Volumetric interpolation of tomographic sequences for accurate 3D reconstruction of anatomical parts

SSDL-an automated semi-supervised deep learning approach for patient-specific 3D reconstruction of proximal femur from QCT images

Deep Learning (DL) techniques have recently been used in medical image segmentation and the reconstruction of 3D anatomies of a human body. In this work, we propose a semi-supervised DL (SSDL) approach utilizing a CNN-based 3D U-Net model for femur segmentation from sparsely annotated quantitative computed tomography (QCT) slices. Specifically, QCT slices at the proximal end of the femur forming ball and socket joint with acetabulum were annotated for precise segmentation, where a segmenting binary mask was generated using a 3D U-Net model to segment the femur accurately. A total of 5474 QCT slices were considered for training among which 2316 slices were annotated. 3D femurs were further reconstructed from segmented slices employing polynomial spline interpolation. Both qualitative and quantitative performance of segmentation and 3D reconstruction were satisfactory with more than 90% accuracy achieved for all of the standard performance metrics considered. The spatial overlap index and reproducibility validation metric for segmentation-Dice Similarity Coefficient was 91.8% for unseen patients and 99.2% for validated patients. An average relative error of 12.02% and 10.75% for volume and surface area, respectively, were computed for 3D reconstructed femurs. The proposed approach demonstrates its effectiveness in accurately segmenting and reconstructing 3D femur from QCT slices.

A novel ear elements segmentation algorithm on depth map images

The growing interest in the auricular anatomy is due to two different strands of research: 1) in the medical field it is associated with autologous ear reconstruction, a surgery adopted following trauma or congenital malformations; 2) in surveillance and law enforcement the ear is used for human detection and recognition. Alternative systems of ear analysis can be differentiated for the type of input data (two-dimensional, three-dimensional or both), for the type of acquisition tools (3D scanner, photographs, video surveillance, etc.) and finally for the adopted algorithms. Although the segmentation and recognition of the ear from the face is a widely discussed topic in literature, the detection and recognition of individual anatomical elements has not yet been studied in depth. To this end, this work lays the foundation for the identification of the auricular elements through image processing algorithms. The proposed algorithm automatically identifies the contours of the main anatomical elements by processing depth map images. The algorithm was tested qualitatively and quantitatively on a dataset composed of 150 ears. The qualitative evaluation was performed with the collaboration of medical staff and the quantitative tests were performed using manually annotated ground truth data.