

Jiashu Liao, Pietro Liò, Marc de Kamps et al.

Vision Transformers face a fundamental limitation: standard self-attention jointly processes spatial and channel dimensions, leading to entangled representations that prevent independent modeling of structural and semantic dependencies. This problem is especially pronounced in hyperspectral imaging, from satellite hyperspectral remote sensing to infrared pathology imaging, where channels capture distinct biophysical or biochemical cues. We propose DisentangleFormer, an architecture that achieves robust multi-channel vision representation through principled spatial-channel decoupling. Motivated by information-theoretic principles of decorrelated representation learning, our parallel design enables independent modeling of structural and semantic cues while minimizing redundancy between spatial and channel streams. Our design integrates three core components: (1) Parallel Disentanglement: Independently processes spatial-token and channel-token streams, enabling decorrelated feature learning across spatial and spectral dimensions, (2) Squeezed Token Enhancer: An adaptive calibration module that dynamically fuses spatial and channel streams, and (3) Multi-Scale FFN: complementing global attention with multi-scale local context to capture fine-grained structural and semantic dependencies. Extensive experiments on hyperspectral benchmarks demonstrate that DisentangleFormer achieves state-of-the-art performance, consistently outperforming existing models on Indian Pine, Pavia University, and Houston, the large-scale BigEarthNet remote sensing dataset, as well as an infrared pathology dataset. Moreover, it retains competitive accuracy on ImageNet while reducing computational cost by 17.8% in FLOPs. The code will be made publicly available upon acceptance.

Tugberk Erol, Duygu Sarikaya

Deep learning models are used to minimize the number of polyps that goes unnoticed by the experts and to accurately segment the detected polyps during interventions. Although state-of-the-art models are proposed, it remains a challenge to define representations that are able to generalize well and that mediate between capturing low-level features and higher-level semantic details without being redundant. Another challenge with these models is that they require too many parameters, which can pose a problem with real-time applications. To address these problems, we propose PlutoNet for polyp segmentation which requires only 2,626,537 parameters, less than 10\% of the parameters required by its counterparts. With PlutoNet, we propose a novel \emph{decoder consistency training} approach that consists of a shared encoder, the modified partial decoder which is a combination of the partial decoder and full-scale connections that capture salient features at different scales without being redundant, and the auxiliary decoder which focuses on higher-level relevant semantic features. We train the modified partial decoder and the auxiliary decoder with a combined loss to enforce consistency, which helps improve the encoders representations. This way we are able to reduce uncertainty and false positive rates. We perform ablation studies and extensive experiments which show that PlutoNet performs significantly better than the state-of-the-art models, particularly on unseen datasets and datasets across different domains.

Tugberk Erol, Duygu Sarikaya

Cancer is a disease that occurs as a result of the uncontrolled division and proliferation of cells. Colon cancer is one of the most common types of cancer in the world. Polyps that can be seen in the large intestine can cause cancer if not removed with early intervention. Deep learning and image segmentation techniques are used to minimize the number of polyps that goes unnoticed by the experts during these interventions. Although these techniques perform well in terms of accuracy, they require too many parameters. We propose a new model to address this problem. Our proposed model requires fewer parameters as well as outperforms the state-of-the-art models. We use EfficientNetB0 for the encoder part, as it performs well in various tasks while requiring fewer parameters. We use partial decoder, which is used to reduce the number of parameters while achieving high accuracy in segmentation. Since polyps have variable appearances and sizes, we use an asymmetric convolution block instead of a classic convolution block. Then, we weight each feature map using a squeeze and excitation block to improve our segmentation results. We used different splits of Kvasir and CVC-ClinicDB datasets for training, validation, and testing, while we use CVC- ColonDB, ETIS, and Endoscene datasets for testing. Our model outperforms state-of-art models with a Dice metric of %71.8 on the ColonDB test dataset, %89.3 on the EndoScene test dataset, and %74.8 on the ETIS test dataset while requiring fewer parameters. Our model requires 2.626.337 parameters in total while the closest model in the state-of-the-art is U-Net++ with 9.042.177 parameters.

