The impact of transfer learning on 3D deep learning convolutional neural network segmentation of the hippocampus in mild cognitive impairment and Alzheimer disease subjects

• Published in: Human brain mapping Vol. 43; no. 11; pp. 3427 - 3438
• Main Authors: Balboni, Erica, Nocetti, Luca, Carbone, Chiara, Dinsdale, Nicola, Genovese, Maurilio, Guidi, Gabriele, Malagoli, Marcella, Chiari, Annalisa, Namburete, Ana I. L., Jenkinson, Mark, Zamboni, Giovanna
• Format: Journal Article
• Language: English
• Published: Hoboken, USA John Wiley & Sons, Inc 01.08.2022
• Subjects: Abnormalities; Alzheimer disease; Alzheimer's disease; Artificial neural networks; Biomarkers; Cognitive ability; Datasets; Deep learning; Domains; Hippocampus; Image processing; Image segmentation; Impairment; Magnetic resonance imaging; Medical imaging; mild cognitive impairment; Neural networks; Neurodegenerative diseases; Training; Transfer learning; deep learning; transfer learning; magnetic resonance imaging; hippocampus; Alzheimer disease; neural networks; mild cognitive impairment
• ISSN: 1065-9471; 1097-0193; 1097-0193
• DOI: 10.1002/hbm.25858

Research on segmentation of the hippocampus in magnetic resonance images through deep learning convolutional neural networks (CNNs) shows promising results, suggesting that these methods can identify small structural abnormalities of the hippocampus, which are among the earliest and most frequent brain changes associated with Alzheimer disease (AD). However, CNNs typically achieve the highest accuracy on datasets acquired from the same domain as the training dataset. Transfer learning allows domain adaptation through further training on a limited dataset. In this study, we applied transfer learning on a network called spatial warping network segmentation (SWANS), developed and trained in a previous study. We used MR images of patients with clinical diagnoses of mild cognitive impairment (MCI) and AD, segmented by two different raters. By using transfer learning techniques, we developed four new models, using different training methods. Testing was performed using 26% of the original dataset, which was excluded from training as a hold‐out test set. In addition, 10% of the overall training dataset was used as a hold‐out validation set. Results showed that all the new models achieved better hippocampal segmentation quality than the baseline SWANS model (ps < .001), with high similarity to the manual segmentations (mean dice [best model] = 0.878 ± 0.003). The best model was chosen based on visual assessment and volume percentage error (VPE). The increased precision in estimating hippocampal volumes allows the detection of small hippocampal abnormalities already present in the MCI phase (SD = [3.9 ± 0.6]%), which may be crucial for early diagnosis. In this study, we used transfer learning technique for the segmentation of the hippocampus, considering three datasets of patients with a clinical diagnosis of mild cognitive impairment and Alzheimer disease, scanned with different protocols. We started from a previously developed deep learning algorithm, trained with a different dataset, and we quantified the benefits given by the transfer learning, both in a numerical and visual way, using manual segmentations from two raters as a gold standard.