Epilepsy is a neurological disorder caused by abnormal electrical discharges in the brain that can be resistant to medication, making an accurate diagnosis essential to determine other treatment options, such as surgery. This study has shown that artificial intelligence could be a powerful tool to improve the detection of focal cortical dysplasia, a type of epilepsy, on MRI scans.
Epilepsy is a neurological disorder characterized by recurrent seizures caused by abnormal electrical discharges in the brain. These seizures can range from brief lapses of consciousness to more intense convulsions.
The condition can have many causes, including brain injuries, genetics, and malformations in brain development. In some cases, epilepsy can be resistant to medication, making an accurate diagnosis essential to determine other treatment options, such as surgery.
This study investigated whether artificial intelligence (AI) can improve the diagnosis of a condition called focal cortical dysplasia (FCD), a leading cause of drug-resistant focal epilepsy. These brain injuries are often difficult to detect on MRI scans, making diagnosis and appropriate treatment difficult.
Researchers used an advanced neural network called MELD Graph, a type of artificial intelligence that analyzes MRI images in detail.
This algorithm was able to identify 64% of lesions that had previously gone unnoticed by human radiologists. It also provided detailed reports on the location, size and shape of the lesions, increasing the reliability of detection.
Focal cortical dysplasia can be treated surgically in some cases, but for this to happen, an accurate diagnosis is essential.
Since artificial intelligence has been shown to be more effective than conventional methods in identifying these lesions, its application can facilitate earlier diagnoses and improve surgical planning. This could lead to better outcomes for patients with focal epilepsy, increasing their chances of seizure control.
The researchers analyzed MRI data from 703 patients with epilepsy caused by focal cortical dysplasia, collected from 23 medical centers around the world between 2018 and 2022.
They trained the AI using data from 20 of these centers and then tested its effectiveness in three different centers to ensure the model was reliable in a variety of scenarios.
The main results showed that MELD Graph was 81.6% accurate in identifying lesions in patients who had undergone surgery and had been seizure-free for one year. For patients whose lesions were not clearly visible on conventional scans, the algorithm was 63.7% accurate.
Compared to another existing algorithm, MELD Graph demonstrated a higher positive predictive value (PPV), meaning that when it did indicate the presence of a lesion, it was more reliable.
This study showed that AI can be a powerful tool for improving the detection of focal cortical dysplasia on MRI scans. Because artificial intelligence is able to identify lesions that often go unnoticed, its implementation could help doctors diagnose this condition more accurately, allowing for more effective treatment for patients with focal epilepsy.
This figure illustrates how artificial intelligence (AI) can help identify hidden brain lesions in patients with epilepsy. (A) Dataset: The study analyzed brain scans from 703 patients with epilepsy and 482 healthy individuals, collected from 23 centers around the world. (B) AI training: A neural network called MELD Graph was trained to recognize patterns in MRI scans and compare its predictions with confirmed lesions. (C) AI testing: MELD Graph detected more lesions and generated fewer false positives than a previous model (Multilayer Perceptron), making its diagnosis more reliable. (D) New patient scan: The AI can highlight suspicious areas on scans, helping doctors identify lesions that might otherwise have gone unnoticed. (E) Interpretable results: The system provides detailed information such as lesion size and confidence level of prediction, as well as features such as brain substance contrast and cortex thickness.
