Category: IPMI 2021 poster

Authors: Xin Ma (The University of Texas at Arlington)*; Guorong Wu (University of North Carolina); Seong Jae Hwang (University of Pittsburgh); Won Hwa Kim (POSTECH)

Poster

Abstract: Tremendous recent literature show that associations between different brain regions, i.e., brain connectivity, provide early symptoms of neurological disorders. Despite significant efforts made for graph neural network (GNN) techniques, their focus on graph nodes makes the state-of-the-art GNN methods not suitable for classifying brain connectivity as graphs where the objective is to characterize disease-relevant network dysfunction patterns on graph links. To address this issue, we propose Multi-resolution Edge Network (MENET) to detect disease-specific connectomic benchmarks with high discrimination power across diagnostic categories. The core of MENET is a novel graph edge-wise transform that we propose, which allows us to capture multi-resolution “connectomic” features. Using a rich set of the connectomic features, we devise a graph learning framework to jointly select discriminative edges and assign diagnostic labels for graphs. Experiments on two real datasets show that MENET accurately predicts diagnostic labels and identify brain connectivities highly associated with neurological disorders such as Alzheimer’s Disease and Attention-Deficit/Hyperactivity Disorder.

Authors: Naresh Nandakumar (JHU)*; Komal Manzoor (JHU); Shruti Agarwal (JHU); Sachin Gujar (Johns Hopkins University); Jay Pillai (Johns Hopkins RAIL); Haris Sair (Johns Hopkins RAIL); Archana Venkataraman (Johns Hopkins University)

Poster

Abstract: We present a deep neural network architecture that combines multi-scale spatial attention with temporal attention to simultaneously localize the language and motor areas of the eloquent cortex from dynamic functional connectivity data. Our multi-scale spatial attention operates on graph-based features extracted from the connectivity matrices, thus honing in on the inter-regional interactions that collectively define the eloquent cortex. At the same time, our temporal attention model selects the intervals during which these interactions are most pronounced.
The final stage of our model employs multi-task learning to differentiate between the eloquent subsystems. Our training strategy enables us to handle missing eloquent class labels by freezing the weights in those branches while updating the rest of the network weights. We evaluate our method on resting-state fMRI data from one synthetic dataset and one in-house brain tumor dataset while using task fMRI activations as ground-truth labels for the eloquent cortex. Our model achieves higher localization accuracies than conventional deep learning approaches. It also produces interpretable spatial and temporal attention features which can provide further insights for presurgical planning. Thus, our model shows translational promise for improving the safety of brain tumor resections.

Authors: Peter A Wijeratne (University College London)*; Daniel Alexander (University College London)

Poster

Abstract: Progressive diseases worsen over time and can be characterised by sequences of events that correspond to changes in observable features of disease progression. Here we connect ideas from two formerly separate methodologies — event-based and hidden Markov modelling — to derive a new generative model of disease progression: the Temporal Event-Based Model (TEBM). TEBM can uniquely infer the most likely group-level sequence and timing of events (natural history) from mixed data types. Moreover, it can infer and predict individual-level trajectories (prognosis) even when data are missing, giving it high clinical utility. Here we derive TEBM and provide an inference scheme based on the expectation maximisation algorithm. We use imaging, clinical and biofluid data from the Alzheimer’s Disease Neuroimaging Initiative to demonstrate the validity and utility of our model. First, we train TEBM to uncover a new sequence and timing of events in Alzheimer’s disease, which are inferred to occur over a period of ~17.6 years. Next, we demonstrate the utility of TEBM in predicting clinical progression, and that TEBM provides improved utility over a comparative disease progression model. Finally, we demonstrate that TEBM maintains predictive accuracy with up to 50% missing data. These results support the clinical validity of TEBM and its broader utility in real-world medical applications.

Authors: Matthias Perkonigg (Medical University of Vienna)*; Johannes Hofmanninger (Medical University of Vienna); Georg Langs (Medical University of Vienna)

Poster

Abstract:

Imaging in clinical routine is subject to changing scanner protocols, hardware, or policies in a typically heterogeneous set of acquisition hardware. Accuracy and reliability of deep learning models suffer from those changes as data and targets become inconsistent with their initial static training set. Continual learning can adapt to a continuous data stream of a changing imaging environment.
Here, we propose a method for continual active learning on a data stream of medical images. It recognizes shifts or additions of new imaging sources – domains -, adapts training accordingly, and selects optimal examples for labelling. Model training has to cope with a limited labelling budget, resembling typical real world scenarios. We demonstrate our method on T1-weighted magnetic resonance images from three different scanners with the task of brain age estimation. Results demonstrate that the proposed method outperforms naive active learning while requiring less manual labelling.

