Improving transparency and representational generalizability through parallel continual learning.

This paper takes a parallel learning approach in continual learning scenarios. We define parallel continual learning as learning a sequence of tasks where the data for the previous tasks, whose distribution may have shifted over time, are also available while learning new tasks. We propose a parallel continual learning method by assigning subnetworks to each task, and simultaneously training only the assigned subnetworks on their corresponding tasks. In doing so, some parts of the network will be shared across multiple tasks. This is unlike the existing literature in continual learning which aims at learning incoming tasks sequentially, with the assumption that the data for the previous tasks have a fixed distribution. Our proposed method offers promises in: (1) Transparency in the network and in the relationship across tasks by enabling examination of the learned representations by independent and shared subnetworks, (2) Representation generalizability through sharing and training subnetworks on multiple tasks simultaneously. Our analysis shows that compared to many competing approaches such as continual learning, neural architecture search, and multi-task learning, parallel continual learning is capable of learning more generalizable representations. Also, (3)Parallel continual learning overcomes the common issue of catastrophic forgetting in continual learning algorithms. This is the first effort to train a neural network on multiple tasks and input domains simultaneously in a continual learning scenario. Our code is available at https://github.com/yours-anonym/PaRT.

Model learning analysis of 3D optoacoustic mesoscopy images for the classification of atopic dermatitis.

Atopic dermatitis (AD) is a skin inflammatory disease affecting 10% of the population worldwide. Raster-scanning optoacoustic mesoscopy (RSOM) has recently shown promise in dermatological imaging. We conducted a comprehensive analysis using three machine-learning models, random forest (RF), support vector machine (SVM), and convolutional neural network (CNN) for classifying healthy versus AD conditions, and sub-classifying different AD severities using RSOM images and clinical information. CNN model successfully differentiates healthy from AD patients with 97% accuracy. With limited data, RF achieved 65% accuracy in sub-classifying AD patients into mild versus moderate-severe cases. Identification of disease severities is vital in managing AD treatment.

Regional registration of whole slide image stacks containing major histological artifacts.

BACKGROUND: High resolution 2D whole slide imaging provides rich information about the tissue structure. This information can be a lot richer if these 2D images can be stacked into a 3D tissue volume. A 3D analysis, however, requires accurate reconstruction of the tissue volume from the 2D image stack. This task is not trivial due to the distortions such as tissue tearing, folding and missing at each slide. Performing registration for the whole tissue slices may be adversely affected by distorted tissue regions. Consequently, regional registration is found to be more effective. In this paper, we propose a new approach to an accurate and robust registration of regions of interest for whole slide images. We introduce the idea of multi-scale attention for registration. RESULTS: Using mean similarity index as the metric, the proposed algorithm (mean ± SD [Formula: see text]) followed by a fine registration algorithm ([Formula: see text]) outperformed the state-of-the-art linear whole tissue registration algorithm ([Formula: see text]) and the regional version of this algorithm ([Formula: see text]). The proposed algorithm also outperforms the state-of-the-art nonlinear registration algorithm (original: [Formula: see text], regional: [Formula: see text]) for whole slide images and a recently proposed patch-based registration algorithm (patch size 256: [Formula: see text] , patch size 512: [Formula: see text]) for medical images. CONCLUSION: Using multi-scale attention mechanism leads to a more robust and accurate solution to the problem of regional registration of whole slide images corrupted in some parts by major histological artifacts in the imaged tissue.

Exploring MRI based radiomics analysis of intratumoral spatial heterogeneity in locally advanced nasopharyngeal carcinoma treated with intensity modulated radiotherapy.

BACKGROUND: We hypothesized that spatial heterogeneity exists between recurrent and non-recurrent regions within a tumor. The aim of this study was to determine if there is a difference between radiomics features derived from recurrent versus non recurrent regions within the tumor based on pre-treatment MRI. METHODS: A total of 14 T4NxM0 NPC patients with histologically proven "in field" recurrence in the post nasal space following curative intent IMRT were included in this study. Pretreatment MRI were co-registered with MRI at the time of recurrence for the delineation of gross tumor volume at diagnosis(GTV) and at recurrence(GTVr). A total of 7 histogram features and 40 texture features were computed from the recurrent(GTVr) and non-recurrent region(GTV-GTVr). Paired t-tests and Wilcoxon signed-rank tests were carried out on the 47 quantified radiomics features. RESULTS: A total of 7 features were significantly different between recurrent and non-recurrent regions. Other than the variance from intensity-based histogram, the remaining six significant features were either from the gray-level size zone matrix (GLSZM) or the neighbourhood gray-tone difference matrix (NGTDM). CONCLUSIONS: The radiomic features extracted from pre-treatment MRI can potentially reflect the difference between recurrent and non-recurrent regions within a tumor and has a potential role in pre-treatment identification of intra-tumoral radio-resistance for selective dose escalation.

Improved tagged cardiac MRI myocardium strain analysis by leveraging cine segmentation.

