Fan Lu

Dr. Fan Lu


Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

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

Representative Publications

View All Publications
31 July 2024 in Royal Society Open Science
Verified
Journal Article
Open access
FLFan Lu
Fan Lu
KZKsenia Zlobina
Ksenia Zlobina
NRNicholas A. Rondoni
Nicholas A. Rondoni
STSam Teymoori
Sam Teymoori
MGMarcella Gomez
Marcella Gomez
Finding the optimal treatment strategy to accelerate wound healing is of utmost importance, but it presents a formidable challenge owing to the intrinsic nonlinear nature of the process. We propose an adaptive closed-loop control framework that incorporates deep learning, optimal control and reinforcement learning to accelerate wound healing. By adaptively learning a linear representation of nonlinear wound healing dynamics using deep learning and interactively training a deep reinforcement learning agent for tracking the optimal signal derived from this representation without the need for intricate mathematical modelling, our approach has not only successfully reduced the wound healing time by 45.56% compared to the one without any treatment, but also demonstrates the advantages of offering a safer and more economical treatment strategy. The proposed methodology showcases a significant potential for expediting wound healing by effectively integrating perception, predictive modelling and optimal adaptive control, eliminating the need for intricate mathematical models.

11 July 2025 in Bioengineering
Verified
Journal Article
Open access
PDF file
FLFan Lu
Fan Lu
KZKsenia Zlobina
Ksenia Zlobina
PBPrabhat Baniya
Prabhat Baniya
Houpu LiHoupu Li
Houpu Li
NRNicholas Rondoni
Nicholas Rondoni
MGMarcella Gomez
Marcella Gomez
Advancements in bioelectronic sensors and actuators have paved the way for real-time monitoring and control of the progression of wound healing. Real-time monitoring allows for precise adjustment of treatment strategies to align them with an individual’s unique biological response. However, due to the complexities of human–drug interactions and a lack of predictive models, it is challenging to determine how one should adjust drug dosage to achieve the desired biological response. This work proposes an adaptive closed-loop control framework that integrates deep learning, optimal control, and reinforcement learning to update treatment strategies in real time, with the goal of accelerating wound closure. The proposed approach eliminates the need for mathematical modeling of complex nonlinear wound-healing dynamics. We demonstrate the convergence of the controller via an in silico experimental setup, where the proposed approach successfully accelerated the wound-healing process by 17.71%. Finally, we share the experimental setup and results of an in vivo implementation to highlight the translational potential of our work. Our data-driven model suggests that the treatment strategy, as determined by our deep reinforcement learning algorithm, results in an accelerated onset of inflammation and subsequent transition to proliferation in a porcine wound model.

20 December 2024 in bioRxiv
Verified
Preprint
Open access
PDF file
FLFan Lu
Fan Lu
KZKsenia Zlobina
Ksenia Zlobina
SOSebastian Osorio
Sebastian Osorio
HYHsin-Ya Yang
Hsin-Ya Yang
ANAlexandra Nava
Alexandra Nava
MGMarcella Gomez
Marcella Gomez
Advancements in bioelectronic sensors and actuators have paved the way for real-time monitoring and control of wound healing progression. Real-time monitoring allows for precise adjustment in treatment strategies that align with an individual’s unique biological response. However, due to the complexities of human-drug interactions and a lack of predictive models it is challenging to determine just how one should adjust drug dosage to achieve the desired biological response. This work proposes an adaptive closed-loop control framework that integrates deep learning, optimal control, and reinforcement learning to update treatment strategies in real-time with the goal of accelerating wound closure. The proposed approach eliminates the need for mathematical modeling of complex nonlinear wound healing dynamics. We demonstrate the convergence of the controller via an in silico experimental setup, where the proposed approach successfully accelerates the wound healing process by 17.71%. Finally, we share the experimental setup and results of an in vivo implementation to highlight the translational potential of our work. Our data-driven model estimates a 40% acceleration in wound closure.

21 December 2024 in bioRxiv
Verified
Preprint
Open access
PDF file
FLFan Lu
Fan Lu
KZKsenia Zlobina
Ksenia Zlobina
PBPrabhat Baniya
Prabhat Baniya
Houpu LiHoupu Li
Houpu Li
NRNicholas A. Rondoni
Nicholas A. Rondoni
MGMarcella Gomez
Marcella Gomez
Advancements in bioelectronic sensors and actuators have paved the way for real-time monitoring and control of wound healing progression. Real-time monitoring allows for precise adjustment in treatment strategies that align with an individual’s unique biological response. However, due to the complexities of human-drug interactions and a lack of predictive models it is challenging to determine just how one should adjust drug dosage to achieve the desired biological response. This work proposes an adaptive closed-loop control framework that integrates deep learning, optimal control, and reinforcement learning to update treatment strategies in real-time with the goal of accelerating wound closure. The proposed approach eliminates the need for mathematical modeling of complex nonlinear wound healing dynamics. We demonstrate the convergence of the controller via an in silico experimental setup, where the proposed approach successfully accelerates the wound healing process by 17.71%. Finally, we share the experimental setup and results of an in vivo implementation to highlight the translational potential of our work. Our data-driven model estimates a 40% acceleration in wound closure.

10 September 2023 in arXiv
Verified
Preprint
Open access
PDF file
FLFan Lu
Fan Lu
SMSean Meyn
Sean Meyn
The paper introduces the first formulation of convex Q-learning for Markov decision processes with function approximation. The algorithms and theory rest on a relaxation of a dual of Manne's celebrated linear programming characterization of optimal control. The main contributions firstly concern properties of the relaxation, described as a deterministic convex program: we identify conditions for a bounded solution, and a significant relationship between the solution to the new convex program, and the solution to standard Q-learning. The second set of contributions concern algorithm design and analysis: (i) A direct model-free method for approximating the convex program for Q-learning shares properties with its ideal. In particular, a bounded solution is ensured subject to a simple property of the basis functions; (ii) The proposed algorithms are convergent and new techniques are introduced to obtain the rate of convergence in a mean-square sense; (iii) The approach can be generalized to a range of performance criteria, and it is found that variance can be reduced by considering ``relative'' dynamic programming equations; (iv) The theory is illustrated with an application to a classical inventory control problem.

14 April 2026 in Advanced Intelligent Discovery
Verified
Journal Article
Open access
PDF file
FLFan Lu
Fan Lu
KZKsenia Zlobina
Ksenia Zlobina
MBMichael Briden
Michael Briden
OAOscar Avalos
Oscar Avalos
HCHéctor Carrión
Héctor Carrión
MGMarcella Gomez
Marcella Gomez
Quantifying wound progression is an important step towards designing intervention strategies to achieve accelerated wound closure. Wound healing is a complex, nonlinear dynamical process that progresses through four key biological stages: hemostasis, inflammation, proliferation, and maturation. Being able to track the system's progression through each of the stages allows one to assess the system's response to specific interventions such as the application of topical drugs or an electric field. Furthermore, if we can learn a model that describes the underlying system, we can leverage predictions to customize an intervention strategy to the individual. Wound stages can be visually identified through wound imaging with the help of deep neural networks. However, accurately predicting these stages from images is challenging due to high variability, the nonlinear nature of wound progression, and the scarcity of labeled data. To address these challenges, we introduce DeepMapper, an unsupervised deep learning approach capable of learning a linear representation of the complex, nonlinear wound healing dynamics from a time series of wound images, through which wound stages are quantified, and future states of the wound can be forecasted.