Assessment of U-Net in the segmentation of short tracts: Transferring to clinical MRI routine.

Accurately studying structural connectivity requires precise tract segmentation strategies. The U-Net network has been widely recognized for its exceptional capacity in image segmentation tasks and provides remarkable results in large tract segmentation when high-quality diffusion-weighted imaging (DWI) data are used. However, short tracts, which are associated with various neurological diseases, pose specific challenges, particularly when high-quality DWI data acquisition within clinical settings is concerned. Here, we aimed to evaluate the U-Net network ability to segment short tracts by using DWI data acquired in different experimental conditions. To this end, we conducted three types of training experiments involving 350 healthy subjects and 11 white matter tracts, including the anterior, posterior, and hippocampal commissure, fornix, and uncinated fasciculus. In the first experiment, the model was exclusively trained with high-quality data of the Human Connectome Project (HCP) dataset. The second experiment focused on images of healthy subjects acquired from a local hospital dataset, representing a typical clinical routine acquisition. In the third experiment, a hybrid training approach was employed, combining data of the HCP and local hospital datasets. Then, the best model was also tested in unseen DWIs of 10 epilepsy patients of the local hospital and 10 healthy subjects acquired on a scanner from another company. The outcomes of the third experiment demonstrated a notable enhancement in performance when contrasted with the preceding trials. Specifically, the short tracts within the local hospital dataset achieved Dice scores ranging between 0.60 and 0.65. Similar intervals were obtained with HCP data in the first experiment, and a substantial improvement compared to the scores between 0.37 and 0.50 obtained with the local hospital dataset at the same experiment. This improvement persisted when the method was applied to diverse scenarios, including different scanner acquisitions and epilepsy patients. These results indicate that combining datasets from different sources, coupled with resolution standardization strengthens the neural network ability to generalize predictions across a spectrum of datasets. Nevertheless, short tract segmentation performance is intricately linked to the training composition, to validation, and to testing data. Moreover, curved tracts have intricate structural nature, which adds complexities to their segmenting. Although the network training approach tested herein has provided promising results, caution must be taken when extrapolating its application to datasets acquired under distinct experimental conditions, even in the case of higher-quality data or analysis of long or short tracts.

Deep learning for automatic mandible segmentation on dental panoramic x-ray images.

Many studies in the last decades have correlated mandible bone structure with systemic diseases like osteoporosis. Mandible segmentation, as well as segmentation of other oral structures, is an essential step in studies that correlate oral structures' conditions with systemic diseases in general. However, manual mandible segmentation is a time-consuming and training-required task that suffers from inter and intra-user variability. Further, the dental panoramic x-ray image (PAN), the most used image in oral studies, contains overlapping of many structures and lacks contrast on structures' interface. Those facts make both manual and automatic mandible segmentation a challenge. In the present study, we propose a precise and robust set of deep learning-based algorithms for automatic mandible segmentation (AMS) on PAN images. Two datasets were considered. An in-house image dataset with 393 image/segmentation pairs was prepared using image data of 321 image patient data and the corresponding manual segmentation performed by an experienced specialist. Additionally, a publicly available third-party image dataset (TPD) composed of 116 image/segmentation pairs was used to train the models. Four deep learning models were trained using U-Net and HRNet architectures with and without data augmentation. An additional morphological refinement routine was proposed to enhance the models' prediction. An ensemble model was proposed combining the four best-trained segmentation models. The ensemble model with morphological refinement achieved the highest scores on the test set (98.27%, 97.60%, 97.18%, ACC, DICE, and IoU respectively), with the other models scoring above 95% in all performance metrics on the test set. The present study achieved the highest ranked performance considering all the previously published results on AMS for PAN images. Additionally, those are the most robust results achieved since it was performed over an image set with considerable gender representativeness, a wide age range, a large variety of oral conditions, and images from different imaging scans.