Diagnostic support in pediatric craniopharyngioma using deep learning.

PURPOSE: We studied a pediatric group of patients with sellar-suprasellar tumors, aiming to develop a convolutional deep learning algorithm for radiological assistance to classify them into their respective cohort. METHODS: T1w and T2w preoperative magnetic resonance images of 226 Chilean patients were collected at the Institute of Neurosurgery Dr. Alfonso Asenjo (INCA), which were divided into three classes: healthy control (68 subjects), craniopharyngioma (58 subjects) and differential sellar/suprasellar tumors (100 subjects). RESULTS: The PPV among classes was 0.828±0.039, and the NPV was 0.919±0.063. Also explainable artificial intelligence (XAI) was used, finding that structures that are relevant during diagnosis and radiological evaluation highly influence the decision-making process of the machine. CONCLUSION: This is the first experience of this kind of study in our institution, and it led to promising results on the task of radiological diagnostic support based on explainable artificial intelligence (AI) and deep learning models.

Utilidad del complejo posterior para el diagnóstico prenatal de sintelencefalia / Utility of the posterior complex for the prenatal diagnosis of syntelencephaly

Introducción y objetivo: Demostrar el valor del plano axial del complejo posterior, como apoyo a la detección antenatal de sintelencefalia, variante de holoprosencefalia. Método: Se incluyeron todas las pacientes con diagnóstico de sintelencefalia evaluadas desde el año 2008. En todos los casos se consignaron los datos clínicos, de neurosonografía (NSG), de resonancia magnética (RM) y genética. Resultados: Cuatro casos fueron diagnosticados en el segundo trimestre y en todos se realizó estudio genético y RM. Tres tuvieron en su evolución anomalías extra-SNC y dos de ellos alteraciones cromosómicas, una de ellas incompatible con la vida extrauterina. Lo hallazgos descritos en neuroimagen para esta afección fueron detectados en la NSG, con una excelente correlación con RM, ya fuera esta última realizada en periodo fetal o posnatal. Conclusión: El diagnóstico prenatal de variantes de holoprosencefalia es difícil, considerando la existencia de una fusión medial más acotada que en las formas clásicas. El presente estudio demuestra la utilidad del plano del complejo posterior para la sospecha diagnóstica de sintelencefalia.

Introduction and objective: To demonstrate the value of the axial plane of the posterior complex, as a clue for the antenatal detection of synthelencephaly, a variant of holoprosencephaly. Method: All patients diagnosed with syntelencephaly evaluated since 2008 were included. In all cases, clinical, neurosonography (NSG), magnetic resonance imaging (MRI) and genetic data were recorded. Results: Four cases were diagnosed in the second trimester and in all of them a genetic study and MRI were performed. Three had extra-CNS anomalies in their evolution and two of them chromosomal anomalies, one of them incompatible with extrauterine life. Neuroimaging findings described for this condition were detected by NSG, with an excellent correlation with MRI, whether the latter was performed in the fetal or postnatal period. Conclusion: The prenatal diagnosis of holoprosencephaly variants is difficult, considering the existence of a more limited medial fusion than in the classical forms. The present study demonstrates the usefulness of the posterior complex plane for the diagnostic suspicion of synthelencephaly.