Deep Learning Framework for Real-Time Label-Free Thyroid Cancer Detection Using Second Harmonic Generation Microscopy

Souradeep Bhattacharya; Wesley Poon; Jun Liu; Hong Zhao; Helmi S. Khadra; Elizabeth M. Jacobi; Raksha Raghunathan; Stephen T. Wong

doi:10.1117/12.3079441

Deep Learning Framework for Real-Time Label-Free Thyroid Cancer Detection Using Second Harmonic Generation Microscopy

Souradeep Bhattacharya, Wesley Poon, Jun Liu, Hong Zhao, Helmi S. Khadra, Elizabeth M. Jacobi, Raksha Raghunathan, Stephen T. Wong

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Fine needle aspiration (FNA) is the primary diagnostic tool for thyroid cancer but has key limitations, including a 21% false-negative rate and 30% indeterminate results, often leading to repeat biopsies, expensive molecular testing, and unnecessary surgeries. Second harmonic generation (SHG) microscopy, a nonlinear optical technique, is being studied as a label-free method of diagnosing thyroid cancer due to its ability to detect malignancy-associated changes in collagen architecture. Although computational methods have demonstrated potential in distinguishing normal from cancerous thyroid tissue, integrating SHG with artificial intelligence (AI)-based analysis can greatly improve diagnostic speed and accuracy. In this study, SHG images acquired from normal and cancerous thyroid sections were analyzed using various deep learning models to assess their diagnostic performance and identify the optimal architecture for clinical deployment. Results showed that DenseNet-121 achieved superior performance with 96.3% AUC (81.3% sensitivity, 92.8% specificity), while ensemble methods reached 96.6% AUC, with all models demonstrating real-time inference capabilities (12ms).

Original language	English (US)
Title of host publication	Multiphoton Microscopy in the Biomedical Sciences XXVI
Editors	Ammasi Periasamy, Peter T. C. So, Karsten Konig
Publisher	SPIE
ISBN (Electronic)	9781510696259
DOIs	https://doi.org/10.1117/12.3079441
State	Published - Mar 5 2026
Event	26th Multiphoton Microscopy in the Biomedical Sciences - San Francisco, United States Duration: Jan 18 2026 → Jan 20 2026

Publication series

Name	Proceedings of SPIE - The International Society for Optical Engineering
Volume	13856
ISSN (Print)	0277-786X
ISSN (Electronic)	1996-756X

Conference

Conference	26th Multiphoton Microscopy in the Biomedical Sciences
Country/Territory	United States
City	San Francisco
Period	1/18/26 → 1/20/26

Keywords

Convolutional neural networks
Deep learning
Knowledge distillation
Label-free imaging
Optical biopsy
Real-time diagnosis
Second harmonic generation microscopy
Thyroid cancer

ASJC Scopus subject areas

Electronic, Optical and Magnetic Materials
Instrumentation
Condensed Matter Physics
Computer Science Applications
Applied Mathematics
Electrical and Electronic Engineering

Access to Document

10.1117/12.3079441

Cite this

Bhattacharya, S., Poon, W., Liu, J., Zhao, H., Khadra, H. S., Jacobi, E. M., Raghunathan, R., & Wong, S. T. (2026). Deep Learning Framework for Real-Time Label-Free Thyroid Cancer Detection Using Second Harmonic Generation Microscopy. In A. Periasamy, P. T. C. So, & K. Konig (Eds.), Multiphoton Microscopy in the Biomedical Sciences XXVI Article 1385606 (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 13856). SPIE. https://doi.org/10.1117/12.3079441

Deep Learning Framework for Real-Time Label-Free Thyroid Cancer Detection Using Second Harmonic Generation Microscopy. / Bhattacharya, Souradeep; Poon, Wesley; Liu, Jun et al.
Multiphoton Microscopy in the Biomedical Sciences XXVI. ed. / Ammasi Periasamy; Peter T. C. So; Karsten Konig. SPIE, 2026. 1385606 (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 13856).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Bhattacharya, S, Poon, W, Liu, J, Zhao, H , Khadra, HS , Jacobi, EM , Raghunathan, R & Wong, ST 2026, Deep Learning Framework for Real-Time Label-Free Thyroid Cancer Detection Using Second Harmonic Generation Microscopy. in A Periasamy, PTC So & K Konig (eds), Multiphoton Microscopy in the Biomedical Sciences XXVI., 1385606, Proceedings of SPIE - The International Society for Optical Engineering, vol. 13856, SPIE, 26th Multiphoton Microscopy in the Biomedical Sciences, San Francisco, United States, 1/18/26. https://doi.org/10.1117/12.3079441

Bhattacharya S, Poon W, Liu J, Zhao H , Khadra HS , Jacobi EM et al. Deep Learning Framework for Real-Time Label-Free Thyroid Cancer Detection Using Second Harmonic Generation Microscopy. In Periasamy A, So PTC, Konig K, editors, Multiphoton Microscopy in the Biomedical Sciences XXVI. SPIE. 2026. 1385606. (Proceedings of SPIE - The International Society for Optical Engineering). doi: 10.1117/12.3079441

Bhattacharya, Souradeep ; Poon, Wesley ; Liu, Jun et al. / Deep Learning Framework for Real-Time Label-Free Thyroid Cancer Detection Using Second Harmonic Generation Microscopy. Multiphoton Microscopy in the Biomedical Sciences XXVI. editor / Ammasi Periasamy ; Peter T. C. So ; Karsten Konig. SPIE, 2026. (Proceedings of SPIE - The International Society for Optical Engineering).

