*Artificial Intelligence in Surgery, Surgical Subspecialties, and Related Disciplines DOI: http://dx.doi.org/10.5772/intechopen.112691*

potentially important role in intraoperative histological analysis. The current procedure/workflow for intraoperative pathology in many oncologic surgeries involves the excision of a portion of the tumor, where the sample is then transported to the laboratory for preparation and interpretation by a pathologist. This process can take 20–30 min, prolonging the overall surgical procedure and also potentially delaying the diagnosis, where each additional step also contributes potential barriers to timely diagnosis [60]. Applications of "computer vision" could potentially address challenges associated with intraoperative interpretation of histology. Data are also emerging on the use of ML algorithms in analyzing images from Raman spectroscopy to identify malignant and benign tumors. The actual algorithm is functionally similar to the process used in radiologic analyses, but Raman spectroscopy imaging can be further processed to provide imaging more similar to hematoxylin and eosin (H&E) staining, which may better allow surgeons and pathologists to verify ML classifications of tissue samples [61]. Intraoperative pathology consultations are quite common in neurosurgical tumor procedures, breast cancer, hepatobiliary and pancreatic resections, lymph node dissections, and dermatopathology [62–66]. These procedures may also benefit from AI-aided streamlining of intraoperative histology and pathology in the future.

The use of computer vision algorithms in surgery can be further expanded to include the characterization of molecular tissue margins. When removing malignant tumors, patient outcomes are optimal with maximal resection of the tumor while sparing as much healthy tissue as possible. Positive margins, or cancerous cells that remain after incomplete resection, are associated with recurrence of cancer, leading to worse patient outcomes. Some estimates indicate that positive margins may be found in approximately 5% of liver and breast cancer resections, so identification of tumor margins is still a significant problem that must be addressed [67, 68]. As mentioned previously, Raman spectroscopy has already been used by pathologists to distinguish neoplastic and normal tissue based on differential Raman scattering, but future advancements could also lead to intraoperative Raman spectroscopy to determine tumor margins [69]. Like with other imaging modalities, computer vision algorithms in the future will be able to identify features such as positive margins. This could allow surgeons to identify tumor margins within the operating room without needing to




**Table 2.**

*Summary of included studies on intraoperative artificial intelligence/machine learning (AI/ML).*

wait for margins to be identified histologically, increasing efficiency and outcomes of tumor resection surgeries (**Table 2**).
