Resumen de Optimization of computer-assisted intraoperative guidance for complex oncological procedures
In 2020, nearly 20 million people were diagnosed with cancer, and almost 10 million died from it. It represents the second leading cause of death in the world. Cancer could be prevented in 30 to 50% of the cases by avoiding tobacco or alcohol use, having healthy diets, or reducing air pollution. Infections are also one of the main causes of cancer, especially in low- and middle-income countries, and only about 5% of cancer cases are caused by inherited genetic mutations. Cancer creates a physical, emotional, and financial impact on individuals and their families and is a significant burden to communities and health care systems. Additionally, limitations in heath access affect cancer diagnosis, treatment, and survival in developing countries. Approximately 70% of deaths caused by cancer occur in these regions due to the high inequalities between countries with a high Human Development Index (HDI) and those with a lower one. Hence, it is necessary to strengthen health systems by finding cost-effective strategies and sustainable solutions for cancer prevention and treatment.
Oncology is the branch of medicine that studies cancer and deals with its diagnosis, treatment, and prevention. In the last century, significant advances have been accomplished in this field to gain a deeper understanding of cancer and better diagnose, prevent, and treat it. There are different options for treatment depending on the type, location, and stage. These options include surgery, radiotherapy, chemotherapy, or a combination. The primary purpose of treatment is to cure cancer or prolong a patient’s life, but it can also focus on improving the patient’s quality of life. Among the options for cancer treatment, surgery represents the oldest one. It is applied at different cancer stages and for various purposes, including diagnosis, prevention, primary treatment, debulking, reconstruction, or palliative care.
When a cancerous tumor is confirmed and localized, the most common treatment is tumor removal through surgery. It is also called primary treatment and is the best chance for cure in most cases. In these interventions, surgeons remove the tumor along with nearby tissue that may also be affected. This extra tissue is called surgical or resection margin, and it is examined after surgery to ensure it is clear of cancer cells. It is defined as a negative margin if no cancer cells are found in the outer edge of the removed tissue. If cancerous cells are found in the outer rim, additional surgery may be necessary.
Cancer treatment has improved significantly in the last century, reducing incidence and mortality thanks to advances in prevention and treatment. Late advancements in imaging techniques and computational power have boosted the development of new technologies introduced in the surgical workflow to reduce the risks, improve precision and enhance the surgeons’ confidence. They can be used preoperatively to plan surgery, during the surgical procedure for guidance and to better identify anatomical structures, or after surgery to assess the intervention. The integration of these technologies in surgical procedures is known as computer-assisted surgery (CAS).
Surgical planning is the process through which the patient images are visualized prior to surgery to predefine the surgical steps. The different existing imaging modalities, such as computed tomography (CT) and magnetic resonance imaging (MRI), can be used for this purpose. Once preoperative planning has concluded, it can be transferred to the operating room to replicate the simulation precisely. Several tools are available to perform this translation, including computer-aided design and manufacturing (CAD/CAM), also known as 3D printing, surgical navigation with conventional tracking systems, or navigation based on augmented reality (AR) technology.
3D-printed anatomical models can be helpful in preoperative planning, surgical simulation, pre-surgical manipulation of surgical equipment, or education. However, the use of 3D printing in medicine is not restricted to the creation of anatomical models. CAD/CAM can also be used to design patient-specific implants, prostheses, splints, or external fixators. It can also produce customized surgical guides and instruments. These tools improve precision and enable a translation of the preoperative plan to the operating room.
Surgical navigation, also known as image-guided surgery, allows the surgeon to display the position of the surgical instruments with respect to the patient’s anatomy through virtual image overlays. The surgeon can use these systems to localize and visualize the anatomical target with precision and decide the path to follow with the navigated instrument to reach the surgical site safely. The position of the tracked instruments is projected onto the preoperative or intraoperative images. It can also be represented with respect to 3D rendered anatomical models, making visualization and guidance easier. Navigation can transform surgery into a safer and less invasive procedure. It also allows the performance of more daring procedures, as less accessible structures can be reached.
AR is the technology through which virtual (computer-generated) objects are overlayed on a user’s view of the real world to enhance it. These virtual objects appear to coexist in the same space as the elements from the physical world. The projection of the virtual elements is usually achieved using cameras, displays, projectors, trackers, or special equipment. AR in surgery has proved to be a helpful tool to improve safety and efficacy. It is primarily used to visualize delicate and vital structures, such as major vessels or nerves, or target elements such as tumors or bones. These anatomical models are projected directly onto the patient, and their visualization can be modified through interactive panels, buttons, gestures, or voice commands, depending on the device. One of the main advantages of this technology is the ability to visualize these elements without deviating the attention from the surgical field to look at an external screen. This behavior reduces time and makes it more intuitive than conventional visualization systems.
