RESUMEN:
Objetivo: El objetivo del estudio fue evaluar el rendimiento diagnóstico de los predictores radiómicos F18-FDG PET/CT en la respuesta patológica completa a la quimioterapia neoadyuvante en pacientes con cáncer de mama. Material y métodos: Se utilizaron 127 registros para la conformación de un simulacro de datos y 19 fueron incluidos en el análisis final. Se analizó edad, respuesta al tratamiento y las características radiómicas. Resultados: La distribución de la respuesta al tratamiento fue parcial en 13 pacientes (68.42%), respuesta total en tres pacientes (15.79%), estabilidad en dos pacientes (19.53%) y progresión en una paciente (5.26%). Se construyó un conjunto empaquetado de un modelo LightGBM en 3 pliegues dadas 85 características originales. Conclusión: Se demostró un rendimiento diagnóstico de las características radiómicas superior al estándar de oro. Aunque aún hay que demostrar la utilidad de este enfoque en la predicción correcta y temprana del tratamiento, este estudio sugiere que las características radiómicas de textura obtenidas por PET/CT pueden ser útiles en la predicción de la respuesta al tratamiento.

ABSTRACT:
Objective: The objective of the study was to evaluate the diagnostic performance of the radiomic predictors F18-FDG PET/CT in the complete pathological response to neoadjuvant chemotherapy in patients with breast cancer. Material and methods: A total of 127 records were used to make up a mock data set and 19 were included in the final analysis. Age, response to treatment and radiomic characteristics were analyzed. Results: The distribution of response to treatment was partial in 13 patients (68.42%), total response in three patients (15.79%), stability in two patients (19.53%) and progression in one patient (5.26 %). A packed ensemble of a LightGBM model was constructed in 3 folds given 85 original features from the input dataset. Conclusion: Diagnostic performance of radiomics features superior to the gold standard was shown. Although the usefulness of this approach in correct and early prediction of treatment has yet to be demonstrated, this study suggests that texture radiomics obtained by PET/CT may be useful in predicting response to treatment.

ABSTRACT:
Objective: The objective of the study was to evaluate the diagnostic performance of the radiomic predictors F18-FDG PET/CT in the complete pathological response to neoadjuvant chemotherapy in patients with breast cancer. Material and methods: A total of 127 records were used to make up a mock data set and 19 were included in the final analysis. Age, response to treatment and radiomic characteristics were analyzed. Results: The distribution of response to treatment was partial in 13 patients (68.42%), total response in three patients (15.79%), stability in two patients (19.53%) and progression in one patient (5.26 %). A packed ensemble of a LightGBM model was constructed in 3 folds given 85 original features from the input dataset. Conclusion: Diagnostic performance of radiomics features superior to the gold standard was shown. Although the usefulness of this approach in correct and early prediction of treatment has yet to be demonstrated, this study suggests that texture radiomics obtained by PET/CT may be useful in predicting response to treatment.