Artificial intelligence in spine surgery: The new kid on the block

Spine surgery like other medical and surgical branches is rapidly evolving.[1] Modern-day spine surgeries consist of minimally invasive/endoscopic approaches and biomechanically superior implants with advanced guidance tools, which allow precise and near-error–free surgeries in regions with complex anatomy.[2,3] The synergistic development of imaging technology (from radiographs to magnetic resonance imaging [MRI], and intraoperative imaging techniques like C-arm to intraoperative computed tomography [CT] MRI) and image analytical software have provided better stereotactic guidance during spine surgeries.[1] The progress, chiefly over the last 50 years, has allowed better understanding and management of spine injuries and diseases. This has also allowed transformation from extensile exposures to minimally invasive surgeries, which has reduced tissue damage, hospital stay, blood loss, and surgical pain, and is allowing quicker and better recovery.[2,3] Advances in Imaging The advancement in imaging in synergy with advancement in image analytical technologies has been pivotal in reforming spine surgeries.[1] The foundation laid by W. Roentgen in 1895 with discovery of X-rays was taken further by the development of X-ray image intensifiers in 1940. However, it was in 1955, when the development of C-arm (Philips, Amsterdam) enhanced the operating room use of X-rays, due to its mobility.[4] It became the third eye of the surgeon and guided surgical incisions, spinal level, alignment of bone, and with time, fluoroscopic-guided implant placement.[3,4] Godfrey Hounsfield et al. in 1971 developed the first CT scanner machine after obtaining cross-sectional images from X-rays. With the advances in its gantry and scanners, 3D reconstruction became a reality, which paved the way toward the modern day navigational technologies.[5,6] On the other hand, Paul Lauterbur in 1973, using principles of rotating magnetic field given by Nikola Tesla, obtained the first MRI. Advances in its technology and development of more MRI sequences (such as diffuse tensor imaging) has allowed early diagnosis, better planning, and more precise assessment of pathoanatomy of spinal pathologies.[3,7] The development of imaging tools with computation techniques along with picture archiving and communication system (PACS) has allowed better analysis of images and integration of MRI and CT images. These integrated images allow assessment and planning for both soft tissue and bone. Softwares for merging Digital Imaging and Communications in Medicine (DICOM) files from CT and MRI and surgical planning are easily available for clinical use, but they have their own learning curve. Stereotactic/Navigational Spine Surgeries Stereotactic surgeries were introduced for brain surgery, where Cartesian coordinates were used to localize anatomical landmarks in the brain using external landmarks. With development of MRI and CT, these techniques were used for stereotactic surgeries, especially brain and spinal biopsies.[8] After the introduction of this technique for spine surgery, image-guided spine surgeries have increased especially in the last 10–15 years.[9] Navigational spine surgery systems (e.g., StealthStation, Medtronic, Memphis, TN) need preoperative patient data (imaging) and intraoperative patient data, which is registered/aligned using probes/trackers, which are attached to fixed (bony) points. The fusion of the preoperative images with intraoperative data provides surgical guidance to the surgeon regarding patient’s anatomy for complex cases and improves precision.[9] But, since the relation of structures are different in both these data, as preoperative data were obtained in supine position, and intraoperative data obtained in prone position, this technique had some issues.[3] Invention of O-arm (Medtronic, Dublin, Ireland), which is an intraoperative imaging system having O-shaped gantry, allowed 2D fluoroscopic images and/or 3D imaging using cone beam CT within seconds.[6] Accessibility to intraoperative CT data for navigation further reduced the error due to positioning for spinal instrumentation. Newer intraoperative CT scan machines and navigation systems have been developed, which have reduced the radiation dose, and have better integration and better mobility; but still errors due to movement of spinous process and blockage of path between trackers and camera happen (line of sight interruption). Newer systems having additional usage of intraoperative images using machine-vision cameras integrated to surgical lights have been introduced, but again, their access to the surgical field throughout the surgery is the limitation.[3,10] Robotic Spine Surgery Surgery around neural tissue always demanded precision as inadvertent injury can lead to lifelong morbidity and even mortality. Therefore, tools, which reduce errors due to planning and physiological issues/human errors, are desirable for spine surgeries. Errors due to hand fatigue and hand tremors can be reduced if the machine/robot takes its place after instructions from a human brain. This has been the reason behind the evolution of robotic surgeries. A robotic system needs visualization system/navigation system (preoperative CT/MRI imaging, intraoperative fluoroscopic images, intraoperative imaging, data from trackers, and/or data from 3D optical camera). This data may be used by the software, which allows surgeon to plan or track the predetermined route of screw insertion and a working arm. Many systems have been developed for patient use, and despite improvement in the precision of surgery, the high initial costs, early technical errors, need of specialized training, and initial learning curve are prohibitory. Moreover, as more and more imaging/inputs are used for planning, there is a potential for error due to software-to-software interpretation. More studies are warranted to validate the robotic platform over only navigate platforms.[3,11] Artifical Intelligence and Augmented Reality in Spine Surgery Artificial intelligence, in particularly deep learning (DL) just like in other medical specialties, has application in spine surgeries. The patient data in the form of DICOM files are integrated and processed by one of the many software available and with the help of DL are used to sync the data between images and patients, and help in creating preoperative plan in the form of images. This data can be visualized in various forms. If the surgeon views the data using specialized glasses (closed, eye-covering headpieces) in a virtual field only, it is called virtual reality (VR). This gives the user an immersive experience, that is, as if the user is in the virtual space. VR may be used for simulation of real world, planning, gaming, surgical training, and education. On the other hand, if these images (3D holographic virtual data) are projected/overlayed in the real field of the user with immersive experience using specialized glasses (augmented reality [AR] glasses [e.g., Microsoft HoloLens]) then it is called AR. The third variant, mixed reality (MR), allows projection of virtual data into the real world just like AR but also allows the interaction of actual physical objects with projected digital objects. MR is also called extended reality or hybrid reality. As VR does not allow real environment view, hence, it cannot be used for intraoperative use. Many studies, mainly cadaveric, have been done using AR for pedicle screw insertion. Many benefits and issues have been noted with this new technology.[12,13] Usage of this technique does not require any bulky equipment, hence allows unobstructed and clear view of field. There is less exposure of radiation, no need to move cumbersome equipment near the surgical field (maintenance of sterility of surgical field is easier) and does not require radiology technician for operation etc. It needs preoperative acquisition of imaging data and planning over specialized software which need technological support and synchronization with AR glasses. Errors in this synchronization of the preoperative data with patient’s anatomy and surgical instrument has led to sometimes wrong placement of screws in cadaveric studies.[13] The AR glasses commonly used are Microsoft HoloLens, but it is currently heavy and with prolonged use can cause neck pain and has issues due to overheating. Better ergonomic design, lighter devices will be needed for prolonged surgeries with improved heat exchange/cooling off mechanisms. Need of smoothly running snag free software is critical along with intraoperative tech-support to avoid unnecessary delays, before they can be advocated for patient care. There is also an additional disadvantage of lack of shared experience, which means the assistant of operating surgeon cannot visualize the holographic image which the operating surgeon is visualizing. Although many cadaveric studies[13-16] regarding the usage of AR and MR in guiding stereotactic surgeries in spine are there, due to concerns mentioned above including the training needed to use it, more refinement is required in it. It definitely holds a lot of promise and can be the future, not only in spine but in all surgical fields. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.