Keigo Kawaji

Prof. Keigo Kawaji

Illinois Institute of Technology, Chicago, Illinois, USA

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 January 2019 in Journal of Cardiovascular Magnetic Resonance
Verified
Journal Article
Open access
PDF file
JEJennifer Erley
Jennifer Erley
DGDavide Genovese
Davide Genovese
NTNatalie Tapaskar
Natalie Tapaskar
NANazia Alvi
Nazia Alvi
NRNina Rashedi
Nina Rashedi
APAmit Patel
Amit Patel
We sought to: (1) determine the agreement in cardiovascular magnetic resonance (CMR) and speckle tracking echocardiography (STE) derived strain measurements, (2) compare their reproducibility, (3) determine which approach is best related to CMR late gadolinium enhancement (LGE). While STE-derived strain is routinely used to assess left ventricular (LV) function, CMR strain measurements are not yet standardized. Strain can be measured using dedicated pulse sequences (strain-encoding, SENC), or post-processing of cine images (feature tracking, FT). It is unclear whether these measurements are interchangeable, and whether strain can be used as an alternative to LGE. Fifty patients underwent 2D echocardiography and 1.5 T CMR. Global longitudinal strain (GLS) was measured by STE (Epsilon), FT (NeoSoft) and SENC (Myocardial Solutions) and circumferential strain (GCS) by FT and SENC. GLS showed good inter-modality agreement (r-values: 0.71–0.75), small biases (< 1%) but considerable limits of agreement (− 7 to 8%). The agreement between the CMR techniques was better for GLS than GCS (r = 0.81 vs 0.67; smaller bias). Repeated measurements showed low intra- and inter-observer variability for both GLS and GCS (intraclass correlations 0.86–0.99; coefficients of variation 3–13%). LGE was present in 22 (44%) of patients. Both SENC- and FT-derived GLS and GCS were associated with LGE, while STE-GLS was not. Irrespective of CMR technique, this association was stronger for GCS (AUC 0.77–0.78) than GLS (AUC 0.67–0.72) and STE-GLS (AUC = 0.58). There is good inter-technique agreement in strain measurements, which were highly reproducible, irrespective of modality or analysis technique. GCS may better reflect the presence of underlying LGE than GLS.

1 January 2020 in Journal of Cardiovascular Magnetic Resonance
Verified
Journal Article
Open access
PDF file
JEJennifer Erley
Jennifer Erley
RTRadu Tanacli
Radu Tanacli
DGDavide Genovese
Davide Genovese
NTNatalie Tapaskar
Natalie Tapaskar
NRNina Rashedi
Nina Rashedi
APAmit Patel
Amit Patel
Right ventricular (RV) strain is a useful predictor of prognosis in various cardiovascular diseases, including those traditionally believed to impact only the left ventricle. We aimed to determine inter-modality and inter-technique agreement in RV longitudinal strain (LS) measurements between currently available cardiovascular magnetic resonance (CMR) and echocardiographic techniques, as well as their reproducibility and the impact of layer-specific strain measurements. RV-LS was determined in 62 patients using 2D speckle tracking echocardiography (STE, Epsilon) and two CMR techniques: feature tracking (FT) and strain-encoding (SENC), and in 17 healthy subjects using FT and SENC only. Measurements included global and free-wall LS (GLS, FWLS). Inter-technique agreement was assessed using linear regression and Bland-Altman analysis. Reproducibility was quantified using intraclass correlation (ICC) and coefficients of variation (CoV). We found similar moderate agreement between both CMR techniques and STE in patients: r = 0.57–0.63 for SENC; r = 0.50–0.62 for FT. The correlation between SENC and STE was better for GLS (r = 0.63) than for FWLS (r = 0.57). Conversely, the correlation between FT and STE was higher for FWLS (r = 0.60–0.62) than GLS (r = 0.50–0.54). FT-midmyocardial strain correlated better with SENC and STE than FT-subendocardial strain. The agreement between SENC and FT was fair (r = 0.36–0.41, bias: − 6.4 to − 10.4%) in the entire study group. All techniques except FT showed excellent reproducibility (ICC: 0.62–0.96, CoV: 0.04–0.30). We found only moderate inter-modality agreement with STE in RV-LS for both FT and SENC and poor agreement when comparing between the CMR techniques. Different modalities and techniques should not be used interchangeably to determine and monitor RV strain.