Abdishakour Awale, Duygu Sarikaya

Modeling and automatically recognizing surgical activities are fundamental steps toward automation in surgery and play important roles in providing timely feedback to surgeons. Accurately recognizing surgical activities in video poses a challenging problem that requires an effective means of learning both spatial and temporal dynamics. Human gaze and visual saliency carry important information about visual attention and can be used to extract more relevant features that better reflect these spatial and temporal dynamics. In this study, we propose to use human gaze with a spatio-temporal attention mechanism for activity recognition in surgical videos. Our model consists of an I3D-based architecture, learns spatio-temporal features using 3D convolutions, as well as learns an attention map using human gaze as supervision. We evaluate our model on the Suturing task of JIGSAWS which is a publicly available surgical video understanding dataset. To our knowledge, we are the first to use human gaze for surgical activity recognition. Our results and ablation studies support the contribution of using human gaze to guide attention by outperforming state-of-the art models with an accuracy of 85.4%.

Haozheng Xu, Alistair Weld, Chi Xu et al.

Accurate instrument pose estimation is a crucial step towards the future of robotic surgery, enabling applications such as autonomous surgical task execution. Vision-based methods for surgical instrument pose estimation provide a practical approach to tool tracking, but they often require markers to be attached to the instruments. Recently, more research has focused on the development of marker-less methods based on deep learning. However, acquiring realistic surgical data, with ground truth instrument poses, required for deep learning training, is challenging. To address the issues in surgical instrument pose estimation, we introduce the Surgical Robot Instrument Pose Estimation (SurgRIPE) challenge, hosted at the 26th International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI) in 2023. The objectives of this challenge are: (1) to provide the surgical vision community with realistic surgical video data paired with ground truth instrument poses, and (2) to establish a benchmark for evaluating markerless pose estimation methods. The challenge led to the development of several novel algorithms that showcased improved accuracy and robustness over existing methods. The performance evaluation study on the SurgRIPE dataset highlights the potential of these advanced algorithms to be integrated into robotic surgery systems, paving the way for more precise and autonomous surgical procedures. The SurgRIPE challenge has successfully established a new benchmark for the field, encouraging further research and development in surgical robot instrument pose estimation.

Arpita Saggar, Jonathan C. Darling, Duygu Sarikaya et al.

Real-time talking head synthesis increasingly relies on deformable 3D Gaussian Splatting (3DGS) due to its low latency. Tri-planes are the standard choice for encoding Gaussians prior to deformation, since they provide a continuous domain with explicit spatial relationships. However, tri-plane representations are limited by grid resolution and approximation errors introduced by projecting 3D volumetric fields onto 2D subspaces. Recent work has shown the superiority of learnt embeddings for driving temporal deformations in 4D scene reconstruction. We introduce $\textbf{EmbedTalk}$, which shows how such embeddings can be leveraged for modelling speech deformations in talking head synthesis. Through comprehensive experiments, we show that EmbedTalk outperforms existing 3DGS-based methods in rendering quality, lip synchronisation, and motion consistency, while remaining competitive with state-of-the-art generative models. Moreover, replacing the tri-plane encoding with learnt embeddings enables significantly more compact models that achieve over 60 FPS on a mobile GPU (RTX 2060 6 GB). Our code will be placed in the public domain on acceptance.

Tianle Zeng, Junlei Hu, Gerardo Loza Galindo et al.

