GradCAM as an explicability method to evaluate the performance of deep learning models in classifying pediatric arteriovenous malformations (AVM) in arterial spin labeling sequences (ASL)
Por:
Romagosa, J, Mata, C, Benítez, R, Valls-Esteve, A, Bernaus, S, Ibnoulkhatib, M, Stephan-Otto, C, Munuera, J
Publicada:
20 oct 2025
Resumen:
PurposeThe study investigates the usefulness of Convolutional Neural Networks (CNNs) in accurately detecting arteriovenous malformations in pediatric medical imaging, particularly using arterial spin labeling sequences. It also aims to offer diagnostic explanations comparable to expert analysis.MethodsThe research analyzed three different CNN architectures to determine their performance in detecting arteriovenous malformations. The study focused on evaluating the relationship between model complexity and performance increase, using data to assess the accuracy and diagnostic usefulness of each model.ResultsThe findings indicated a nonlinear link between model complexity and performance. Sur- prisingly, more complex models frequently produced poor results and diagnostically useless answers. The simplest CNN models achieved the highest accuracy rate (90%), demonstrating the effectiveness of minimal complexity in model construction. Heat maps showed a strong association with the real locations of irregularities, indicating that the models were interpretable.ConclusionThe study highlights the usefulness of CNNs in medical diagnostics, emphasizing the importance of model simplicity and interpretability in clinical applications. It suggests a need for balancing technical sophistication with clinical value and presents options for future research into refining CNN structures for increased diagnostic precision in various medical imaging modalities.
Filiaciones:
Romagosa, J:
Inst Recerca Sant Joan de Deu, Pediat Computat Imaging Ctr, Esplugas de Llobregat, Spain
Mata, C:
Inst Recerca Sant Joan de Deu, Pediat Computat Imaging Ctr, Esplugas de Llobregat, Spain
Univ Politecn Cataluna, Inst Res & Innovat Hlth IRIS, BIOCOM SC, Barcelona, Spain
Benítez, R:
Inst Recerca Sant Joan de Deu, Pediat Computat Imaging Ctr, Esplugas de Llobregat, Spain
Univ Politecn Cataluna, Inst Res & Innovat Hlth IRIS, BIOCOM SC, Barcelona, Spain
Valls-Esteve, A:
Inst Recerca Sant Joan de Deu, Pediat Computat Imaging Ctr, Esplugas de Llobregat, Spain
Hosp San Juan Dios, Innovat Dept, Esplugues Del Llobregat, Spain
Bernaus, S:
Inst Recerca Sant Joan de Deu, Pediat Computat Imaging Ctr, Esplugas de Llobregat, Spain
Ibnoulkhatib, M:
Hosp San Juan Dios, Dept Diagnost Imaging, Esplugues Del Llobregat, Spain
Stephan-Otto, C:
Inst Recerca Sant Joan de Deu, Pediat Computat Imaging Ctr, Esplugas de Llobregat, Spain
Ctr Invest Biomed Red Salud Mental, Madrid, Spain
Munuera, J:
Inst Recerca Sant Pau, Adv Med Imaging Artificial Intelligence & Imaging, Barcelona, Spain
Hosp Santa Creu & Sant Pau, Diagnost Imaging Dept, Barcelona, Spain
Green Accepted
|