Search papers, labs, and topics across Lattice.
This study introduces a Physics-Flavored Neural Network (PFNN) that automates the kinetic phenotyping of engineered skeletal muscle tissues (ESMs) by integrating a stretched-exponential physical model into a CNN-Transformer architecture. The PFNN addresses the challenge of extracting meaningful kinetic parameters from force-time profiles while overcoming the scarcity of labeled biological data through a hybrid training paradigm that combines synthetic data with unsupervised self-alignment on real-world measurements. The results show that this approach achieves high-fidelity parameterization across various contractile phenotypes and cell lines, including models for Duchenne Muscular Dystrophy, thus enhancing the functional characterization of ESMs for biomedical applications.
A novel neural network architecture captures complex contraction dynamics of engineered muscle tissues, achieving high fidelity in parameterization even with limited labeled data.
Engineered Skeletal Muscle Tissues (ESMs) have become a key structure for biomedical disease modeling and pharmacological screening, yet their functional characterization often relies on simplistic metrics like peak force, discarding critical kinetic information. This is partially due to the high level of mathematical complexity which mechanistic models introduce to capture these dynamics. Hence, exactly the complexity prevents scalable application and widespread adaptation in the field. Here we present a Physics-Flavored Neural Network (PFNN) that automates the kinetic phenotyping of ESMs. Our architecture integrates a stretched-exponential physical model into a CNN-Transformer, enabling the extraction of physically meaningful parameters directly from force-time profiles. To address the scarcity of labeled biological data, we employ a hybrid training paradigm: the model develops a "physical intuition" on synthetic data before undergoing unsupervised self-alignment on unlabeled real-world measurements. Our results demonstrate that this physics-flavored approach achieves high-fidelity parameterization across diverse contractile phenotypes and cell lines, including Duchenne Muscular Dystrophy models. Our scalable, self-improving pipeline bridges the gap between idealized biophysics and noisy \emph{in vitro} data, providing a robust tool for high-throughput biophysical research.
