Radiomics and artificial intelligence for predicting pituitary neuroendocrine tumor consistency: a systematic review and meta-analysis.
Authors
Affiliations (7)
Affiliations (7)
- Department of Neurosurgery, Pontificia Universidad Javeriana, Bogotá, Colombia. [email protected].
- Neurosurgery Research Group, Pontificia Universidad Javeriana, Bogotá, Colombia. [email protected].
- Department of Neurosurgery, Hospital Universitario San Ignacio, Bogotá, Colombia. [email protected].
- Department of Neurosurgery, Pontificia Universidad Javeriana, Bogotá, Colombia.
- Neurosurgery Research Group, Pontificia Universidad Javeriana, Bogotá, Colombia.
- Department of Neurosurgery, Hospital Universitario San Ignacio, Bogotá, Colombia.
- Medical Student, Faculty of Medicine, Santa Casa de São Paulo School of Medical Sciences, São Paulo, Brasil.
Abstract
Pituitary neuroendocrine tumors (PitNETs) represent approximately 16% of primary brain tumors. Tumor consistency, whether soft or hard, directly affects surgical strategy, extent of resection, and risk of complications. This study aimed to perform a systematic review and meta-analysis on the use of artificial intelligence and radiomics to predict the consistency of PitNETs. A systematic search was conducted to identify studies that evaluated the prediction of PitNET consistency using radiomics and artificial intelligence algorithms. Study quality was assessed using the Radiomics Quality Score and the QUADAS-2 tool. Random-effects meta-analysis was conducted. Nine studies comprising 935 patients were included, with 67.8% having soft and 32.2% hard tumors. Radiomics models, using varied MRI protocols and primarily manual segmentation, selected an average of 10.4 features from 691.8 extracted. Pooled diagnostic performance showed an accuracy of 84%, sensitivity 86%, specificity 78%, and area under the curve 0.91. Considerable heterogeneity (I² up to 100%) was observed due to differences in MRI acquisition, segmentation, and subjective definitions of tumor consistency. None of the studies included external validation. Radiomics and artificial intelligence show potential in the preoperative prediction of PitNET consistency, which may assist surgical planning and reduce complications. However, methodological heterogeneity, subjective reference standards, and lack of external validation limit the generalizability of current results. Standardized imaging protocols and multicenter validation are required before clinical implementation.