Authors: Xianfeng GU (Stony Brook University)*; Yalin Wang (Arizona State University); Dongsheng An (Stony Brook University); Min Zhang (Brigham and Women’s Hospital, Harvard Medical School); Na Lei (Dalian University of Technology); Tong Zhao (inria); Jianfeng Wu (Arizona State University); Xiaoyin Xu (Brigham and Women’s Hospital, Harvard Medical School)

Poster

Abstract: Biomarkers play an important role in preclinical/early detection and intervention in Alzheimer’s disease (AD). However, obtaining effective biomarkers for AD is still a big challenge. In this work, we propose to use the worst transportation cost as the biomarker for the A beta brain surfaces.
The worst transportation (WT) aims to find the least economical way to transport one measure to the other, which contrasts to the optimal transportation (OT) that find the most economical way between measures. In contrast, maps find the The transportation cost of an OT map is the Wasserstein distance between the measures, which has been broadly applied for shape analysis.
To compute the WT cost, we generalize the Brenier theorem for the OT map to the WT map, and show that the WT map is the gradient of a concave function satisfying the Monge-Ampere equation. We also develop an algorithm to compute the WT map based on computational geometry. Finally, we successfully use the WT cost to study the group difference between the A beta positive AD patients and A beta negative cognitively unimpaired subjects.

Authors: Mobarakol Islam (Imperial College London)*; Ben Glocker (Imperial College London)

Poster

Abstract: The task of image segmentation is inherently noisy due to ambiguities regarding the exact location of boundaries between anatomical structures. We argue that this information can be extracted from the expert annotations at no extra cost, and when integrated into state-of-the-art neural networks, it can lead to improved calibration between soft probabilistic predictions and the underlying uncertainty. We built upon label smoothing (LS) where a network is trained on `blurred’ versions of the ground truth labels which has been shown to be effective for calibrating output predictions. However, LS is not taking the local structure into account and results in overly smoothed predictions with low confidence even for non-ambiguous regions. Here, we propose Spatially Varying Label Smoothing (SVLS), a soft labeling technique that captures the structural uncertainty in semantic segmentation. SVLS also naturally lends itself to incorporate inter-rater uncertainty when multiple labelmaps are available. The proposed approach is extensively validated on four clinical segmentation tasks with different imaging modalities, number of classes and single and multi-rater expert annotations. The results demonstrate that SVLS, despite its simplicity, obtains superior boundary prediction with improved uncertainty and model calibration.

Authors: Xingyu Li (University of Alberta)*

Poster

Abstract: Multiple stains are usually used to highlight biological substances in biomedical image analysis. To decompose multiple stains for co-localization quantification, blind source separation is usually performed. Prior model-based stain separation methods usually rely on stains’ spatial distributions over an image and may fail to solve the co-localization problem. With the advantage of machine learning, deep generative models are used for this purpose. Since prior knowledge of imaging models is ignored in purely data-driven solutions, these methods may be sub-optimal. In this study, a novel learning-based blind source separation framework is proposed, where the physical model of biomedical imaging is incorporated to regularize the learning process. The introduced model-relevant adversarial loss couples all generators in the framework and limits the capacities of the generative models. Further more, a training algorithm is innovated for the proposed framework to avoid inter-generator confusion during learning. This paper particularly takes fluorescence unmixing in fluorescence microscopy images as an application example of the proposed framework. Qualitative and quantitative experimentation on a public fluorescence microscopy image set demonstrates the superiority of the proposed method over both prior model-based approaches and learning-based methods.

