Transform Surgery Now with AI Visual Guidance

Simply put, computer vision interprets and extracts patterns using AI-powered algorithms to analyze images and surgery recordings.¹ While still relatively new, computer vision shows promise in providing precise visual assistance intraoperatively based on global expertise.² It works by utilizing deep neural networks and has emerged as an integral component of machine learning.¹ Furthermore, through visual segmentation, which is the ability to segment an unknown image into different parts, computer vision helps in the precise demarcation of the surgical site using anatomical localization by swiftly analyzing and interpreting imaging data.³

AI-driven 3-dimensional (3D)-guided surgery

Besides computer vision, AI-driven 3D anatomical reconstruction enhances surgical planning and intraoperative navigation for complex procedures.² A study by Porpiglia and associates found that elastic 3D virtual models successfully simulated prostate deformation and identified lesion locations in a dynamic setting.⁴ When 2-dimensional images were used, the study participants reported experiencing a harder process of mental imagination.⁴ Additionally, accuracy in identifying the lesion intraoperatively was poorer.

In the same study, the researchers compared data from surgeries that utilized 3D-guided surgery and those that did not. The study found that the 3D-guided surgery model significantly improved the identification of extracapsular extension with an increase in detection rate from 47.0% to 100% (p<0.002).⁴

Navigating surgery with augmented reality (AR)

AR is when a device superimposes data or images over the real environment to form a blended view.⁵ In the context of surgery, the preoperative or intraoperative field is superimposed over the operative field, which allows surgeons to see pertinent data in real time, such as those derived from computed tomography (CT) or magnetic resonance imaging (MRI), while performing surgery.⁵ This can aid in the identification of important anatomical landmarks or tissue planes that might be otherwise hard to identify.⁵

As we are aware, the cornerstone of optimizing post-surgical patient recovery is the preservation of neurovascular bundles (NVB).⁶ Checcucci and associates developed an AI-powered automatic augmented reality (AAR) system that aided the surgeon in performing a 3D AAR-guided robot-assisted radical prostatectomy.⁶ The study yielded 3 notable findings:⁶

• The 33D AAR-guided biopsy correctly identified cancer cells at the NVB level in 87.5% of pT3 patients.
• Excisional biopsy at the NVB level recorded 7.1% of positive surgical margins even in cases of locally advanced disease.
• A large number of neurovascular fibers were preserved with a potency recovery rate of 58.8% at 3 months post-surgery.

Real-time surgical monitoring

Monitoring the patient's anatomy in real time must take into consideration a myriad of factors such as equipment, blood, smoke, adipose tissue, and tissue deformation that can obstruct the visual field.³ Despite these challenges, researchers are making significant strides in automating the recognition of patient anatomy.

In line with this, there is significant ongoing research on the development of machine learning (ML) systems that can be utilized intraoperatively. A recent study utilized ML systems for the analysis of color and texture to identify anatomical components during guided surgery.³ The automated skill assessment showed a significant association with professional evaluation of dissection quality with an impressive accuracy rate of 83.3%.³

Intraoperative bleeding is considered a leading challenge in laparoscopic and robotic surgery, accounting for 23% of all adverse events.⁷ Immediate identification of bleeding during endoscopic procedures would help optimize surgical outcomes.⁷ The Bleeding Artificial Intelligence-Based Detector (BLAIR) system was created and powered by ML with this end goal in mind.⁷ It works by receiving video inputs during live surgery, which is fed into the system and provides the surgeon with information on the probability of bleeding in the next few instants of surgery.⁷ A study that utilized the BLAIR software showed it accurately predicted bleeding occurrences during surgery with an accuracy rate of more than 90%.⁷

Real-time haptic feedback

While 3D visualization has been shown to aid surgical precision, optimal surgical outcomes also depend on visual and tactile cues.³ Factors such as the technique of dissection, amount of pressure exerted, and assessment of tissue responses can have an impact on surgery results.³ Excessive force in surgery could have poor effects on the NVB and prolong recovery periods. On the other hand, insufficient force could cause poor suture retention.³

A study was carried out by Dai and associates in which a warning system was designed that combined biaxial shear detection and haptic feedback to detect suture breakage.^3,8 This system helped warn the surgeon of potential suture ruptures based on suture tightness.^3,8 The study found a significant reduction in the rates of suture breakage (59%). Notably, the Da Vinci robotic surgery system also uses biaxial shear sensors to offer vibrotactile feedback to the operating surgeon.^3,8