There has recently been great progress in automatic segmentation of medical images with deep learning algorithms. In most works observer variation is acknowledged to be a problem as it makes training data heterogeneous but so far no attempts have been made to explicitly capture this variation. Here, we propose an approach capable of mimicking different styles of segmentation, which potentially can improve quality and clinical acceptance of automatic segmentation methods. In this work, instead of training one neural network on all available data, we train several neural networks on subgroups of data belonging to different segmentation variations separately. Because a priori it may be unclear what styles of segmentation exist in the data and because different styles do not necessarily map one-on-one to different observers, the subgroups should be automatically determined. We achieve this by searching for the best data partition with a genetic algorithm. Therefore, each network can learn a specific style of segmentation from grouped training data. We provide proof of principle results for open-sourced prostate segmentation MRI data with simulated observer variations. Our approach provides an improvement of up to 23% (depending on simulated variations) in terms of Dice and surface Dice coefficients compared to one network trained on all data.

@en

We introduce a novel surrogate-assisted Genetic Algorithm (GA) for expensive optimization of problems with discrete categorical variables. Specifically, we leverage the strengths of the Gene-pool Optimal Mixing Evolutionary Algorithm (GOMEA), a state-of-the-art GA, and, for the first time, propose to use a convolutional neural network (CNN) as a surrogate model. We propose to train the model on pairwise fitness differences to decrease the number of evaluated solutions that is required to achieve adequate surrogate model training. In providing a proof of principle, we consider relatively standard CNNs, and demonstrate that their capacity is already sufficient to accurately learn fitness landscapes of various well-known benchmark functions. The proposed CS-GOMEA is compared with GOMEA and the widely-used Bayesian-optimization-based expensive optimization frameworks SMAC and Hyperopt, in terms of the number of evaluations that is required to achieve the optimum. In our experiments on binary problems with dimensionalities up to 400 variables, CS-GOMEA always found the optimum, whereas SMAC and Hyperopt failed for problem sizes over 16 variables. Moreover, the number of evaluated solutions required by CS-GOMEA to find the optimum was found to scale much better than GOMEA.

@en

Generalization of voxelwise classifiers is hampered by differences between MRI-scanners, e.g. different acquisition protocols and field strengths. To address this limitation, we propose a Siamese neural network (MRAI-net) that extracts acquisition-invariant feature vectors. These can consequently be used by task-specific methods, such as voxelwise classifiers for tissue segmentation. MRAI-net is tested on both simulated and real patient data. Experiments show that MRAI-net outperforms voxelwise classifiers trained on the source or target scanner data when a small number of labeled samples is available.@en

Background: The most tedious and time-consuming task in medical additive manufacturing (AM) is image segmentation. The aim of the present study was to develop and train a convolutional neural network (CNN) for bone segmentation in computed tomography (CT) scans.

Method: The CNN was trained with CT scans acquired using six different scanners. Standard tessellation language (STL) models of 20 patients who had previously undergone craniotomy and cranioplasty using additively manufactured skull implants served as “gold standard” models during CNN training. The CNN segmented all patient CT scans using a leave-2-out scheme. All segmented CT scans were converted into STL models and geometrically compared with the gold standard STL models.

Results: The CT scans segmented using the CNN demonstrated a large overlap with the gold standard segmentation and resulted in a mean Dice similarity coefficient of 0.92 ± 0.04. The CNN-based STL models demonstrated mean surface deviations ranging between −0.19 mm ± 0.86 mm and 1.22 mm ± 1.75 mm, when compared to the gold standard STL models. No major differences were observed between the mean deviations of the CNN-based STL models acquired using six different CT scanners. Conclusions: The fully-automated CNN was able to accurately segment the skull. CNNs thus offer the opportunity of removing the current prohibitive barriers of time and effort during CT image segmentation, making patient-specific AM constructs more accesible.

@en