LEIA MAIS:
Detection of Epileptogenic Focal Cortical Dysplasia Using Graph Neural Networks: A MELD Study
Mathilde Ripart, Hannah Spitzer, Logan ZJ Williams, Lennart Walger, Andrew Chen, Antonio Napolitano, Camilla Rossi-Espagnet, Stephen T. Foldes, Wenhan Hu, Jiajie Mo, Marcus Likeman, Theodor Rüber, Maria Eugenia Caligiuri, Antonio Gambardella, Christopher Guttler, Anna Tietze, Matteo Lenge, Renzo Guerrini, Nathan T. Cohen, Irene Wang, Ane Kloster, Lars H. Pinborg, Khalid Hamandi, Graeme Jackson, Domenico Tortora, Martin Tisdall, Estefania Conde-Blanco, Jose C. Pariente, Carmen Perez-Enriquez, Sofia Gonzalez-Ortiz, Nandini Mullatti, Katy Vecchiato, Yawu Liu, Reetta Kalviainen, Drahoslav Sokol, Jay Shetty, Benjamin Sinclair, Lucy Vivash, Anna Willard, Gavin P. Winston, Clarissa Yasuda, Fernando Cendes, Russell T. Shinohara, John S. Duncan, J. Helen Cross, Torsten Baldeweg, Emma C. Robinson, Juan Eugenio Iglesias, Sophie Adler, Konrad Wagstyl, MELD FCD writing group, Abdulah Fawaz, Alessandro De Benedictis, Luca De Palma, Kai Zhang, Angelo Labate, Carmen Barba, Xiaozhen You, William D. Gaillard, Yingying Tang, Shan Wang, Shirin Davies, Mira Semmelroch, Mariasavina Severino, Pasquale Striano, Aswin Chari, Felice D’Arco, Kshitij Mankad, Nuria Bargallo, Saul Pascual-Diaz, Ignacio Delgado-Martinez, Jonathan O’Muircheartaigh, Eugenio Abela, Jothy Kandasamy, Ailsa McLellan, Patricia Desmond, Elaine Lui, Terence J. O’Brien and Kirstie Whitaker
JAMA Neurology, 24 February 2025
DOI: 10.1001/jamaneurol.2024.5406
Abstract:
A leading cause of surgically remediable, drug-resistant focal epilepsy is focal cortical dysplasia (FCD). FCD is challenging to visualize and often considered magnetic resonance imaging (MRI) negative. Existing automated methods for FCD detection are limited by high numbers of false-positive predictions, hampering their clinical utility. To evaluate the efficacy and interpretability of graph neural networks in automatically detecting FCD lesions on MRI scans. In this multicenter diagnostic study, retrospective MRI data were collated from 23 epilepsy centers worldwide between 2018 and 2022, as part of the Multicenter Epilepsy Lesion Detection (MELD) Project, and analyzed in 2023. Data from 20 centers were split equally into training and testing cohorts, with data from 3 centers withheld for site-independent testing. A graph neural network (MELD Graph) was trained to identify FCD on surface-based features. Network performance was compared with an existing algorithm. Feature analysis, saliencies, and confidence scores were used to interpret network predictions. In total, 34 surface-based MRI features and manual lesion masks were collated from participants, 703 patients with FCD–related epilepsy and 482 controls, and 57 participants were excluded during MRI quality control. Sensitivity, specificity, and positive predictive value (PPV) of automatically identified lesions. In the test dataset, the MELD Graph had a sensitivity of 81.6% in histopathologically confirmed patients seizure-free 1 year after surgery and 63.7% in MRI–negative patients with FCD. The PPV of putative lesions from the 260 patients in the test dataset (125 female [48%] and 135 male [52%]; mean age, 18.0 [IQR, 11.0-29.0] years) was 67% (70% sensitivity; 60% specificity), compared with 39% (67% sensitivity; 54% specificity) using an existing baseline algorithm. In the independent test cohort (116 patients; 62 female [53%] and 54 male [47%]; mean age, 22.5 [IQR, 13.5-27.5] years), the PPV was 76% (72% sensitivity; 56% specificity), compared with 46% (77% sensitivity; 47% specificity) using the baseline algorithm. Interpretable reports characterize lesion location, size, confidence, and salient features. In this study, the MELD Graph represented a state-of-the-art, openly available, and interpretable tool for FCD detection on MRI scans with significant improvements in PPV. Its clinical implementation holds promise for early diagnosis and improved management of focal epilepsy, potentially leading to better patient outcomes.
Commentaires