Enhancing predictive reliability in medical laser procedures through physics-informed neural networks for bioheat modeling

Name: Teema Leangarun

Organization : King Mongkut’s University of Technology Thonburi. Dept. of Control Systems and Instrumentation Engineering

Email : teema.lea@kmutt.ac.th

Name: Mangkorn Chunvijitra

Organization : King Mongkut’s University of Technology Thonburi. Dept. of Control Systems and Instrumentation Engineering

Email : mangkorn.chunvi@kmutt.ac.th

Name: Wutipong Preechaphonkul

Organization : King Mongkut’s University of Technology Thonburi. Institute of Field Robotics (FIBO)

Email : wutipong.pree@kmutt.ac.th

keyword: Bioheat transfer

LCSH: Lasers in medicine -- Safety measures

LCSH: Neural networks (Computer science)

LCSH: Finite differences

LCSH: Medical instruments and apparatus -- Reliability -- Testing

Abstract: Laser applications in medical procedures require precise thermal control to achieve selective photothermolysis while strictly preventing thermal necrosis in surrounding healthy tissues. While traditional numerical approaches such as the Finite Difference Method (FDM) provide the high accuracy necessary to model these interactions, they are often computationally prohibitive for real-time control applications. Conversely, standard Deep Neural Networks (DNNs) provide rapid inference but suffer from poor generalization in data-sparse regimes. This paper presents a comparative study of FDM, DNN, and Physics-Informed Neural Networks (PINNs) for modeling 1D laser-tissue thermal interactions based on the Pennes Bioheat Equation. We investigated two scenarios: a homogeneous simple model and a realistic multi-layer skin model. Our results demonstrate that PINNs significantly outperform standard DNNs in extreme sparse regimes (N=20), achieving an RMSE approximately 38 times lower in the simple model. While a performance crossover occurs in the complex model at N>=100 where DNNs yield higher numerical precision, the unconstrained flexibility of DNNs leads to non-physical deviations during sensor gaps. In contrast, PINNs maintain high fidelity to biological heat transfer laws, providing a physically plausible and robust framework for real-time thermal monitoring in safety-critical medical procedures.

King Mongkut's University of Technology North Bangkok. Central Library

Address: BANGKOK

Email: library@kmutnb.ac.th

Created: 2026

Modified: 2026-08-10

Issued: 2026-08-10

บทความ/Article

application/pdf

BibliograpyCitation : In ECTI Association and Burapha University. Faculty of Engineering. 2026 23rd International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology (ECTI-CON 2026) (pp.384-389). Piscataway, NJ : IEEE

eng

©copyrights King Mongkut's University of Technology North Bangkok

ลำดับที่.	ชื่อแฟ้มข้อมูล	ขนาดแฟ้มข้อมูล	จำนวนเข้าถึง	วัน-เวลาเข้าถึงล่าสุด
1	ECTI-CON 2026pp.384-389.pdf	1.22 MB

ใช้เวลา

-0.968442 วินาที