What is role of artificial intelligence in regional anaesthesia?

A surgical technique known as ultrasound-guided regional anaesthesia (USG-RA) uses real-time USG pictures with a colour overlay applied to highlight important anatomical features. It also entails the best possible collection and analysis of ultrasonography data to define sonoanatomy. However, for trainees and anesthesiologists , anatomical knowledge and the capacity to understand sonoanatomy are challenging and incomplete. The creation of computer systems that are able to carry out operations that typically require human intellect, such as making decisions, identifying objects, and resolving challenging issues, is known as artificial intelligence (AI). Deep learning (DL) is a subset of machine learning (ML) that makes use of artificial networks that mimic the neural networks found in human brains. ML enables computers to learn by subjecting them to an algorithm. Recently published narrative review in IJA evaluated the role of AI in regional anesthesia.

Medical imaging and diagnosis, medical education, medication creation, treatment, and monitoring, medical data management, digital consultation, individual health monitoring, illness diagnosis, health plan analysis, and medical audits are just a few of the healthcare domains where artificial intelligence (AI) is now being employed. AI is able to identify objects and structures with a high degree of sensitivity and precision, generate reports quickly using preprogrammed algorithms, and provide findings that are highly consistent. Numerous strategies have been proposed to address the use of AI in USG-RA, including raising the regional anaesthesia (RA) success rate, enhancing safety, and lowering the risk of complications.

AI helps RA practitioners detect anatomical features accurately and minimise problems. Sonographic picture optimisation, interpretation, and needle visibility are all enhanced by AI-guided models. AI-assisted USG-RA may help non-experts find the right USG anatomy to execute regional blocks and make it easier to identify anatomical components. Prior research brought attention to the apparent gaps in our understanding of anatomy, which USG image interpretation aid may help to support.

According to reports, AI was useful in 99.7% of the situations. The performance of blocks employing USG varies statistically significantly depending on the area. The bones and veins were the most highly evaluated anatomical locations, indicating the possibility of AI's therapeutic value in USG-RA, particularly when used by non-experts.

It is difficult to create AI systems that can recognise every anatomical characteristic with USG because of operator reliance, complexity, and variety. Machine learning (ML) may be used to teach automated image interpretation systems to distinguish between different structures, such as blood arteries, bones, soft tissues, and nerves. By locating targets and mapping the best insertion locations by recognising pertinent landmarks and guide structures, this technique may enhance the understanding of USG anatomy.

Artificial intelligence is used in USG-RA, or AI-generated automated target detection, to monitor and identify nerve structures, help identify blood arteries, anatomical features, and ensure that needles are inserted correctly. Improved interpretation of the spinal USG and real-time interpretation of anatomical structures for prompt decision-making during blocks are two advantages of this method. AI-assisted USG-RA is not without dangers and shortcomings, however, including the possibility of misinterpreting images, an excessive dependence on professionals, and problems with tracking and detection.

Increased likelihood of block failure and needle damage to adjacent structures, along with consequences such nerve injury, systemic toxicity of local anaesthetic, peripheral injury, and peritoneal injury, are the most significant drawbacks of AI-assisted USG-RA. Research has shown that there is a modest impression of increased danger when using AI help; nonetheless, problems might have significant clinical implications. Errors may arise from incorrect highlighting, misrepresented screen colour, and device use.

Correct anatomical structure identification may be useful to anesthesiologists, but it does not guarantee safe USG-RA or direct needle insertion. AI-assisted technology should be utilised as a source of extra information rather than a decision-maker. It may be difficult to track anatomical targets during USG-guided operations because of variations in light, occlusion, noise, and target distortion. Moreover, there's a chance that the object's speed may change suddenly, multiplicative noise will contaminate the photos, and some characteristics found won't really belong there.

AI-guided USG-RA has the potential to lower failure rates, increase safety and effectiveness, and optimise the advantages of USG guidance. While technical robots use motion and visual measurements to create sensors for training, computer vision and cognitive robots may provide greater accuracy and dexterity.

Even with AI's increasing advantages, there are still moral and legal questions to be answered. Practitioners using AI in RA practice should be aware of both these advantages and possible drawbacks. To sum up, AI has the potential to revolutionise RA treatment, enhance patient care, and increase pain management accuracy.

Reference –

Balavenkatasubramanian, J; Kumar, Senthil1; Sanjayan, R.D.2. Artificial intelligence in regional anaesthesia. Indian Journal of Anaesthesia 68(1):p 100-104, January 2024. | DOI: 10.4103/ija.ija_1274_23.