World journal of urology · 2026
AI-driven model for early failure prediction after HIFU integrating immediate post-ablation ultrasound and clinical data.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Cella Ludovica, Avolio Pier Paolo, Mascaro Luca, Loiacono Daniele, Paciotti Marco, Fasulo Vittorio, Piccolini Andrea, Saita Alberto, Lazzeri Massimo, Buffi Nicolò Maria, Casale Paolo, Lughezzani Giovanni
Аннотация
PURPOSE: High-Intensity Focused Ultrasound (HIFU) is an emerging focal therapy for localized prostate cancer, offering an alternative to radical prostatectomy and radiotherapy while balancing oncological control and functional preservation. However, early treatment failure remains a critical challenge, and tools to support early risk stratification are lacking. We developed a two-stage model integrating immediate post-ablation ultrasound with routinely available clinical variables to predict early recurrence (≤ 12 months) after HIFU.
METHODS: Seventy-two patients treated with HIFU were analyzed. In Stage 1, a deep-learning model (DenseNet121) analyzed immediate post-ablation ultrasound images to generate a patient-level probability of treatment failure. In Stage 2, this image-derived probability was combined with routinely available clinical variables (age, PSA density, ISUP grade, percentage of positive biopsy cores, and treated volume) in a Support Vector Machine (SVM) to predict early treatment failure. Model performance was evaluated using cross-validation.
RESULTS: The integrated model achieved the highest predictive accuracy, with an area under the receiver-operating characteristic curve of 0.79, compared with 0.75 for the model based on clinical variables alone and 0.65 for the model based on imaging alone. Clinical variables accounted for most of the predictive signal, with the ultrasound-derived probability providing a modest but consistent additional value, particularly in larger treated volumes (≥ 5.7 cc).
CONCLUSION: This two-stage framework combining deep learning and clinical variables improves prediction of early treatment failure after HIFU, supporting point-of-care risk stratification and individualized follow-up.
METHODS: Seventy-two patients treated with HIFU were analyzed. In Stage 1, a deep-learning model (DenseNet121) analyzed immediate post-ablation ultrasound images to generate a patient-level probability of treatment failure. In Stage 2, this image-derived probability was combined with routinely available clinical variables (age, PSA density, ISUP grade, percentage of positive biopsy cores, and treated volume) in a Support Vector Machine (SVM) to predict early treatment failure. Model performance was evaluated using cross-validation.
RESULTS: The integrated model achieved the highest predictive accuracy, with an area under the receiver-operating characteristic curve of 0.79, compared with 0.75 for the model based on clinical variables alone and 0.65 for the model based on imaging alone. Clinical variables accounted for most of the predictive signal, with the ultrasound-derived probability providing a modest but consistent additional value, particularly in larger treated volumes (≥ 5.7 cc).
CONCLUSION: This two-stage framework combining deep learning and clinical variables improves prediction of early treatment failure after HIFU, supporting point-of-care risk stratification and individualized follow-up.