skip to main content

This content will become publicly available on November 16, 2023

Title: Calibrated bagging deep learning for image semantic segmentation: A case study on COVID-19 chest X-ray image
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) causes coronavirus disease 2019 (COVID-19). Imaging tests such as chest X-ray (CXR) and computed tomography (CT) can provide useful information to clinical staff for facilitating a diagnosis of COVID-19 in a more efficient and comprehensive manner. As a breakthrough of artificial intelligence (AI), deep learning has been applied to perform COVID-19 infection region segmentation and disease classification by analyzing CXR and CT data. However, prediction uncertainty of deep learning models for these tasks, which is very important to safety-critical applications like medical image processing, has not been comprehensively investigated. In this work, we propose a novel ensemble deep learning model through integrating bagging deep learning and model calibration to not only enhance segmentation performance, but also reduce prediction uncertainty. The proposed method has been validated on a large dataset that is associated with CXR image segmentation. Experimental results demonstrate that the proposed method can improve the segmentation performance, as well as decrease prediction uncertainty.
; ; ; ;
Hemanth, Jude
Award ID(s):
Publication Date:
Journal Name:
Page Range or eLocation-ID:
Sponsoring Org:
National Science Foundation
More Like this
  1. The newly discovered Coronavirus Disease 2019 (COVID-19) has been globally spreading and causing hundreds of thousands of deaths around the world as of its first emergence in late 2019. The rapid outbreak of this disease has overwhelmed health care infrastructures and arises the need to allocate medical equipment and resources more efficiently. The early diagnosis of this disease will lead to the rapid separation of COVID-19 and non-COVID cases, which will be helpful for health care authorities to optimize resource allocation plans and early prevention of the disease. In this regard, a growing number of studies are investigating the capability of deep learning for early diagnosis of COVID-19. Computed tomography (CT) scans have shown distinctive features and higher sensitivity compared to other diagnostic tests, in particular the current gold standard, i.e., the Reverse Transcription Polymerase Chain Reaction (RT-PCR) test. Current deep learning-based algorithms are mainly developed based on Convolutional Neural Networks (CNNs) to identify COVID-19 pneumonia cases. CNNs, however, require extensive data augmentation and large datasets to identify detailed spatial relations between image instances. Furthermore, existing algorithms utilizing CT scans, either extend slice-level predictions to patient-level ones using a simple thresholding mechanism or rely on a sophisticated infection segmentation tomore »identify the disease. In this paper, we propose a two-stage fully automated CT-based framework for identification of COVID-19 positive cases referred to as the “COVID-FACT”. COVID-FACT utilizes Capsule Networks, as its main building blocks and is, therefore, capable of capturing spatial information. In particular, to make the proposed COVID-FACT independent from sophisticated segmentations of the area of infection, slices demonstrating infection are detected at the first stage and the second stage is responsible for classifying patients into COVID and non-COVID cases. COVID-FACT detects slices with infection, and identifies positive COVID-19 cases using an in-house CT scan dataset, containing COVID-19, community acquired pneumonia, and normal cases. Based on our experiments, COVID-FACT achieves an accuracy of 90.82 % , a sensitivity of 94.55 % , a specificity of 86.04 % , and an Area Under the Curve (AUC) of 0.98, while depending on far less supervision and annotation, in comparison to its counterparts.« less
  2. Computer vision techniques always had played a salient role in numerous medical fields, especially in image diagnosis. Amidst a global pandemic situation, one of the archetypal methods assisting healthcare professionals in diagnosing various types of lung cancers, heart diseases, and COVID-19 infection is the Computed Tomography (CT) medical imaging technique. Segmentation of Lung and Infection with high accuracy in COVID-19 CT scans can play a vital role in the prognosis and diagnosis of a mass population of infected patients. Most of the existing works are predominately based on large private data sets that are practically impossible to obtain during a pandemic situation. Moreover, it is difficult to compare the segmentation methods as the data set are obtained in various geographical areas and developed and implemented in different environments. To help the current global pandemic situation, we are proposing a highly data-efficient method that gets trained on 20 expert annotated COVID-19 cases. To increase the efficiency rate further, the proposed model has been implemented on NVIDIA - Jetson Nano (System-on-Chip) to completely exploit the GPU performance for a medical application machine learning module. To compare the results, we tested the performance with conventional U-Net architecture and calculated the performance metrics. Themore »proposed state-of-art method proves better than the conventional architecture delivering a Dice Similarity Coefficient of 99%.« less
