1
0
Coturel AE y cols. Inteligencia arꢂꢃcial en la estadiꢃcación y toma de decisiones en cáncer esófago gástrico Rev Argent Cir. 2026;118(2):e-1963
detecꢀng distant metastases, CT remains the standard
In our series, staging was primarily based on
imaging modality, although its sensiꢀvity and speciꢁcity CT and PET-CT scans, since endoscopic ultrasound
for peritoneal involvement are moderate (66% and 77%, was not rouꢀnely available at the ꢀme of diagnosis.
respecꢀvely). Findings suggesꢀve of peritoneal seeding Even so, the chatbot achieved adequate concordance
include ascites, peritoneal nodules or plaques, bowel with the tumor board, suggesꢀng that these tools can
wall thickening, or peritoneal hyperenhancement.
In our pracꢀce, we used oral and intravenous not all the ideal diagnosꢀc methods are available.
contrast-enhanced CT scans for staging. Endoscopic Thisstudyhassomelimitaꢀons.Itsretrospecꢀve
be incorporated into real-world pracꢀce even when
ultrasound is not available in our seꢆng, and we design suggests heterogeneity and the possibility
do not rouꢀnely request PET-CT scans for gastric of data loss, along with variability in the reporꢀng of
tumors, although some paꢀents bring results from studies conducted at different centers, which hinders
other insꢀtuꢀons. We do request PET-CT scans more standardizaꢀon. Furthermore, the analysis was based
frequently for esophageal and GEJ cancers.
on wriꢄen reports rather than original images, which
Endoscopic ultrasound demonstrates high prevents the conversaꢀonal model from conducꢀng an
accuracy for staging esophageal cancer T-status, with a objecꢀve reassessment.
sensiꢀvity and speciꢁcity of 85% and 87%, respecꢀvely,
No paꢀents underwent endoscopic ultrasound
for idenꢀfying T1a lesions. The accuracy of endoscopic (EUS), the gold standard for assessing tumor depth and
ultrasound for detecꢀng lymph node involvement is lymph node involvement, which may have limited the
more variable; however, it can be signiꢁcantly enhanced accuracy of the cTNM staging in some cases. However,
2
1
by incorporaꢀng ꢁne-needle aspiraꢀon .
the limited availability of this technology in Argenꢀna
While CT is the most widely used imaging means that the results are pragmaꢀc and can be
modality for esophageal cancer, it lacks sensiꢀvity for extrapolated to most local centers. Furthermore, since
early-stage disease and demonstrates poor diagnosꢀc this is a single-center terꢀary-level study, the results
performance for lymph node assessment, with high cannot be generalized without external validaꢀon,
2
2
false-negaꢀve rates in the presence of micrometastases which is an essenꢀal requirement for AI models .
and false-posiꢀve rates due to concurrent infecꢀon or Finally, the small sample size limits the staꢀsꢀcal power
inflammaꢀon. PET-CT is more sensiꢀve for idenꢀfying and the ability to detect rare paꢄerns.
primary tumors, lymph node involvement, and distant
In conclusion, based on the results presented,
metastases in this type of cancer. However, the low it can be stated that AI, implemented through a chatbot
spaꢀal resoluꢀon makes it unreliable for assessing the trained using clinical guidelines, proved to be useful
extent and depth of the lesion (T). Regarding N-staging in supporꢀng staging and iniꢀal treatment planning
(
lymph nodes), PET-CT may face challenges in detecꢀng for gastroesophageal cancer. Despite the menꢀoned
lymph nodes in close proximity to the primary tumor, limitaꢀons, the study demonstrated high concordance
as it oꢈen fails to disꢀnguish them from the primary with the treatment decisions made by the treaꢀng
lesion. In addiꢀon, micrometastases may not uptake medical team. These ꢁndings posiꢀon conversaꢀonal
FDG. PET-CT scans perform best for M-staging models as promising and accessible tools, and their
(
metastasis), with a sensiꢀvity of 71% and a speciꢁcity integraꢀon should be further evaluated in prospecꢀve,
of 93%.
mulꢀcenter studies.
Referencias bibliográꢁcas /References
1
.
Marra A, Morganꢀ S, Pareja F, Campanella G, Bibeau F, FuchsT,
et al. Arꢀꢁcial intelligence entering the pathology arena in on-
cology: current applicaꢀons and future perspecꢀves. Ann Oncol.
J
Gastroenterol. 2025;31(24):108021. doi:10.3748/wjg.v31.
i24.108021.
4. Petracchi EJ, Olivieri SE, Varela J, Canullán CM, Zandalazini H,
Ocampo C, et al. Use of arꢀꢁcial intelligence in the detecꢀon
of the criꢀcal view of safety during laparoscopic cholecystec-
tomy. J Gastrointest Surg. 2024;28(6):877-9. doi:10.1016/j.gas-
sur.2024.03.018.
5. Pereyra L, Schloꢄmann F, Steinberg L, Lasa J. Colorectal Cancer
Prevenꢀon: Is Chat Generaꢀve Pretrained Transformer (Chat GPT)
ready to Assist Physicians in Determining Appropriate Scree-
ning and Surveillance Recommendaꢀons?.J Clin Gastroenterol.
2024;58(10):1022-7. doi:10.1097/MCG.0000000000001979.
2
025;36(7):712-25. doi:10.1016/j.annonc.2025.03.006.
2
.
Chong PL, Vaigeshwari V, Mohammed Reyasudin BK, Noor Hida-
yah BRA, Tatchanaamoorꢀ P, Yeow JA, et al. Integraꢀng arꢀꢁcial
intelligence in healthcare: applicaꢀons, challenges, and future
direcꢀons. Future Sci OA. 2025;11(1):2527505. doi:10.1080/205
6
5623.2025.2527505.
3
.
Chen ZL, Wang C, Wang F. Revoluꢀonizing gastroenterology and
hepatology with arꢀꢁcial intelligence: From precision diagnosis
to equitable healthcare through interdisciplinary pracꢀce. World