Page 29 - AIH-2-2

Artificial Intelligence in Health Machine learning in arthroplasty

adverse events, blood transfusions, and extended hospital its ability to be integrated uniformly across medical
stays maximized at 95.4%, 95.6%, and 82.3%, respectively. centers. Furthermore, ongoing validation, computational
78
In addition to adverse intra- and perioperative events, upgrades, and integration with existing systems can add
ML was also used to address patient discharge. Lopez to a health system’s overall cost. In addition, ensuring
et al. created a boosted decision tree and an ANN to regulatory compliance for AI in healthcare adds complexity
79
anticipate non-home discharges following elective TSA. and expense to already burdened compliance personnel.
The boosted decision tree displayed an AUROC of 0.788 While the adoption of AI in orthopedic surgery involves
and an accuracy of 90.3%, while the ANN had an AUROC significant upfront costs and training, the technology offers
of 0.851 with an accuracy of 89.9%. Table 3 summarizes a horizon of benefits. AI has the potential to revolutionize
79
the results of key studies investigating the application of AI orthopedic care. As the technology continues to evolve
in predicting complications in patients undergoing TSA. and become more integrated into healthcare systems, its
4.3. Future direction cost-effectiveness and value in improving patient care are
expected to become even more pronounced.
The use of AI, specifically ML, has demonstrated
encouraging benefits when applied to TSA. Preparation There is potential for AI to be integrated into modern
is paramount to successful operations, and AI can aid orthopedic clinical practices. For example, it can be used to
in pre-operative assessment through shoulder implant predict post-operative clinical outcomes and complication
identification in revision cases and anatomic planning for rates for patients. This may be useful in patient counseling
primary operations. These advancements enable surgeons on expectations following surgery. A surgeon can input
a patient’s information and the algorithm may be able to
to increase efficiency and potentially reduce costs to predict if a patient will have full forward elevation, limited
patients by reducing the use of unnecessary equipment. internal rotation, and the percentage of chance of specific
Furthermore, AI use has branched to surgical outcomes
prediction, perioperative complications, hospital stay complications. Thus, it will be a great tool when advising
patients on expectations after surgery.
length, functional improvement following TSA, and cost
burden on patients. Employing these tools enables surgeons 5. Ethical considerations
to provide transparent information to their patients, thus
empowering patients to make informed decisions that are Although AI usage in healthcare and surgery offers many
suited for them and their families. opportunities for optimizing physician preparation and
patient outcomes, several pitfalls exist in merging AI with
The application of AI in TSA has also opened doors medicine, providing ethical considerations that require
to the use of this technology beyond arthroplasty. For sincere attention. Healthcare data represents roughly 30%
example, robotic surgery is one of the many sectors within of the world’s data, with its compound annual growth rate
surgery that has begun to adopt AI technology. More projected to reach 36% by 2025. As health data continues
84
recently, it has been shown that AI can be used to identify to grow, so does the risk of compromising patient
“surgical planes” by differentiating tissue types. While confidentiality and anonymity. Capelli et al. highlighted
80
84
tissue recognition has yet to transition into orthopedic that AI functionality has expanded to potentially
arthroplasties, it can be hypothesized that this technology re-identify patient data that were previously scrubbed of all
could greatly assist surgeons, especially when attempting to identifying information. Furthermore, understanding the
identify critical neurovascular structures. Another aspect complex algorithms of AI and how outputs and decisions
of ML being applied to robotic surgery is intraoperative are derived can be difficult for humans, as the process
assessments. By merging automated performance for generating output is unclear. 84,85 This obscurity raises
metrics with ML, objective assessments can be utilized ethical issues, as it is difficult for bias evaluation and clinical
to offer standardized insights into the development of experience application to evaluate AI programs’ logic, and
a surgeon’s skills. Arthroplasty-derived innovations inform patients of the decision-making process. 84-86
81
have contributed to robotic surgery’s usage of AI. Image, When applying these concepts to the use of AI and ML
denoizing, deblurring, and color-correcting have proven in arthroplasty, a significant concern is understanding the
to be effective tools for arthroplasty, and thus, similar ML decision-making process involved in evaluating patients
models have been created for robotic surgery. 82,83
for arthroplasty. It is essential that surgeons’ expertise
However, implementing AI technology in orthopedic informs and guides AI computations. Until these processes
surgery entails significant initial investments in equipment, are perfected and clearly understood by surgeons, there is
software, and training. Surgeons may be apprehensive a risk of patients being excluded from the decision-making
about the required education to effectively use AI, limiting process of their surgical care.

Volume 2 Issue 2 (2025) 23 doi: 10.36922/aih.3278