Page 41 - Read Online

Ding et al. Art Int Surg 2024;4:109-38 https://dx.doi.org/10.20517/ais.2024.16 Page 125

Furthermore, evaluating and enhancing surgical workflows and skills through these technologies can
significantly advance surgeon training. By providing objective, quantifiable feedback on surgical techniques,
DTs can support a comprehensive approach to assessing and improving surgical proficiency. This not only
aids in training novices but also enhances performance evaluation across a spectrum of surgeons, from
novices to experts, including those performing robot-assisted procedures [20,22,23,283-285] . The development of
advanced cognitive surgical assistance technologies, based on the analysis of surgical workflows and skills,
represents another opportunity. These technologies have the potential to enhance operational safety and
foster semi-autonomous robotic systems that anticipate and adapt to the surgical team’s needs, thereby
improving the collaborative efficiency of the surgical environment [1,3,286] .

[6]
Intraoperative guidance technologies offer surgeons improved precision and real-time feedback .
Innovations such as mixed reality overlays and virtual fixtures could see their utility and efficacy greatly
enhanced through integration with DTs. This synergy could further refine surgical accuracy and patient
safety. Moreover, advancements like tool and needle guidance systems, alongside automated image
acquisition , exemplify progress in geometric understanding and digital innovation. Integrating DTs with
[4,8]
these surgical technologies holds the potential to improve standards for minimally invasive procedures and
overall surgical quality.

Table 2 summarizes important methods for geometric scene understanding tasks and applications.

DISCUSSION
The concept of DTs is rapidly gaining momentum in various surgical procedures, showcasing its
transformative potential in shaping the future of surgery. The introduction of DT across various surgical
[278]
[279]
procedures such as skull base surgery [7,272-275] , liver surgery , cardiovascular intervention , and orthopedic
surgery [280,281] demonstrate the broad effectiveness of DT technology in improving surgical precision and
patient outcomes across different medical specialties. As stated in Section “APPLICATIONS OF
GEOMETRIC SCENE UNDERSTANDING EMPOWERED DIGITAL TWINS”, the emergence of novel
geometric scene understanding applications, such as phase recognition and gesture classification, could
further empower DT models to interpret the broader context and flow of surgical operations. The potential
ability to derive context-aware intelligence from a deep understanding of surgical dynamics further
emphasizes DTs’ potential to advance robot-assisted surgeries and procedural planning. Geometric scene
understanding-empowered DTs offer an alternate approach to the holistic understanding of the surgical
scene through virtual models and have the potential to subsequently enhance the surgical process, from
planning and execution to training and postoperative analysis, driving the digital revolution in surgery.

Geometric scene understanding forms the backbone of DTs that incorporate and process diverse enriched
data from different stages of surgery. The evolution of geometric information processing has transitioned
from simple low-level feature processing to neural network-based methods, introducing innovative
geometric representations like neural fields. Various geometric scene understanding tasks have also been
established, with corresponding methods achieving significant performance improvements. However,
challenges persist in the representation of geometric information, the development of geometric scene
understanding within the surgical domain, and its application to the DT paradigm.

In geometric representations, there is no single form that can meet all the requirements of DT in terms of
accuracy, applicability, efficiency, interactivity, and reliability. Grid-based methods compromise the
accuracy and processing efficiency for convenient structure and representation ability. It also lacks
interactivity and reliability, as the geometric information is mainly represented by the aggregation of