Computer vision-based technologies significantly enhance surgical automation by advancing tool tracking, detection, and localization. However, Current data-driven approaches are data-voracious, requiring large, high-quality labeled image datasets, which limits their application in surgical data science. Our Work introduces a novel dynamic Gaussian Splatting technique to address the data scarcity in surgical image datasets. We propose a dynamic Gaussian model to represent dynamic surgical scenes, enabling the rendering of surgical instruments from unseen viewpoints and deformations with real tissue backgrounds. We utilize a dynamic training adjustment strategy to address challenges posed by poorly calibrated camera poses from real-world scenarios. Additionally, we propose a method based on dynamic Gaussians for automatically generating annotations for our synthetic data. For evaluation, we constructed a new dataset featuring seven scenes with 14,000 frames of tool and camera motion and tool jaw articulation, with a background of an ex-vivo porcine model. Using this dataset, we synthetically replicate the scene deformation from the ground truth data, allowing direct comparisons of synthetic image quality. Experimental results illustrate that our method generates photo-realistic labeled image datasets with the highest values in Peak-Signal-to-Noise Ratio (29.87). We further evaluate the performance of medical-specific neural networks trained on real and synthetic images using an unseen real-world image dataset. Our results show that the performance of models trained on synthetic images generated by the proposed method outperforms those trained with state-of-the-art standard data augmentation by 10%, leading to an overall improvement in model performances by nearly 15%.

Arpita Saggar, Jonathan C. Darling, Vania Dimitrova et al.

Persona-based dialogue generation is an important milestone towards building conversational artificial intelligence. Despite the ever-improving capabilities of large language models (LLMs), effectively integrating persona fidelity in conversations remains challenging due to the limited diversity in existing dialogue data. We propose a novel framework SBS (Score-Before-Speaking), which outperforms previous methods and yields improvements for both million and billion-parameter models. Unlike previous methods, SBS unifies the learning of responses and their relative quality into a single step. The key innovation is to train a dialogue model to correlate augmented responses with a quality score during training and then leverage this knowledge at inference. We use noun-based substitution for augmentation and semantic similarity-based scores as a proxy for response quality. Through extensive experiments with benchmark datasets (PERSONA-CHAT and ConvAI2), we show that score-conditioned training allows existing models to better capture a spectrum of persona-consistent dialogues. Our ablation studies also demonstrate that including scores in the input prompt during training is superior to conventional training setups. Code and further details are available at https://arpita2512.github.io/score_before_you_speak

Xin Ci Wong, Duygu Sarikaya, Kieran Zucker et al.

Magnetic resonance (MR) imaging, including cardiac MR, is prone to domain shift due to variations in imaging devices and acquisition protocols. This challenge limits the deployment of trained AI models in real-world scenarios, where performance degrades on unseen domains. Traditional solutions involve increasing the size of the dataset through ad-hoc image augmentation or additional online training/transfer learning, which have several limitations. Synthetic data offers a promising alternative, but anatomical/structural consistency constraints limit the effectiveness of generative models in creating image-label pairs. To address this, we propose a diffusion model (DM) trained on a source domain that generates synthetic cardiac MR images that resemble a given reference. The synthetic data maintains spatial and structural fidelity, ensuring similarity to the source domain and compatibility with the segmentation mask. We assess the utility of our generative approach in multi-centre cardiac MR segmentation, using the 2D nnU-Net, 3D nnU-Net and vanilla U-Net segmentation networks. We explore domain generalisation, where, domain-invariant segmentation models are trained on synthetic source domain data, and domain adaptation, where, we shift target domain data towards the source domain using the DM. Both strategies significantly improved segmentation performance on data from an unseen target domain, in terms of surface-based metrics (Welch's t-test, p < 0.01), compared to training segmentation models on real data alone. The proposed method ameliorates the need for transfer learning or online training to address domain shift challenges in cardiac MR image analysis, especially useful in data-scarce settings.