Initially, the application of technologies such as surgical navigation was restricted to neurosurgical procedures. In these interventions, accuracy is crucial to obtain satisfactory surgical outcomes and minimize the risks. Also, the surgical field is surrounded by the skull, a rigid structure that can be fixated to provide a well-defined and reliable reference frame necessary for precise surgical navigation. In the last decades, these tools have been translated to other disciplines such as orthopedic and maxillofacial oncology. Nevertheless, this translation requires an adaptation to each surgical scenario. These procedures mainly involve rigid bony structures, facilitating the translation of the preoperative plan. However, in most cases, the anatomical structures are mobile or cannot be fixed due to the surgical approach. Consequently, new solutions have been introduced involving dynamic reference frames and new registration methods.
Although many solutions have been proposed adapted for each surgical scenario, there is still room for improvement. In many cases, the proposed solutions are expensive, invasive, and entail a complex setup that increases operative time. We have to pursue feasible and convenient setups that do not impose a more invasive procedure and are accessible. Additionally, more studies should be performed to analyze their accuracy and define the optimal setups to ensure precision. The aim of this thesis was to tackle the limitations of current solutions and present new alternatives that meet these criteria, analyzing the accuracy provided and comparing the results with conventional methods.
We first focused on using 3D-printed patient-specific instruments (PSIs) as surgical guides for orthopedic oncology procedures. Their application is increasing as they are easily introduced in the surgical room and do not require extra hardware. However, they are designed with large sizes to ensure an adequate installation, requiring additional bone exposure and, therefore, increasing intraoperative time and invasiveness. Their use in the pelvis is of significant interest due to its large and complex shape, with adjacent delicate structures. However, no study measures the accuracy they provide in realistic scenarios with a sufficiently large number of cases.
In this thesis, we performed an experimental cadaveric study with a large sample to characterize the accuracy of PSIs in the four most common locations for pelvic osteotomies. We followed a realistic workflow for preoperative planning and intraoperative procedure. To minimize the invasiveness of previous solutions, we reduced the size of PSIs, limiting the bone exposure. Additionally, PSIs were fabricated in a biocompatible material with a desktop 3D printer to make them more accessible, as they present a significantly lower price than industrial 3D printers. We analyzed the placement errors, computing translations and rotations with respect to the planned position. These errors were represented in a local reference frame to improve their characterization and interpretation. We also measured the osteotomy errors resulting from PSIs’ incorrect placements. The results demonstrate similar accuracy in short osteotomies to previous studies with lower samples, performed with larger PSIs in less realistic setups. However, we recorded errors higher than 5 mm for larger osteotomies located in the ilium, which are not acceptable for oncologic resections.
An alternative to the use of PSIs in pelvic oncology is surgical navigation. It allows not only to guide osteotomies but also to identify other anatomical structures during the procedure, enhancing guidance and avoiding damaging vital structures. However, it introduces additional and expensive devices in the operating room. The patient-to-image registration step requires exposure of bone areas not necessarily involved in the procedure as anatomical landmarks and large surfaces are used, becoming more invasive and increasing intraoperative time. Also, installing a dynamic reference frame fixed to the patient’s bone increases the operative time and can be obtrusive.
To overcome these limitations, we propose the introduction of the small PSIs presented in our previous work to use them as artificial landmarks, limiting bone exposure. We also simplify the reference frame installation by including a socket in the PSIs design to introduce the dynamic reference frame, allowing to attach and detach it on request. This solution reduces the invasiveness, as well as the time and the obtrusiveness. We tested the accuracy and feasibility of the proposed setup in a cadaveric study. The installation of the dynamic reference frame demonstrated repeatability, which allows removing it without the need to repeat registration every time it is reinserted. We also analyzed the errors introduced by the PSIs and the navigation system combined and separately, defining the optimum configurations of PSIs for three common surgical scenarios to achieve high accuracy in the surgical area following a non-invasive setup. We concluded that adequate navigation accuracy and osteotomy errors below 2 mm could be achieved when at least two small PSIs are placed surrounding the target region. The highest errors measured with the data collected from the experiment were again a consequence of incorrect placements of PSIs located in the ilium.
