Resting-state functional magnetic resonance imaging (rs-fMRI) has significantly advanced the diagnosis of brain diseases. However, existing methods are generally limited to small, disease-specific datasets with less convincing outcomes or lack the interpretability needed to identify reliable disease-associated biomarkers. In this paper, we introduce a novel generative inference model that integrates a Variational Autoencoder (VAE) with Non-negative Matrix Factorization (NMF). Our model comprises three key components: an encoder for learning spatiotemporal dynamic feature embeddings within fMRI data, a decoder to reconstruct the input data from the encoded latent space, and a classifier to distinguish between neurological disorders and normal controls. The three components are simultaneously optimized to perform inference by estimating the posterior distribution of the latent variables from the input fMRI, yielding predictive and interpretable biomarkers for the diagnosis of neurological disorders. We extensively evaluated our method for identifying Autism Spectrum Disorder (ASD) and Alzheimer’s Disease (AD) using two public datasets, ABIDE and ADNI. Experimental results show that our method achieves state-of-the-art performance across various metrics.

Existing methods for diagnosing middle ear diseases using temporal bone computed tomography (CT) imaging primarily focus on adult datasets and require labor-intensive manual input from radiologists to label and select regions of interest (ROIs). These methods rely on prior knowledge and introduce inter- and intra-observer variability. Additionally, the selected ROIs typically consist of a few 2D slices, underutilizing the 3D capabilities of CT imaging. Moreover, these methods are not reproducible on pediatric datasets, where rapidly developing heads exhibit greater morphological variability. To address these challenges, we developed a fully automated framework for precise diagnosis of pediatric chronic suppurative otitis media (CSOM) and cholesteatoma (MEC) using temporal bone CT imaging. Our method automatically detects the most informative 3D ROIs by calculating entropy changes in the 3D anatomical structures of the middle ear, eliminating the need for annotations or prior knowledge. We also introduce a multi-scale classification network that incorporates a global-local feature interaction strategy and uses a wide and deep multi-layer transformer for feature extraction, effectively learning feature dependencies. Experimental results on a dataset of CT images from 603 pediatric patients show that our method achieves a classification accuracy of 93.17% and an AUC-ROC of 96.84%, outperforming state-of-the-art methods. This innovative approach reduces radiologists’ workload, aids automated surgical navigation, and has the potential to transform diagnostic workflows in otolaryngology.

Conventional one-shot segmentation methods typically employ either registration methods for label propagation from a reference atlas or utilize synthetically labeled data to augment the training of segmentation networks. However, these approaches often fail to accurately capture anatomical structural information from real images, resulting in suboptimal segmentation performance and limited generalizability across different medical imaging modalities. In this paper, we propose a robust one-shot brain tissue segmentation framework, which requires only a single labeled image and a few unlabeled ones. Our approach features a novel synthesis module based on patch-wise contrastive learning for generating realistic, well-labeled training samples, followed by a dual-head Variational AutoEncoder (VAE) module for the joint learning of reconstruction and segmentation. During the final inference, the well-trained VAE is capable of precisely segmenting new, unseen images. Our method demonstrates versatility across various imaging modalities. Evaluations on two public MRI datasets and one in-house CT dataset reveal that our method achieves superior performance and enhanced generalization capabilities compared to existing state-of-the-art methods.

Early detection of facial dysmorphology - variations of the normal facial geometry - is essential for the timely detection of genetic conditions, which has a significant impact in the reduction of the mortality and morbidity associated with them. A model encoding the normal variability in the healthy population can serve as a reference to quantify the often subtle facial abnormalities that are present in young patients with such conditions. In this paper, we present the first facial model constructed exclusively from newborn data, the Baby Face Model (BabyFM). Our model is built from 3D scans with an innovative pipeline based on least squared conformal maps (LSCM). LSCM are piece-wise linear mappings that project the training faces to a common 2D space minimising the conformal distortion. This process allows improving the correspondences between 3D faces, which is particularly important for the identification of subtle dysmorphology. We evaluate the ability of our BabyFM to recover the babys facial morphology from a set of 2D images by comparing it to state-of-the-art facial models. We also compare it to models built following an analogous pipeline to the one proposed in this paper but using nonrigid iterative closest point (NICP) to establish dense correspondences between the training faces. The results show that our model reconstructs the facial morphology of babies with significantly smaller errors than the state-of-the-art models (p = 10-4) and the “NICP models” (p <; 0.01).

We present a novel approach for improving the shape statistics of medical image objects by generating correspondence of skeletal points. Each object's interior is modeled by an s-rep, i.e., by a sampled, folded, two-sided skeletal sheet with spoke vectors proceeding from the skeletal sheet to the boundary. The skeleton is divided into three parts: the up side, the down side, and the fold curve. The spokes on each part are treated separately and, using spoke interpolation, are shifted along that skeleton in each training sample so as to tighten the probability distribution on those spokes' geometric properties while sampling the object interior regularly. As with the surface/boundary-based correspondence method of Cates et al., entropy is used to measure both the probability distribution tightness and the sampling regularity, here of the spokes' geometric properties. Evaluation on synthetic and real world lateral ventricle and hippocampus data sets demonstrate improvement in the performance of statistics using the resulting probability distributions. This improvement is greater than that achieved by an entropy-based correspondence method on the boundary points.

We present a scheme that propagates a reference skeletal model (s-rep) into a particular case of an object, thereby propagating the initial shape-related layout of the skeleton-to-boundary vectors, called spokes. The scheme represents the surfaces of the template as well as the target objects by spherical harmonics and computes a warp between these via a thin plate spline. To form the propagated s-rep, it applies the warp to the spokes of the template s-rep and then statistically refines. This automatic approach promises to make s-rep fitting robust for complicated objects, which allows s-rep based statistics to be available to all. The improvement in fitting and statistics is significant compared with the previous methods and in statistics compared with a state-of-the-art boundary based method.