Skip to content
DraconiticResearch
Search Sources API
197,787 studies31 sourcesUpdated 8 min ago
Peer-reviewedCC BYJournal of biomedical optics

Preliminary regression-based feature selection for optimizing spatial frequency domain imaging in burn-severity assessment

  • Thinh Phan1
  • Christopher A. Campbell1
  • Gordon T. Kennedy1
  • Nicole Wakida1
  • Nataliya Makeeva1,2
  • Gabriela Tabone1,2
  • Robert J. Christy3
  • Theresa L. Chin4
  • Victor C. Joe4
  • Bernard Choi1,2,4
  • Anthony J. Durkin1,2
4 affiliations
  1. 1University of California – Irvine, Beckman Laser Institute and Medical Clinic, Irvine, California, United States
  2. 2University of California – Irvine, Department of Biomedical Engineering, Irvine, California, United States
  3. 3University of Texas – San Antonio, Military Health Institute, San Antonio, Texas, United States
  4. 4University of California – Irvine, Department of Surgery, Irvine, California, United States
Read full textPDFDOICite

Abstract

Significance: We identified key spatial frequency domain imaging (SFDI) features for assessing burn wound severity. We built upon our previous two-step regression framework, using SFDI-calibrated reflectance (Rd) maps to predict collagen denaturation depth (CDD) and healing outcome of burn wounds. We used feature selection methods to identify key parameters from a large data set (40 features, including eight wavelengths, each at five spatial frequencies).

Aim: We aimed to identify a reduced set of key SFDI features (i.e., pairs of spatial frequencies and wavelengths) for assessing burn wound severity.

Approach: We utilized a graded-severity porcine burn model (n=4) with eight burn contact durations (5 to 40 s). Multiple linear regression (MLR) was employed to predict CDD from Rd obtained within 24 h post-burn. Subsequently, logistic regression (LogR) was used to predict day-28 healing outcomes based on the predicted CDD. In this study, we fielded three feature selection algorithms (i.e., standardized ordinary least squares, least absolute shrinkage and selection operator, and minimum redundancy maximum relevance) to identify key SFDI features in the MLR model. The resulting prediction then dictates the LogR outputs. We assessed the models for accuracy and fit performance against those of the 40-feature original model.

Results: A subset of 13 LASSO features matched the predictive performance of the 40-feature baseline (RMSE 98 versus 105 μm; within fold-to-fold noise), and a minimum set of eight LASSO features (minimum features with acceptable errors) closely approximated it. Across all three methods, the consistently top-ranked features were planar (0 mm−1) reflectance at 526, 591, and 621 nm.

Conclusions: Using feature selection algorithms, we identified preliminary candidate features that may be suitable for developing a future SFDI system to assess burn wound severity with reduced complexity. This has the potential to enable a more compact design compared with the current research-grade SFDI system.

Related

  1. Optimizing Behavioral Strategy Selection during Cognitive Behavior Therapy for Procrastination using Machine Learning
    Devon LoParo, Magdalena Pietruch, Weronika Maria Browarczyk +3
    PsyArXivPreprintCC BY 4.0Full textPsychiatry & mental healthOSF
  2. A preliminary study on the assessment of bone age in children and adolescents in frigid-zone regions by the CHN-05 method
    Yue Qi, Xinyi Li, Dongmei Lv +4
    Frontiers in medicinePeer-reviewedCC BYFull textGeneral medicineEurope PMC
  3. Designing and optimizing an intelligent health informatics management platform for patients undergoing chemotherapy for breast cancer: from needs assessment to pilot testing
    Xiaosong Yu, Lin Zhang, Li Yang +2
    Frontiers in OncologyPeer-reviewedCC BY 4.0Full textOncologyCrossref
  4. A quantitative treatment assessment for optimizing combined virotherapy-chemotherapy regimens against HPV-induced cervical cancer
    Salil Ghosh, Amit Kumar Bag, Satyajit Mukherjee +3
    PLOS OnePeer-reviewedCC BY 4.0Full textOncologyPLOS
  5. Exercise for tension-type headache: protocol for a systematic review and meta-analysis with meta-regression and assessment of exercise reporting completeness.
    Ana Bravo-Vazquez, Carlos Bernal-Utrera, Ernesto Anarte-Lazo +2
    BMJ openPeer-reviewedCC BYFull textPsychiatry & mental healthEurope PMCCrossref

Cite

BibTeXRIS
BibTeX
@article{phan2026preliminary,
  title = {{Preliminary regression-based feature selection for optimizing spatial frequency domain imaging in burn-severity assessment}},
  author = {Thinh Phan and Christopher A. Campbell and Gordon T. Kennedy and Nicole Wakida and Nataliya Makeeva and Gabriela Tabone and Robert J. Christy and Theresa L. Chin and Victor C. Joe and Bernard Choi and Anthony J. Durkin},
  journal = {Journal of biomedical optics},
  year = {2026},
  month = oct,
  doi = {10.1117/1.jbo.31.10.106007},
  url = {https://europepmc.org/article/PMC/PMC13722507},
  note = {CC BY},
}
RIS
TY  - JOUR
TI  - Preliminary regression-based feature selection for optimizing spatial frequency domain imaging in burn-severity assessment
AU  - Thinh Phan
AU  - Christopher A. Campbell
AU  - Gordon T. Kennedy
AU  - Nicole Wakida
AU  - Nataliya Makeeva
AU  - Gabriela Tabone
AU  - Robert J. Christy
AU  - Theresa L. Chin
AU  - Victor C. Joe
AU  - Bernard Choi
AU  - Anthony J. Durkin
PY  - 2026
DA  - 2026/10/01
JO  - Journal of biomedical optics
DO  - 10.1117/1.jbo.31.10.106007
UR  - https://europepmc.org/article/PMC/PMC13722507
AB  - Significance: We identified key spatial frequency domain imaging (SFDI) features for assessing burn wound severity. We built upon our previous two-step regression framework, using SFDI-calibrated reflectance (Rd) maps to predict collagen denaturation depth (CDD) and healing outcome of burn wounds. We used feature selection methods to identify key parameters from a large data set (40 features, including eight wavelengths, each at five spatial frequencies).  Aim: We aimed to identify a reduced set of key SFDI features (i.e., pairs of spatial frequencies and wavelengths) for assessing burn wound severity.  Approach: We utilized a graded-severity porcine burn model (n=4) with eight burn contact durations (5 to 40 s). Multiple linear regression (MLR) was employed to predict CDD from Rd obtained within 24 h post-burn. Subsequently, logistic regression (LogR) was used to predict day-28 healing outcomes based on the predicted CDD. In this study, we fielded three feature selection algorithms (i.e., standardized ordinary least squares, least absolute shrinkage and selection operator, and minimum redundancy maximum relevance) to identify key SFDI features in the MLR model. The resulting prediction then dictates the LogR outputs. We assessed the models for accuracy and fit performance against those of the 40-feature original model.  Results: A subset of 13 LASSO features matched the predictive performance of the 40-feature baseline (RMSE 98 versus 105 μm; within fold-to-fold noise), and a minimum set of eight LASSO features (minimum features with acceptable errors) closely approximated it. Across all three methods, the consistently top-ranked features were planar (0 mm−1) reflectance at 526, 591, and 621 nm.  Conclusions: Using feature selection algorithms, we identified preliminary candidate features that may be suitable for developing a future SFDI system to assess burn wound severity with reduced complexity. This has the potential to enable a more compact design compared with the current research-grade SFDI system.
SN  - 1083-3668
SN  - 1560-2281
ER  -