Surgformer: Surgical Transformer with Hierarchical Temporal Attention for Surgical Phase Recognition

TL;DR

The paper tackles surgical phase recognition by addressing limitations of current one-stage and two-stage frameworks in modeling cross-frame spatial-temporal dependencies and reducing redundancy. It introduces Surgformer, a one-stage Transformer that uses divided spatial-temporal attention and sparse frame input, complemented by Hierarchical Temporal Attention for multi-resolution temporal modeling and Aggregated Spatial Attention for spatial information propagation. The authors show HTA and ASA yield consistent performance gains on two challenging datasets, Cholec80 and Autolaparo, with HTA providing stronger benefits on longer sequences and Autolaparo; TFA improves performance on Cholec80. Empirical results demonstrate Surgformer competitiveness with or superiority over state-of-the-art methods, particularly against two-stage baselines, and code is released for reproducibility.

Abstract

Existing state-of-the-art methods for surgical phase recognition either rely on the extraction of spatial-temporal features at a short-range temporal resolution or adopt the sequential extraction of the spatial and temporal features across the entire temporal resolution. However, these methods have limitations in modeling spatial-temporal dependency and addressing spatial-temporal redundancy: 1) These methods fail to effectively model spatial-temporal dependency, due to the lack of long-range information or joint spatial-temporal modeling. 2) These methods utilize dense spatial features across the entire temporal resolution, resulting in significant spatial-temporal redundancy. In this paper, we propose the Surgical Transformer (Surgformer) to address the issues of spatial-temporal modeling and redundancy in an end-to-end manner, which employs divided spatial-temporal attention and takes a limited set of sparse frames as input. Moreover, we propose a novel Hierarchical Temporal Attention (HTA) to capture both global and local information within varied temporal resolutions from a target frame-centric perspective. Distinct from conventional temporal attention that primarily emphasizes dense long-range similarity, HTA not only captures long-term information but also considers local latent consistency among informative frames. HTA then employs pyramid feature aggregation to effectively utilize temporal information across diverse temporal resolutions, thereby enhancing the overall temporal representation. Extensive experiments on two challenging benchmark datasets verify that our proposed Surgformer performs favorably against the state-of-the-art methods. The code is released at https://github.com/isyangshu/Surgformer.

Figures (4)

• Figure 1: Illustration of architectures of different paradigms. left: the one-stage paradigm with a unified spatial-temporal feature extractor. right: the two-stage paradigm with separate spatial and temporal feature extractors.
• Figure 2: Illustration of temporal attention mechanisms. In contrast to temporal attention only establishing dense global dependencies among frames, Hierarchical Temporal Attention constructs multiple temporal segments with different temporal resolutions to capture the long-term and short-term information.
• Figure 3: Overview of Surgformer. Given spatial-temporal tokens, we sequentially utilize Hierarchical Temporal Attention and Aggregated Spatial Attention to facilitate the learning of spatial-temporal feature representations.
• Figure 4: Qualitative results on four video sequences from the Cholec80 dataset.