29 January 2021 in Advanced Healthcare Materials
Verified
Journal Article
Closed access
PDF file
YHYusheng J. He
Yusheng J. He
MSMartin F. Santana
Martin F. Santana
ASAusteja Staneviciute
Austeja Staneviciute
MPMarja B. Pimentel
Marja B. Pimentel
FYFeipeng Yang
Feipeng Yang
GPGeorgia Papavasiliou
Georgia Papavasiliou

15 June 2022 in Journal of Diabetes Science and Technology
Verified
Journal Article
Closed access
MAMohammad Reza Reza Askari
Mohammad Reza Reza Askari
Mudassir M. RashidMudassir M. Rashid
Mudassir M. Rashid
XSXiaoyu Sun
Xiaoyu Sun
Mert SevilMert Sevil
Mert Sevil
ASAndrew Shahidehpour
Andrew Shahidehpour
ACAli Cinar
Ali Cinar
Background: Predicting carbohydrate intake and physical activity in people with diabetes is crucial for improving blood glucose concentration regulation. Patterns of individual behavior can be detected from historical free-living data to predict meal and exercise times. Data collected in free-living may have missing values and forgotten manual entries. While machine learning (ML) can capture meal and exercise times, missing values, noise, and errors in data can reduce the accuracy of ML algorithms. Methods: Two recurrent neural networks (RNNs) are developed with original and imputed data sets to assess detection accuracy of meal and exercise events. Continuous glucose monitoring (CGM) data, insulin infused from pump data, and manual meal and exercise entries from free-living data are used to predict meals, exercise, and their concurrent occurrence. They contain missing values of various lengths in time, noise, and outliers. Results: The accuracy of RNN models range from 89.9% to 95.7% for identifying the state of event (meal, exercise, both, or neither) for various users. “No meal or exercise” state is determined with 94.58% accuracy by using the best RNN (long short-term memory [LSTM] with 1D Convolution). Detection accuracy with this RNN is 98.05% for meals, 93.42% for exercise, and 55.56% for concurrent meal-exercise events. Conclusions: The meal and exercise times detected by the RNN models can be used to warn people for entering meal and exercise information to hybrid closed-loop automated insulin delivery systems. Reliable accuracy for event detection necessitates powerful ML and large data sets. The use of additional sensors and algorithms for detecting these events and their characteristics provides a more accurate alternative.

1 June 2022 in BioMedInformatics
Verified
Journal Article
Open access
PDF file
MAMohammad Reza Reza Askari
Mohammad Reza Reza Askari
Mudassir M. RashidMudassir M. Rashid
Mudassir M. Rashid
Xiaoyu SunXiaoyu Sun
Xiaoyu Sun
Mert SevilMert Sevil
Mert Sevil
ASAndrew Shahidehpour
Andrew Shahidehpour
ACAli Cinar
Ali Cinar
Objective: The interpretation of time series data collected in free-living has gained importance in chronic disease management. Some data are collected objectively from sensors and some are estimated and entered by the individual. In type 1 diabetes (T1D), blood glucose concentration (BGC) data measured by continuous glucose monitoring (CGM) systems and insulin doses administered can be used to detect the occurrences of meals and physical activities and generate the personal daily living patterns for use in automated insulin delivery (AID). Methods: Two challenges in time-series data collected in daily living are addressed: data quality improvement and the detection of unannounced disturbances of BGC. CGM data have missing values for varying periods of time and outliers. People may neglect reporting their meal and physical activity information. In this work, novel methods for preprocessing real-world data collected from people with T1D and the detection of meal and exercise events are presented. Four recurrent neural network (RNN) models are investigated to detect the occurrences of meals and physical activities disjointly or concurrently. Results: RNNs with long short-term memory (LSTM) with 1D convolution layers and bidirectional LSTM with 1D convolution layers have average accuracy scores of 92.32% and 92.29%, and outperform other RNN models. The F1 scores for each individual range from 96.06% to 91.41% for these two RNNs. Conclusions: RNNs with LSTM and 1D convolution layers and bidirectional LSTM with 1D convolution layers provide accurate personalized information about the daily routines of individuals. Significance: Capturing daily behavior patterns enables more accurate future BGC predictions in AID systems and improves BGC regulation.

1 November 2018 in Clinical Imaging
Verified
Journal Article
Closed access
SZS. Javed Zaidi
S. Javed Zaidi
WCWaseem Cossor
Waseem Cossor
ASAmita Singh
Amita Singh
FMFrancesco Maffesanti
Francesco Maffesanti
KKKeigo Kawaji
Keigo Kawaji
APAmit Patel
Amit Patel