  3. Abstract The global spread of COVID-19, the disease caused by the novel coronavirus SARS-CoV-2, has casted a significant threat to mankind. As the COVID-19 situation continues to evolve, predicting localized disease severity is crucial for advanced resource allocation. This paper proposes a method named COURAGE (COUnty aggRegation mixup AuGmEntation) to generate a short-term prediction of 2-week-ahead COVID-19 related deaths for each county in the United States, leveraging modern deep learning techniques. Specifically, our method adopts a self-attention model from Natural Language Processing, known as the transformer model, to capture both short-term and long-term dependencies within the time series while enjoying computational efficiency. Our model solely utilizes publicly available information for COVID-19 related confirmed cases, deaths, community mobility trends and demographic information, and can produce state-level predictions as an aggregation of the corresponding county-level predictions. Our numerical experiments demonstrate that our model achieves the state-of-the-art performance among the publicly available benchmark models.
  4. Abstract We consider semantic image segmentation. Our method is inspired by Bayesian deep learning which improves image segmentation accuracy by modeling the uncertainty of the network output. In contrast to uncertainty, our method directly learns to predict the erroneous pixels of a segmentation network, which is modeled as a binary classification problem. It can speed up training comparing to the Monte Carlo integration often used in Bayesian deep learning. It also allows us to train a branch to correct the labels of erroneous pixels. Our method consists of three stages: (i) predict pixel-wise error probability of the initial result, (ii) redetermine new labels for pixels with high error probability, and (iii) fuse the initial result and the redetermined result with respect to the error probability. We formulate the error-pixel prediction problem as a classification task and employ an error-prediction branch in the network to predict pixel-wise error probabilities. We also introduce a detail branch to focus the training process on the erroneous pixels. We have experimentally validated our method on the Cityscapes and ADE20K datasets. Our model can be easily added to various advanced segmentation networks to improve their performance. Taking DeepLabv3+ as an example, our network can achieve 82.88%more »of mIoU on Cityscapes testing dataset and 45.73% on ADE20K validation dataset, improving corresponding DeepLabv3+ results by 0.74% and 0.13% respectively.« less
  5. Objective and Impact Statement . We propose an automated method of predicting Normal Pressure Hydrocephalus (NPH) from CT scans. A deep convolutional network segments regions of interest from the scans. These regions are then combined with MRI information to predict NPH. To our knowledge, this is the first method which automatically predicts NPH from CT scans and incorporates diffusion tractography information for prediction. Introduction . Due to their low cost and high versatility, CT scans are often used in NPH diagnosis. No well-defined and effective protocol currently exists for analysis of CT scans for NPH. Evans’ index, an approximation of the ventricle to brain volume using one 2D image slice, has been proposed but is not robust. The proposed approach is an effective way to quantify regions of interest and offers a computational method for predicting NPH. Methods . We propose a novel method to predict NPH by combining regions of interest segmented from CT scans with connectome data to compute features which capture the impact of enlarged ventricles by excluding fiber tracts passing through these regions. The segmentation and network features are used to train a model for NPH prediction. Results . Our method outperforms the current state-of-the-art bymore »9 precision points and 29 recall points. Our segmentation model outperforms the current state-of-the-art in segmenting the ventricle, gray-white matter, and subarachnoid space in CT scans. Conclusion . Our experimental results demonstrate that fast and accurate volumetric segmentation of CT brain scans can help improve the NPH diagnosis process, and network properties can increase NPH prediction accuracy.« less