The ilium region presents an extensive and regular shape that hinders the correct placement of PSIs and can lead to high deviations from the planned position. Whether PSIs are used independently or as a tool for surgical navigation, this uncertainty in their placement is the most critical source of error.
To prevent this, we proposed to use AR as a tool to guide PSIs placement. The planned position can be displayed on top of the patient’s bone, indicating the surgeon where to install the PSI. We developed an AR application based on detecting a 3D-printed optical marker placed on a PSI and tested it with four users who placed six pairs of PSIs in the ilium (one in the iliac crest and the other in the supra-acetabular region). Users placed the PSIs freehand, using a smartphone, and the HoloLens 2. The results demonstrate how AR can significantly reduce the risk of high placement errors, ensuring placements close to the target. Both devices present similar results, and the selection of one over the other is more subject to surgeons’ preferences. Additionally, we performed the experiment in two phantom versions, where one was the traditional version 3D-printed in PLA, and the other included a layer of silicone simulating tissue to provide realism. The results demonstrated that PSIs are easier to place in the non-realistic phantom, which should be considered for further studies focused on measuring the precision of these devices. We also studied the differences among cases where smaller PSIs located in smoother regions presented higher errors than larger ones placed in more distinct regions. Hence, there exists a trade-off between reducing invasiveness and ensuring precision for the design of PSIs.
Finally, with the same objective of optimizing the current technique, we wanted to apply these same technologies to another discipline where surgical guidance is also of great value to achieve an accurate oncological resection with low risks. This is the case of tumor resections in deep regions of the mouth where the visibility and access are limited and vital structures are close to the affected area. In these interventions, the patient’s head cannot be fixed as surgeons need to move it to access the surgical region. Therefore, alternative setups based on dynamic reference frames or new registration techniques must be applied. Some studies have presented solutions based on dynamic reference frames screwed to the head. However, these setups are invasive and use non-precise registration points.
For this thesis, we focused on a real clinical case of a patient presenting an adenoid cystic carcinoma in the hard palate. In this scenario, introducing surgical navigation enabled a more conservative approach and avoided the complete removal of the maxilla, minimizing the need for reconstruction. We studied the case and proposed three different non-invasive navigation solutions. Two solutions were based on an optical tracking system (OTS) with different registration methods and one on AR. All solutions used 3D-printed tools fabricated with desktop 3D printers to track the patient and surgical instruments. The accuracy of each system was measured with a 3D-printed patient-specific phantom. We obtained similar results for all solutions with errors below or close to 1 mm. Finally, one of the solutions based on the OTS was applied for surgical navigation during the intervention, although AR was also tested for visualization. The results obtained from the postoperative CT demonstrated high accuracy, with errors below 2 mm in 90% of the points recorded intraoperatively along the resection margins. The surgical outcomes demonstrated clear margins, and after two years, the patient was free of disease.
The four studies included in this thesis demonstrate the multiple options 3D printing presents in surgical scenarios. We have used 3D printing not only to fabricate surgical guides but also to create AR markers, fabricate patient-specific phantoms, create patient-specific splints, or manufacture other surgical tools such as reference frames to track the movements of the patient or surgical instruments. All these devices have been fabricated using a desktop 3D printer, obtaining accurate results at a lower cost than industrial 3D printers. These printers present an affordable price and can be included in the hospitals for faster production, allowing cost-effective on-demand manufacturing. However, 3D printing requires the introduction of well-trained engineers in the hospital and the implication of surgeons in the production process.
One factor worth mentioning is that the application of all these systems requires extra preoperative time from both engineers and surgeons to plan the surgery, develop the software, design the 3D-printed tools, fabricate, and test them. This is a common entry barrier for the use of these technologies. However, it presents several potential benefits. Intraoperative time is reduced, and the procedure is performed with a higher control without increasing the risk of complications.
To conclude, in this thesis we have proposed new setups for intraoperative navigation in two complex surgical scenarios for tumor resection. We analyzed their navigation precision, defining the optimal configurations to ensure accuracy. With this, we have demonstrated that computer-assisted surgery techniques can be integrated into the surgical workflow with accessible and non-invasive setups. Although these solutions have been applied to two specific scenarios, selected for their complexity, similar solutions could be translated to other procedures where CAS can improve surgical outcomes. These results are a step further towards optimizing the procedures and continue improving surgical outcomes in complex surgical scenarios.