BACKGROUND AND OBJECTIVES: Tagged MR images provide an effective way for regional analysis of the myocardium strain. A reliable myocardium strain analysis requires both correct segmentation and accurate motion tracking of the myocardium during the cardiac cycle. While many algorithms have been proposed for accurate tracking of the myocardium in tagged MR images, little focus has been placed on ensuring correct segmentation of the tagged myocardium during the cardiac cycle. Myocardial strain analysis is usually done by segmenting the myocardium in end-diastole, generating a mesh from the segmentation, propagating the mesh through the cardiac cycle using the output deformation field from motion tracking, and measuring strain on the deforming mesh. Due to the imposed tag strips on the anatomy, identification of the myocardium boundaries is challenging in tagged MR images. As a result, there is no guarantee that the propagated mesh is annotating the myocardium accurately through the cardiac cycle. Moreover, clinical studies indicate that incorrect myocardium annotation can result in overestimation of myocardial strains. METHODS: We introduce a method to improve reliability of strain analysis by proposing a mesh which correctly segments the myocardium in tagged MRI by leveraging the available cine MRI segmentation. In particular, we generate a series of mesh proposals using the cine MRI segmentation and find the propagated mesh proposal which gives the most accurate full-cycle myocardium segmentation. RESULTS: The mesh selection algorithm was tested on 22 2D MRI scans of diseased and healthy hearts. The proposed algorithm provided more accurate whole-cycle myocardium segmentation compared to the propagated end-diastolic mesh. Regional myocardium strain was measured for 10 3D MRI scans of healthy volunteers using the proposed mesh and the end-diastolic mesh. The measured strain using the proposed mesh was more similar to the expected myocardium strain for a healthy heart than the measured strain using the end-diastolic mesh. CONCLUSION: The proposed approach provides accurate whole-cycle tagged myocardium segmentation and more reliable myocardium strain analysis.

Comparison of wrist-worn Fitbit Flex and waist-worn ActiGraph for measuring steps in free-living adults.

INTRODUCTION: Accelerometers are commonly used to assess physical activity. Consumer activity trackers have become increasingly popular today, such as the Fitbit. This study aimed to compare the average number of steps per day using the wrist-worn Fitbit Flex and waist-worn ActiGraph (wGT3X-BT) in free-living conditions. METHODS: 104 adult participants (n = 35 males; n = 69 females) were asked to wear a Fitbit Flex and an ActiGraph concurrently for 7 days. Daily step counts were used to classify inactive (<10,000 steps) and active (≥10,000 steps) days, which is one of the commonly used physical activity guidelines to maintain health. Proportion of agreement between physical activity categorizations from ActiGraph and Fitbit Flex was assessed. Statistical analyses included Spearman's rho, intraclass correlation (ICC), median absolute percentage error (MAPE), Kappa statistics, and Bland-Altman plots. Analyses were performed among all participants, by each step-defined daily physical activity category and gender. RESULTS: The median average steps/day recorded by Fitbit Flex and ActiGraph were 10193 and 8812, respectively. Strong positive correlations and agreement were found for all participants, both genders, as well as daily physical activity categories (Spearman's rho: 0.76-0.91; ICC: 0.73-0.87). The MAPE was: 15.5% (95% confidence interval [CI]: 5.8-28.1%) for overall steps, 16.9% (6.8-30.3%) vs. 15.1% (4.5-27.3%) in males and females, and 20.4% (8.7-35.9%) vs. 9.6% (1.0-18.4%) during inactive days and active days. Bland-Altman plot indicated a median overestimation of 1300 steps/day by the Fitbit Flex in all participants. Fitbit Flex and ActiGraph respectively classified 51.5% and 37.5% of the days as active (Kappa: 0.66). CONCLUSIONS: There were high correlations and agreement in steps between Fitbit Flex and ActiGraph. However, findings suggested discrepancies in steps between devices. This imposed a challenge that needs to be considered when using Fibit Flex in research and health promotion programs.

Automatic basal slice detection for cardiac analysis.

Identification of the basal slice in cardiac imaging is a key step to measuring the ejection fraction of the left ventricle. Despite all the effort placed on automatic cardiac segmentation, basal slice identification is routinely performed manually. Manual identification, however, suffers from high interobserver variability. As a result, an automatic algorithm for basal slice identification is required. Guidelines published in 2013 identify the basal slice based on the percentage of myocardium surrounding the blood cavity in the short-axis view. Existing methods, however, assume that the basal slice is the first short-axis view slice below the mitral valve and are consequently at times identifying the incorrect short-axis slice. Correct identification of the basal slice under the Society for Cardiovascular Magnetic Resonance guidelines is challenging due to the poor image quality and blood movement during image acquisition. This paper proposes an automatic tool that utilizes the two-chamber view to determine the basal slice while following the guidelines. To this end, an active shape model is trained to segment the two-chamber view and create temporal binary profiles from which the basal slice is identified. From the 51 tested cases, our method obtains 92% and 84% accurate basal slice detection for the end-systole and the end-diastole, respectively.