Velocity Optimization via Thermal Imaging Enhances Autonomous Tumor Resection

Naveed D. Riaziat; Jiawei Ge; Ethan Kilmer; Michael W. Mikula; Paul Maria Scheikl; Michael Kam; Leila J. Mady; Axel Krieger

cs.ROOct 20, 2026

Velocity Optimization via Thermal Imaging Enhances Autonomous Tumor Resection

Authors: Naveed D. Riaziat, Jiawei Ge, Ethan Kilmer, Michael W. Mikula, Paul Maria Scheikl, Michael Kam, Leila J. Mady, Axel Krieger,

Organizations: Laboratory for Computational Sensing and Robotics, Johns Hopkins University, Baltimore, Maryland · Department of Pathology, Johns Hopkins Hospital, Baltimore, Maryland

Abstract

Objective: To demonstrate and evaluate THERMO, a thermal-velocity optimization algorithm for autonomous electrosurgery, in ex vivo porcine tissue using the UR10e-based ASTR system. Methods: Thermography for Electrosurgical Rate Modulation via Optimization (ThERMO), a closed-loop algorithm that uses real-time thermal imaging data to minimize thermal injury while ensuring accurate cut margins, was tested on ex vivo porcine tissue. Histological analysis was performed to detect thermal injury extent at a cellular level, and 3D photogrammetric measurement was used to verify margin accuracy. ThERMO's performance was compared against two previously used fixed velocity methods. Results: ThERMO outperformed both fixed velocity comparison methods without requiring data-driven training or iterative tuning, reducing thermal injury by up to 52%, improving cut accuracy by up to 59%, and maintaining task completion speed. Conclusions: Closed-loop thermal feedback via ThERMO improves autonomous robotic electrosurgery by simultaneously reducing thermal injury and improving margin accuracy, reinforcing its value over fixed velocity approaches for tumor resection applications. Significance: This work advances autonomous robotic tumor resection by demonstrating that real-time closed-loop thermal feedback control can substantially reduce collateral tissue damage and improve surgical margin accuracy.

Figures & tables

Figure 1: The ASTR system is an autonomous robotic surgical system for conducting autonomous electrosurgical tumor resection on ex vivo tissue. It consists of a UR10e arm (Universal Robots, Odense, Denmark), a Zivid 2+ RGB-D camera (Zivid, Oslo, Norway), and an electrosurgical system. A T3 Pro thermal camera (InfiRay, Yantai, China) is attached to the tool flange to provide thermal feedback for ThERMO . The porcine tongue sample is placed on the grounding pad on the table and the UR10e autonomously performs tumor resection using information from the Zivid and T3 cameras.

Figure 2: (A) A thermal frame is shown from the T3 thermal camera, looking down the shaft of the electrocautery needle. The white location corresponds to tissue contact from the needle, and the hot trail shows the cutting path. (B) The L2 distance transform is shown, in which the color corresponds to the distance a given pixel is from the $60\text{\,}\mathrm{\SIUnitSymbolCelsius}$ temperature boundary. The thermal width is the maximum value of this array.

Thermal Properties		Cost Function Weights
$\rho$	$1.09\times 10^{-6}$ kg/mm 3	$a$	1
$C_{p}$	3800 J/kg/K	$b$	1
$k$	$5.6\times 10^{-4}$ W/mm/K	$c$	5
$\alpha$	$1.34\times 10^{-7}$ mm 2 /s	$r$	0.01

Table 1: Thermal properties:

\rho

(tissue density),

C_{p}

(specific heat),

k

(thermal conductivity),

\alpha

(thermal diffusivity). Thermal properties and cost function weights

a,b,c,r

as defined in Eq. 5 .

Figure 3: Commanded velocity, measured velocity, thermal width, and deflection are plotted with respect to time for a selected trial of ThERMO . Thermal and RGB frames are displayed at three timestamps above the plot, showing the correlation between the visible and thermal cutting states.

Figure 4: (A) Tongues are placed on an electrosurgical grounding pad and sutured as in the partial glossectomy procedure to prevent bulk tissue motion. (B) Pointcloud scan from the Zivid 2+ camera. (C) Projection of tumor outline onto pointcloud.

Figure 5: (A) Tongue sample after 72 hours of formalin fixation. (B) Tongue sample cut to rough size, with red dashed lines indicating how each sample was cut into eight wedge-shaped blocks that evenly sampled the electrosurgical cut spatially. (C) A cross-section of one of the wedge-shaped blocks resulting from the previous step. This face is cut on a microtome to generate three $5\text{\,}\mathrm{\SIUnitSymbolMicro m}$ thick sections from each block. (D) A slide with a H&E stained and mounted tissue section.

Figure 6: (A) A grid of photos shows each tongue sample immediately after cutting. Each cut is shown as a picture on the left and as a point-cloud scan on the right, overlayed with signed margin error (SME) . Each column represents one cutting method: RDFS ( $2\text{\,}\mathrm{mm}\text{/}\mathrm{s}$ ), Clinician-Derived Fixed Speed (CDFS) ( $7\text{\,}\mathrm{mm}\text{/}\mathrm{s}$ ), and ThERMO . The black path in the center of each tongue shows the virtual tumor boundary. The cut trajectory is $5\text{\,}\mathrm{mm}$ from the center of the black edge. (B) One representative H&E -stained slide was selected for each method by the pathologist and is shown from least to most thermal injury. A black line denotes the extent of thermal injury, as measured by a pathologist. Eight measures are shown along this path. Due to the thermal injury width, only the left half of the cut is shown for RDFS .