@inproceedings{5eb22d756eb643b998085de9ec669b1f,

title = "Deep Learning Framework for Real-Time Label-Free Thyroid Cancer Detection Using Second Harmonic Generation Microscopy",

abstract = "Fine needle aspiration (FNA) is the primary diagnostic tool for thyroid cancer but has key limitations, including a 21\% false-negative rate and 30\% indeterminate results, often leading to repeat biopsies, expensive molecular testing, and unnecessary surgeries. Second harmonic generation (SHG) microscopy, a nonlinear optical technique, is being studied as a label-free method of diagnosing thyroid cancer due to its ability to detect malignancy-associated changes in collagen architecture. Although computational methods have demonstrated potential in distinguishing normal from cancerous thyroid tissue, integrating SHG with artificial intelligence (AI)-based analysis can greatly improve diagnostic speed and accuracy. In this study, SHG images acquired from normal and cancerous thyroid sections were analyzed using various deep learning models to assess their diagnostic performance and identify the optimal architecture for clinical deployment. Results showed that DenseNet-121 achieved superior performance with 96.3\% AUC (81.3\% sensitivity, 92.8\% specificity), while ensemble methods reached 96.6\% AUC, with all models demonstrating real-time inference capabilities (12ms).",

keywords = "Convolutional neural networks, Deep learning, Knowledge distillation, Label-free imaging, Optical biopsy, Real-time diagnosis, Second harmonic generation microscopy, Thyroid cancer",

author = "Souradeep Bhattacharya and Wesley Poon and Jun Liu and Hong Zhao and Khadra, \{Helmi S.\} and Jacobi, \{Elizabeth M.\} and Raksha Raghunathan and Wong, \{Stephen T.\}",

year = "2026",

month = mar,

day = "5",

doi = "10.1117/12.3079441",

language = "English (US)",

series = "Proceedings of SPIE - The International Society for Optical Engineering",

publisher = "SPIE",

editor = "Ammasi Periasamy and So, \{Peter T. C.\} and Karsten Konig",

booktitle = "Multiphoton Microscopy in the Biomedical Sciences XXVI",

address = "United States",

}

TY - GEN

T1 - Deep Learning Framework for Real-Time Label-Free Thyroid Cancer Detection Using Second Harmonic Generation Microscopy

AU - Bhattacharya, Souradeep

AU - Poon, Wesley

AU - Liu, Jun

AU - Zhao, Hong

AU - Khadra, Helmi S.

AU - Jacobi, Elizabeth M.

AU - Raghunathan, Raksha

AU - Wong, Stephen T.

PY - 2026/3/5

Y1 - 2026/3/5

N2 - Fine needle aspiration (FNA) is the primary diagnostic tool for thyroid cancer but has key limitations, including a 21% false-negative rate and 30% indeterminate results, often leading to repeat biopsies, expensive molecular testing, and unnecessary surgeries. Second harmonic generation (SHG) microscopy, a nonlinear optical technique, is being studied as a label-free method of diagnosing thyroid cancer due to its ability to detect malignancy-associated changes in collagen architecture. Although computational methods have demonstrated potential in distinguishing normal from cancerous thyroid tissue, integrating SHG with artificial intelligence (AI)-based analysis can greatly improve diagnostic speed and accuracy. In this study, SHG images acquired from normal and cancerous thyroid sections were analyzed using various deep learning models to assess their diagnostic performance and identify the optimal architecture for clinical deployment. Results showed that DenseNet-121 achieved superior performance with 96.3% AUC (81.3% sensitivity, 92.8% specificity), while ensemble methods reached 96.6% AUC, with all models demonstrating real-time inference capabilities (12ms).

AB - Fine needle aspiration (FNA) is the primary diagnostic tool for thyroid cancer but has key limitations, including a 21% false-negative rate and 30% indeterminate results, often leading to repeat biopsies, expensive molecular testing, and unnecessary surgeries. Second harmonic generation (SHG) microscopy, a nonlinear optical technique, is being studied as a label-free method of diagnosing thyroid cancer due to its ability to detect malignancy-associated changes in collagen architecture. Although computational methods have demonstrated potential in distinguishing normal from cancerous thyroid tissue, integrating SHG with artificial intelligence (AI)-based analysis can greatly improve diagnostic speed and accuracy. In this study, SHG images acquired from normal and cancerous thyroid sections were analyzed using various deep learning models to assess their diagnostic performance and identify the optimal architecture for clinical deployment. Results showed that DenseNet-121 achieved superior performance with 96.3% AUC (81.3% sensitivity, 92.8% specificity), while ensemble methods reached 96.6% AUC, with all models demonstrating real-time inference capabilities (12ms).

KW - Convolutional neural networks

KW - Deep learning

KW - Knowledge distillation

KW - Label-free imaging

KW - Optical biopsy

KW - Real-time diagnosis

KW - Second harmonic generation microscopy

KW - Thyroid cancer

UR - https://www.scopus.com/pages/publications/105035725712

UR - https://www.scopus.com/inward/citedby.url?scp=105035725712&partnerID=8YFLogxK

U2 - 10.1117/12.3079441

DO - 10.1117/12.3079441

M3 - Conference contribution

AN - SCOPUS:105035725712

T3 - Proceedings of SPIE - The International Society for Optical Engineering

BT - Multiphoton Microscopy in the Biomedical Sciences XXVI

A2 - Periasamy, Ammasi

A2 - So, Peter T. C.

A2 - Konig, Karsten

PB - SPIE

T2 - 26th Multiphoton Microscopy in the Biomedical Sciences

Y2 - 18 January 2026 through 20 January 2026

ER -

Deep Learning Framework for Real-Time Label-Free Thyroid Cancer Detection Using Second Harmonic Generation Microscopy

Abstract

Publication series

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this