@inbook{Abdi:2027:10.1007/978-3-032-35393-1_10,
author = {Abdi, A and Jeviskov, J and Hussain, A and Alibakshi, A and Ufumaka, I and Alajrami, E and Francis, DP and Serej, ND and Zolgharni, M},
doi = {10.1007/978-3-032-35393-1_10},
pages = {131--142},
title = {Regional Myocardial Strain Estimation with MyoTrackerChrono},
url = {http://dx.doi.org/10.1007/978-3-032-35393-1_10},
year = {2027}
}

Download

TY  - CHAP
AB  - Cardiovascular diseases require precise, non-invasive diagnostic tools to assess regional myocardial function. While Speckle Tracking Echocardiography (STE) is widely used to measure myocardial strain, traditional block-matching and optical flow algorithms frequently fail due to acoustic noise and out-of-plane speckle decorrelation. This paper introduces MyoTrackerChrono, a hybrid deep learning architecture designed for robust, sparse point tracking in echocardiography. By integrating a Temporal Adapter between a spatial convolutional encoder and a tracking transformer, the model explicitly aggregates temporal context early in the feature extraction phase. This allows for stable tracking even during rapid cardiac phases or momentary signal loss. Validated on physics-based synthetic benchmarks, MyoTrackerChrono improves pixel-level tracking accuracy (Mean Absolute Error) by 33.0% over baseline models and reduces strain estimation error in mechanically complex mid-wall segments by over 64%. Furthermore, the architecture operates at an effective inference rate of 1382 FPS, confirming its feasibility for rapid post-acquisition clinical deployment.
AU  - Abdi,A
AU  - Jeviskov,J
AU  - Hussain,A
AU  - Alibakshi,A
AU  - Ufumaka,I
AU  - Alajrami,E
AU  - Francis,DP
AU  - Serej,ND
AU  - Zolgharni,M
DO  - 10.1007/978-3-032-35393-1_10
EP  - 142
PY  - 2027///
SP  - 131
TI  - Regional Myocardial Strain Estimation with MyoTrackerChrono
UR  - http://dx.doi.org/10.1007/978-3-032-35393-1_10
ER  - 

Download