Same same but different: A Web-based deep learning application revealed classifying features for the histopathologic...

Area of Science:

• Neuroscience
• Computational Pathology
• Artificial Intelligence in Medicine

Background:

• Focal cortical dysplasia type IIb and cortical tuber of tuberous sclerosis complex are distinct malformations of cortical development.
• Histopathological review of these conditions presents diagnostic challenges due to overlapping features like neuronal dyslamination and balloon cells.
• Accurate differentiation is crucial for patient management and understanding disease mechanisms.

Purpose of the Study:

• To develop and train a convolutional neural network (CNN) for classifying focal cortical dysplasia type IIb and cortical tuber.
• To visualize the morphological features utilized by the CNN for classification.
• To create a proof-of-concept web-based deep learning application for routine histopathologic use.

Main Methods:

• A digital processing pipeline was used to analyze 56 cases, generating numerous regions of interest and subsamples for CNN training.
• Guided gradient-weighted class activation maps (Guided Grad-CAMs) were employed to visualize CNN decision-making processes.
• A classification score was derived from identified patterns and validated by expert and non-expert neuropathologists.

Main Results:

• The best-performing CNN achieved 91% accuracy and an 0.88 area under the ROC curve on an unseen test set.
• Guided Grad-CAMs revealed novel histopathologic patterns aiding in distinguishing the two entities.
• The classification score significantly improved diagnostic performance, particularly for non-experts.

Conclusions:

• Deep learning, particularly CNNs with visualization techniques, can effectively differentiate challenging histopathological entities like focal cortical dysplasia IIb and tuberous sclerosis complex.
• The developed web application demonstrates the potential for integrating AI into routine pathology workflows.
• This approach enhances diagnostic accuracy and aids in the understanding of rare brain malformations.