Tugberk Erol, Tuba Caglikantar, Duygu Sarikaya

Deep learning models have been proposed for automatic polyp detection and precise segmentation of polyps during colonoscopy procedures. Although these state-of-the-art models achieve high performance, they often require a large number of parameters. Their complexity can make them prone to overfitting, particularly when trained on biased datasets, and can result in poor generalization across diverse datasets. Knowledge distillation and self-distillation are proposed as promising strategies to mitigate the limitations of large, over-parameterized models. These approaches, however, are resource-intensive, often requiring multiple models and significant memory during training. We propose a confidence-based self-distillation approach that outperforms state-of-the-art models by utilizing only previous iteration data storage during training, without requiring extra computation or memory usage during testing. Our approach calculates the loss between the previous and current iterations within a batch using a dynamic confidence coefficient. To evaluate the effectiveness of our approach, we conduct comprehensive experiments on the task of polyp segmentation. Our approach outperforms state-of-the-art models and generalizes well across datasets collected from multiple clinical centers. The code will be released to the public once the paper is accepted.

Arnaud Huaulmé, Duygu Sarikaya, Kévin Le Mut et al.

The "MIcro-Surgical Anastomose Workflow recognition on training sessions" (MISAW) challenge provided a data set of 27 sequences of micro-surgical anastomosis on artificial blood vessels. This data set was composed of videos, kinematics, and workflow annotations described at three different granularity levels: phase, step, and activity. The participants were given the option to use kinematic data and videos to develop workflow recognition models. Four tasks were proposed to the participants: three of them were related to the recognition of surgical workflow at three different granularity levels, while the last one addressed the recognition of all granularity levels in the same model. One ranking was made for each task. We used the average application-dependent balanced accuracy (AD-Accuracy) as the evaluation metric. This takes unbalanced classes into account and it is more clinically relevant than a frame-by-frame score. Six teams, including a non-competing team, participated in at least one task. All models employed deep learning models, such as CNN or RNN. The best models achieved more than 95% AD-Accuracy for phase recognition, 80% for step recognition, 60% for activity recognition, and 75% for all granularity levels. For high levels of granularity (i.e., phases and steps), the best models had a recognition rate that may be sufficient for applications such as prediction of remaining surgical time or resource management. However, for activities, the recognition rate was still low for applications that can be employed clinically. The MISAW data set is publicly available to encourage further research in surgical workflow recognition. It can be found at www.synapse.org/MISAW

Lena Maier-Hein, Matthias Eisenmann, Duygu Sarikaya et al.

Recent developments in data science in general and machine learning in particular have transformed the way experts envision the future of surgery. Surgical Data Science (SDS) is a new research field that aims to improve the quality of interventional healthcare through the capture, organization, analysis and modeling of data. While an increasing number of data-driven approaches and clinical applications have been studied in the fields of radiological and clinical data science, translational success stories are still lacking in surgery. In this publication, we shed light on the underlying reasons and provide a roadmap for future advances in the field. Based on an international workshop involving leading researchers in the field of SDS, we review current practice, key achievements and initiatives as well as available standards and tools for a number of topics relevant to the field, namely (1) infrastructure for data acquisition, storage and access in the presence of regulatory constraints, (2) data annotation and sharing and (3) data analytics. We further complement this technical perspective with (4) a review of currently available SDS products and the translational progress from academia and (5) a roadmap for faster clinical translation and exploitation of the full potential of SDS, based on an international multi-round Delphi process.

Duygu Sarikaya, Jason J. Corso, Khurshid A. Guru

Video understanding of robot-assisted surgery (RAS) videos is an active research area. Modeling the gestures and skill level of surgeons presents an interesting problem. The insights drawn may be applied in effective skill acquisition, objective skill assessment, real-time feedback, and human-robot collaborative surgeries. We propose a solution to the tool detection and localization open problem in RAS video understanding, using a strictly computer vision approach and the recent advances of deep learning. We propose an architecture using multimodal convolutional neural networks for fast detection and localization of tools in RAS videos. To our knowledge, this approach will be the first to incorporate deep neural networks for tool detection and localization in RAS videos. Our architecture applies a Region Proposal Network (RPN), and a multi-modal two stream convolutional network for object detection, to jointly predict objectness and localization on a fusion of image and temporal motion cues. Our results with an Average Precision (AP) of 91% and a mean computation time of 0.1 seconds per test frame detection indicate that our study is superior to conventionally used methods for medical imaging while also emphasizing the benefits of using RPN for precision and efficiency. We also introduce a new dataset, ATLAS Dione, for RAS video understanding. Our dataset provides video data of ten surgeons from Roswell Park Cancer Institute (RPCI) (Buffalo, NY) performing six different surgical tasks on the daVinci Surgical System (dVSS R ) with annotations of robotic tools per frame.

