Artikel
Automated detection and segmentation of focal cortical dysplasias (FCDs) – presentation of a novel artificial neural network with a prospective clinical validation
Automatisierte Erkennung und Segmentation von fokalen kortikalen Dysplasien (FCDs) – Vorstellung eines neuartigen künstlichen neuronalen Netzwerks und dessen prospektiver klinischer Validierung
Suche in Medline nach
Autoren
Veröffentlicht: | 4. Juni 2021 |
---|
Gliederung
Text
Objective: Focal cortical dysplasias (FCDs) are highly epileptogenic lesions frequently accounting for pharmaco-resistant focal epilepsy. Although MRI techniques have improved significantly over the past years, FCD detection remains challenging, as FCDs vary in location, size, and shape and commonly blend into surrounding tissues without clear definable boundaries. We developed a novel convolutional neural network for FCD detection and segmentation and validated it prospectively on daily-routine MRIs.
Methods: The neural network was trained on 201 T1 and FLAIR 3T MRI volume sequences of 158 patients with at least one FCD, regardless of type. Non-FCD MRIs, drawn from 100 normal MRIs and 50 MRIs with non-FCD pathologies, were added to the training. We applied the algorithm prospectively on 100 consecutive MRIs of patients with focal epilepsy from daily clinical practice. The results were compared with associated neuroradiological reports and morphometric analyses evaluated by an experienced epileptologist.
Results: Best training results reached a sensitivity (recall) of 70.1% and a precision of 54.3% for detecting FCDs. Applied on the daily-routine MRIs, 7 out of 9 FCDs were detected and segmented correctly with a sensitivity of 77.8% and a specificity of 5.5%. The results of conventional visual analyses were 33.3% and 94.5%, respectively (3/9 FCDs detected); the results of morphometric analyses with overall epileptologic evaluation were both 100% (9/9 FCDs detected) and thus served as reference.
Conclusion: We developed a 3D convolutional neural network with autoencoder regularization for FCD detection and segmentation. Our algorithm employs the largest FCD training cohort to date consisting of 201 MRIs with various types of FCDs. The higher sensitivity in detecting and segmenting FCDs compared to visual analyses allows our algorithm to be used as a convenient FCD pre-screening tool. However, its current low specificity, leading to a high rate of false positive predictions, still calls for validation by visual and morphometric analyses as well as the need for further algorithm training.
Figure 1 [Fig. 1]