Figure 7: For all plots, ** refers to

p<0.01

and * refers to

p<0.05

for contrasts from the three linear mixed models (LMMs) (

\alpha=0.05

). (A) Absolute margin error (AME) is plotted in millimeters against cutting method (

n=3240

). ThERMO vs. CDFS showed a significant difference in AME at

p<0.05

. (B) Signed margin error (SME) is plotted in millimeters against cutting method (

n=3240

). Margins with SME less than

-2.2\text{\,}\mathrm{mm}

(red region) are “close margins” [ 5 , 43 ] . RDFS vs. CDFS were significantly different at

p<0.05

. (C) Thermal injury extent (TIE) is plotted in microns against cutting method (

n=216

). The difference between RDFS vs ThERMO and RDFS vs CDFS are significant at

p<0.01

Figure 8: Cutting depth varied slightly during cutting. Statistical analysis revealed that depth was a significant predictor of thermal injury extent (TIE) (

p=0.034

). Cut depth impacted TIE at

32.1\text{\,}\mathrm{\SIUnitSymbolMicro m}\text{/}\mathrm{mm}

. The plotted lines show the linear mixed model (LMM) fit with a 95% confidence interval. ThERMO produced the most accurate and most consistent cut depth (

4.13\text{\,}\mathrm{mm}

), followed by RDFS (

3.65\text{\,}\mathrm{mm}

) and CDFS (

3.20\text{\,}\mathrm{mm}

)

Surgical Robotics

Surgical robotics develops robotic tools and systems for minimally invasive, image-guided, and microsurgical procedures.

See all 187 papers

Occupancy Network-Guided Autonomous Robotic Partial Nephrectomy
Sep 14, 2026Ethan Kilmer, Pit Henrich, Jiawei Ge +13Vision-Based Robot Control Robotic Perception
Laboratory of Computational Sensing and Robotics (LCSR), Johns Hopkins University, Baltimore, MD 21218, USA · Department Artificial Intelligence in Biomedical Engineering (AIBE), Friedrich-Alexander University Erlangen-Nürnberg (FAU), 91052 Erlangen, Germany · Brady Urological Institute, Johns Hopkins University, Baltimore, MD 21287, USA +1
Incision trajectory tracing for electrosurgical navigation by CNN-based knife contacting frames extraction method
Aug 14, 2026Yu Chun Wang, Kaixu Chen, Naoto Ienaga +1Surgical Video Understanding
Degree Programs in Systems and Information Engineering, University of Tsukuba, Tsukuba, 305-8573, Ibaraki, Japan · Center for Computational Sciences, University of Tsukuba, Tsukuba, 305-8577, Ibaraki, Japan · Institute of Systems and Information Engineering, University of Tsukuba, Tsukuba, 305-8573, Ibaraki, Japan
RL-MACRO: A Cybernetic Closed-Loop Intelligence Framework for Multimodal Adaptive Robotic Craniotomy
Jul 23, 2026Xiao Zhang, Jiaxuan Li, Renzhen Le +9Robotic Control Surgical Robotics
Dalian University of Technology, Dalian 116024, China · Second Hospital of Dalian Medical University, Dalian 116024, China · Department of Mechanical Engineering, The University of Tokyo, Tokyo 113-8656, Japan

Sep 14, 2026J/K move · Enter open · S save

Occupancy Network-Guided Autonomous Robotic Partial Nephrectomy

Ethan Kilmer, Pit Henrich, Jiawei Ge, Paul M. Scheikl+12

Laboratory of Computational Sensing and Robotics (LCSR), Johns Hopkins University, Baltimore, MD 21218, USA · Department Artificial Intelligence in Biomedical Engineering (AIBE), Friedrich-Alexander University Erlangen-Nürnberg (FAU), 91052 Erlangen, Germany · Brady Urological Institute, Johns Hopkins University, Baltimore, MD 21287, USA+1

Vision-Based Robot Control Robotic Perception Surgical Robotics

Open paper AskPDF arXiv

Autonomous soft-tissue cancer surgery has been limited to interventions on organ surfaces, because current systems cannot perceive and adapt to anatomy once it deforms or is cut. We introduce the first vision-guided autonomous system capable of performing complete tumor resections for partial nephrectomy. Our system integrates conditional occupancy networks, trained entirely in a physics-based simulation, that infer full 3-D anatomy (tumor, margin tissue, and kidney) from single-view partial point clouds. These occupancy networks maintain intraoperative tracking even as tissue is cut and deformed, enabling adaptive planning and execution. The surgical platform combines a depth camera for capturing surface point clouds, dual robotic arms for electrosurgical cutting and vacuum-based tissue manipulation, and an autonomous control strategy for tumor resection. In patient-derived hydrogel phantoms under an open partial nephrectomy setting, the robot performed eight consecutive autonomous tumor resections comprising 77 electrosurgical cuts, with all cuts achieving negative surgical margins and 1.61

\pm

0.48 mm mean absolute margin error. This work demonstrates, for the first time, a foundation for supervised autonomous closed-loop, imaging-driven, margin-negative tumor removal in phantoms.

Abstract

Surgical Robotics

Explore similar work

Occupancy Network-Guided Autonomous Robotic Partial Nephrectomy

Incision trajectory tracing for electrosurgical navigation by CNN-based knife contacting frames extraction method

RL-MACRO: A Cybernetic Closed-Loop Intelligence Framework for Multimodal Adaptive Robotic Craniotomy