Duygu Sarikaya, Pierre Jannin

Modeling and recognition of surgical activities poses an interesting research problem. Although a number of recent works studied automatic recognition of surgical activities, generalizability of these works across different tasks and different datasets remains a challenge. We introduce a modality that is robust to scene variation, and that is able to infer part information such as orientational and relative spatial relationships. The proposed modality is based on spatial temporal graph representations of surgical tools in videos, for surgical activity recognition. To explore its effectiveness, we model and recognize surgical gestures with the proposed modality. We construct spatial graphs connecting the joint pose estimations of surgical tools. Then, we connect each joint to the corresponding joint in the consecutive frames forming inter-frame edges representing the trajectory of the joint over time. We then learn hierarchical spatial temporal graph representations using Spatial Temporal Graph Convolutional Networks (ST-GCN). Our experiments show that learned spatial temporal graph representations perform well in surgical gesture recognition even when used individually. We experiment with the Suturing task of the JIGSAWS dataset where the chance baseline for gesture recognition is 10%. Our results demonstrate 68% average accuracy which suggests a significant improvement. Learned hierarchical spatial temporal graph representations can be used either individually, in cascades or as a complementary modality in surgical activity recognition, therefore provide a benchmark for future studies. To our knowledge, our paper is the first to use spatial temporal graph representations of surgical tools, and pose-based skeleton representations in general, for surgical activity recognition.

Duygu Sarikaya, Pierre Jannin

In this paper, we address the open research problem of surgical gesture recognition using motion cues from video data only. We adapt Optical flow ConvNets initially proposed by Simonyan et al.. While Simonyan uses both RGB frames and dense optical flow, we use only dense optical flow representations as input to emphasize the role of motion in surgical gesture recognition, and present it as a robust alternative to kinematic data. We also overcome one of the limitations of Optical flow ConvNets by initializing our model with cross modality pre-training. A large number of promising studies that address surgical gesture recognition highly rely on kinematic data which requires additional recording devices. To our knowledge, this is the first paper that addresses surgical gesture recognition using dense optical flow information only. We achieve competitive results on JIGSAWS dataset, moreover, our model achieves more robust results with less standard deviation, which suggests optical flow information can be used as an alternative to kinematic data for the recognition of surgical gestures.

Duygu Sarikaya, Khurshid A. Guru, Jason J. Corso

We propose a novel multi-modal and multi-task architecture for simultaneous low level gesture and surgical task classification in Robot Assisted Surgery (RAS) videos.Our end-to-end architecture is based on the principles of a long short-term memory network (LSTM) that jointly learns temporal dynamics on rich representations of visual and motion features, while simultaneously classifying activities of low-level gestures and surgical tasks. Our experimental results show that our approach is superior compared to an ar- chitecture that classifies the gestures and surgical tasks separately on visual cues and motion cues respectively. We train our model on a fixed random set of 1200 gesture video segments and use the rest 422 for testing. This results in around 42,000 gesture frames sampled for training and 14,500 for testing. For a 6 split experimentation, while the conventional approach reaches an Average Precision (AP) of only 29% (29.13%), our architecture reaches an AP of 51% (50.83%) for 3 tasks and 14 possible gesture labels, resulting in an improvement of 22% (21.7%). Our architecture learns temporal dynamics on rich representations of visual and motion features that compliment each other for classification of low-level gestures and surgical tasks. Its multi-task learning nature makes use of learned joint re- lationships and combinations of shared and task-specific representations. While benchmark studies focus on recognizing gestures that take place under specific tasks, we focus on recognizing common gestures that reoccur across different tasks and settings and significantly perform better compared to conventional architectures.