Authors: Yuexiang Li (Youtu Lab, Tencent)*; Jiawei Chen (Tencent); Kai Ma (Tencent); Yefeng Zheng (Tencent)

Poster

Abstract: Cervical cancer causes the fourth most cancer-related deaths of women worldwide. One of the most commonly-used clinical tools for the diagnosis of cervical intraepithelial neoplasia (CIN) and cervical cancer is colposcopy examination. However, due to the challenging imaging conditions such as light reflection on the cervix surface, the clinical accuracy of colposcopy examination is relatively low. In this paper, we propose a computer-aided diagnosis (CAD) system to accurately segment the lesion areas (i.e., CIN and cancer) from colposcopic images, which can not only assist colposcopists for clinical decision, but also provide the guideline for the location of biopsy sites. In clinical practice, colposcopists often need to zoom in the potential lesion area for clearer observation. The colposcopic images with multi-scale views result in a difficulty for current straight-forward deep learning networks to process. To address the problem, we propose a novel attention mechanism, namely feature library, which treats the whole backbone network as a pool of features and extract the useful features on different scales from the pool to recalibrate the most informative representation. Furthermore, to well-train and evaluate our deep learning network, we collect a large-scale colposcopic image dataset for CervIcal lesioN sEgMentAtion (CINEMA), consisting of 34,337 images from 9,652 patients. The lesion areas in the colposcopic images are manually annotated by experienced colposcopists. Extensive experiments are conducted on the CINEMA dataset, which demonstrate the effectiveness of our feature library dealing with cervical lesions of varying sizes.

Authors: Jiazhou Chen (South China University of Technology)*; Defu Yang (University of North Carolina at Chapel Hill); Hongmin Cai (South China University of Technology); Martin Styner (University of North Carolina); Guorong Wu (UNC-CH)

Poster

Abstract: A plethora of neuroscience studies show that neurodegenerative diseases such as Alzheimer’s disease (AD) manifest network dysfunction much earlier before the onset of clinical symptoms, where neuropathological burdens often propagate across brain networks in a prion-like manner. In this context, characterizing the in-vivo spreading pathway of neuropathological events provides a new window to understand the pathophysiological mechanism of AD progression. However, little attention has been paid to the intrinsic geometry associated with the spreading pathway of neuropathological events, which indeed requires a network-specific algebra to quantify the associated propagation profile across individuals. To address this challenge, we propose a novel manifold harmonic approach to construct a set of region-adaptive harmonic wavelets, which allow us to capture diverse local network geometry in the aging population using the “swiss army knife” of the network topology. The learned common harmonic wavelets constitute the most representative spreading pathways that can be used to characterize the local propagation patterns of neuropathology on an individual basis. We have evaluated the power of our novel harmonic wavelets in identifying AD-related spreading pathways that lead to cognitive decline. Compared with other popular empirical biomarkers, the harmonic-based propagation patterns show much high sensitivity and specificity in stratifying cognitive normal (CN), early-stage mild cognitive impairment (MCI), and late-stage MCI, which indicates the great potential of being a putative biomarker in predicting the risk of developing AD at the pre-clinical stage.

Authors: Hong-Yu Zhou (The University of Hong Kong)*; Hualuo Liu (Jilin University); Shilei Cao (Tencent); DONG WEI (Tencent Jarvis Lab); Chixiang Lu (Huazhong University of Science and Technology); Yizhou Yu (The University of Hong Kong); Kai Ma (Tencent); Yefeng Zheng (Tencent)

Poster

Abstract: Learning by imitation is one of the most significant abilities of human beings and plays a vital role in human’s computational neural system. In medical image analysis, given several exemplars (anchors), experienced radiologist has the ability to delineate unfamiliar organs by imitating the reasoning process learned from existing types of organs. Inspired by this observation, we propose \emph{OrganNet} which learns a generalized organ concept from a set of annotated organ classes and then transfer this concept to unseen classes. In this paper, we show that such process can be integrated into the one-shot segmentation task which is a very challenging but meaningful topic. We propose pyramid reasoning modules (PRMs) to model the anatomical correlation between anchor and target volumes. In practice, the proposed module first computes a correlation matrix between target and anchor computerized tomography (CT) volumes. Then, this matrix is used to transform the feature representations of both anchor volume and its segmentation mask. Finally, OrganNet learns to fuse the representations from various inputs and predicts segmentation results for target volume. Extensive experiments show that OrganNet can effectively resist the wide variations in organ morphology and produce state-of-the-art results in one-shot segmentation task. Moreover, even when compared with fully-supervised segmentation models, OrganNet is still able to produce satisfying segmentation results.