-
Objective dyspnea evaluation on COVID-19 patients learning from exertion-induced dyspnea scores
Authors:
Zijing Zhang,
Jianlin Zhou,
Thomas B. Conroy,
Samuel Chung,
Justin Choi,
Patrick Chau,
Daniel B. Green,
Ana C. Krieger,
Edwin C. Kan
Abstract:
Objective: Dyspnea is one of the most common symptoms for many pulmonary diseases including COVID-19. Clinical assessment of dyspnea is mainly performed by subjective self-report, which has limited accuracy and is challenging for continuous monitoring. The objective of this research study is to determine if dyspnea progression in COVID patients can be assessed using a non-invasive wearable sensor…
▽ More
Objective: Dyspnea is one of the most common symptoms for many pulmonary diseases including COVID-19. Clinical assessment of dyspnea is mainly performed by subjective self-report, which has limited accuracy and is challenging for continuous monitoring. The objective of this research study is to determine if dyspnea progression in COVID patients can be assessed using a non-invasive wearable sensor and if the findings are comparable to a learning model of physiologically induced dyspnea on healthy subjects. Methods: Non-invasive wearable respiratory sensors were employed to retrieve continuous respiratory characteristics with user comfort and convenience. Overnight (~16h) respiratory waveforms were collected on 12 COVID-19 patients, and a benchmark on 13 healthy subjects with exertion-induced dyspnea were also performed for blind comparison. The learning model was built from the respiratory features with self report on 32 healthy subjects under exertion and airway blockage. Results: High similarity between dyspnea on COVID patients and physiologically induced dyspnea on healthy subjects was established. COVID patients have consistently high objective dyspnea scores in comparison with normal breathing of healthy subjects. We also exhibited continuous dyspnea scoring capability for 12-16 hours on patients. Conclusion: This paper validates the viability to use our objective dyspnea scoring for clinical dyspnea assessment on COVID patients. Significance: The proposed system can help the identification of dyspneic exacerbation in conditions such as COVID, leading to early intervention and possibly improving their outcome. This approach can be potentially applied to other pulmonary disorders such as asthma, emphysema, and pneumonia.
△ Less
Submitted 7 November, 2022;
originally announced November 2022.
-
Hop-Spanners for Geometric Intersection Graphs
Authors:
Jonathan B. Conroy,
Csaba D. Tóth
Abstract:
A $t$-spanner of a graph $G=(V,E)$ is a subgraph $H=(V,E')$ that contains a $uv$-path of length at most $t$ for every $uv\in E$. It is known that every $n$-vertex graph admits a $(2k-1)$-spanner with $O(n^{1+1/k})$ edges for $k\geq 1$. This bound is the best possible for $1\leq k\leq 9$ and is conjectured to be optimal due to Erdős' girth conjecture.
We study $t$-spanners for $t\in \{2,3\}$ for…
▽ More
A $t$-spanner of a graph $G=(V,E)$ is a subgraph $H=(V,E')$ that contains a $uv$-path of length at most $t$ for every $uv\in E$. It is known that every $n$-vertex graph admits a $(2k-1)$-spanner with $O(n^{1+1/k})$ edges for $k\geq 1$. This bound is the best possible for $1\leq k\leq 9$ and is conjectured to be optimal due to Erdős' girth conjecture.
We study $t$-spanners for $t\in \{2,3\}$ for geometric intersection graphs in the plane. These spanners are also known as \emph{$t$-hop spanners} to emphasize the use of graph-theoretic distances (as opposed to Euclidean distances between the geometric objects or their centers). We obtain the following results: (1) Every $n$-vertex unit disk graph (UDG) admits a 2-hop spanner with $O(n)$ edges; improving upon the previous bound of $O(n\log n)$. (2) The intersection graph of $n$ axis-aligned fat rectangles admits a 2-hop spanner with $O(n\log n)$ edges, and this bound is tight up to a factor of $\log \log n$. (3) The intersection graph of $n$ fat convex bodies in the plane admits a 3-hop spanner with $O(n\log n)$ edges. (4) The intersection graph of $n$ axis-aligned rectangles admits a 3-hop spanner with $O(n\log^2 n)$ edges.
△ Less
Submitted 30 October, 2023; v1 submitted 13 December, 2021;
originally announced December 2021.
-
Densely Connected Convolutional Networks and Signal Quality Analysis to Detect Atrial Fibrillation Using Short Single-Lead ECG Recordings
Authors:
Jonathan Rubin,
Saman Parvaneh,
Asif Rahman,
Bryan Conroy,
Saeed Babaeizadeh
Abstract:
The development of new technology such as wearables that record high-quality single channel ECG, provides an opportunity for ECG screening in a larger population, especially for atrial fibrillation screening. The main goal of this study is to develop an automatic classification algorithm for normal sinus rhythm (NSR), atrial fibrillation (AF), other rhythms (O), and noise from a single channel sho…
▽ More
The development of new technology such as wearables that record high-quality single channel ECG, provides an opportunity for ECG screening in a larger population, especially for atrial fibrillation screening. The main goal of this study is to develop an automatic classification algorithm for normal sinus rhythm (NSR), atrial fibrillation (AF), other rhythms (O), and noise from a single channel short ECG segment (9-60 seconds). For this purpose, signal quality index (SQI) along with dense convolutional neural networks was used. Two convolutional neural network (CNN) models (main model that accepts 15 seconds ECG and secondary model that processes 9 seconds shorter ECG) were trained using the training data set. If the recording is determined to be of low quality by SQI, it is immediately classified as noisy. Otherwise, it is transformed to a time-frequency representation and classified with the CNN as NSR, AF, O, or noise. At the final step, a feature-based post-processing algorithm classifies the rhythm as either NSR or O in case the CNN model's discrimination between the two is indeterminate. The best result achieved at the official phase of the PhysioNet/CinC challenge on the blind test set was 0.80 (F1 for NSR, AF, and O were 0.90, 0.80, and 0.70, respectively).
△ Less
Submitted 10 October, 2017;
originally announced October 2017.
-
Fast Simultaneous Training of Generalized Linear Models (FaSTGLZ)
Authors:
Bryan R. Conroy,
Jennifer M. Walz,
Brian Cheung,
Paul Sajda
Abstract:
We present an efficient algorithm for simultaneously training sparse generalized linear models across many related problems, which may arise from bootstrapping, cross-validation and nonparametric permutation testing. Our approach leverages the redundancies across problems to obtain significant computational improvements relative to solving the problems sequentially by a conventional algorithm. We…
▽ More
We present an efficient algorithm for simultaneously training sparse generalized linear models across many related problems, which may arise from bootstrapping, cross-validation and nonparametric permutation testing. Our approach leverages the redundancies across problems to obtain significant computational improvements relative to solving the problems sequentially by a conventional algorithm. We demonstrate our fast simultaneous training of generalized linear models (FaSTGLZ) algorithm on a number of real-world datasets, and we run otherwise computationally intensive bootstrapping and permutation test analyses that are typically necessary for obtaining statistically rigorous classification results and meaningful interpretation. Code is freely available at http://liinc.bme.columbia.edu/fastglz.
△ Less
Submitted 31 July, 2013;
originally announced July 2013.
