Page 112 - Read Online
P. 112
Shen et al. Art Int Surg. 2025;5:150-9 https://dx.doi.org/10.20517/ais.2024.71 Page 154
routes directly on the patient's abdomen, effectively remedying the limitations with perforator depth and its
[16]
respective course changes mentioned by DeFazio et al. . While the sample size was limited to 5 cases,
preoperative perforator route identification was successful for all patients, as verified by Doppler. Map
alignment time may prove to be prohibitive to large-scale implementation, but other studies using
coordinate systems centered around the umbilicus have shown high accuracy without the same tedious
[23]
alignment process .
The economic considerations of implementing 3D modeling as an intraoperative aid vary significantly
between physical and virtual approaches. Ghasroddashti et al. reported that the cost of printing each 3DVM
differed significantly, ranging from $6.50 to $829.72 . While plastic and reconstructive surgery lacks formal
[24]
cost-benefit analyses specific to the field, insights can be drawn from other surgical specialties. Ravi et al.
and Ballard et al. have demonstrated that within the field of general surgery, orthopedic, and maxillofacial
surgery, the reduced operative times do generally outweigh the model production costs [25,26] . In contrast,
virtual 3D modeling approaches, particularly those utilizing DIRT technology, show promising cost
advantages. The growing adoption of DIRT can be attributed to its operational simplicity and potential for
further cost reduction through integration with smartphone-based cameras, though comprehensive cost
analyses for virtual modeling implementations remain limited .
[27]
AI TOOLS TO ENHANCE POSTOPERATIVE OUTCOMES
Symptom monitoring and postoperative recovery
AI is rapidly revolutionizing breast reconstruction procedures by transitioning postoperative care from
traditional in-person flap checks and clinical visits to portable and ambient tools, providing closer
monitoring, quicker responses, and improved surgical outcomes. One such application is the use of
smartphone-based tools for managing and monitoring microvascular flaps . Supervised ML was able to
[28]
predict vascular compromise in free flaps with a high accuracy of 98.4% using images taken on a
smartphone and clinical variables . In the study by Hsu et al., a deep learning model integrated
[29]
smartphone application was used for free flap monitoring, which demonstrated high accuracy (95.3%),
sensitivity (95.2%), and specificity (95.3%) in predicting venous congestion in free flaps, with an area under
the curve being 0.99 . More importantly, in several cases, the smartphone application was able to detect an
[30]
increase in congestion probability on average two hours earlier than human observers. Another study by
Kim et al. similarly harnessed the power of CNNs and the convenience of smartphone cameras for free flap
monitoring . Their best model detected venous insufficiency and arterial insufficiency with sensitivity of
[31]
97.5% and 92.8%, respectively. Pilot implementation of a smartphone application equipped with the same
AI algorithms on 10 patients demonstrated great utility in the clinical setting, with prompt feedback in less
than one second. In autologous breast reconstruction, the time at which potential flap compromise is
detected may determine what surgical or medical interventions are possible, and the likelihood of those
treatments working to salvage a flap from progressing to failure. The current practice of detecting free flap
failure primarily involves frequent clinical examination of the flap combined with Doppler ultrasound to
monitor blood flow. This process requires specially trained medical professionals and needs to be done in
the medical setting, which is laborious and resource-intensive. Digital monitoring tools, especially when
equipped with predictive ML algorithms to identify early risk factors, could provide cost-effective, real-time,
and non-invasive monitoring. These emerging applications of AI will not only allow for prompt
intervention by plastic surgeons to improve flap survival, but also enhance patient outcomes and optimize
resource allocation within the healthcare system.
While smartphone-based tools and ML show promising clinical utility, comprehensive cost analysis specific
to those technologies when used in postoperative flap monitoring is limited. Lee et al. noted a significant