The Computational Chemical Imaging (CCI) Lab focuses on the development of computation-driven vibrational chemical imaging technologies, to non-invasively measure the biochemical processes in living systems and identify novel biomarkers for diseases, including:

• Developing vibrational chemical imaging platforms beyond the physical limits by computation-driven system design and image reconstruction. 

• Discovering unknown biochemical mechanisms and disease biomarkers using novel chemical imaging data and AI / ML algorithms.

Human health and diseases are governed by the biochemical processes across scales and systems. Yet, tools to visualize the intracellular small biomolecules and pathways in action remain limited. Our lab is at the forefront of developing vibrational chemical imaging, including stimulated Raman scattering (SRS) and vibrational photothermal imaging (VIP), to directly measure the intrinsic chemical bond vibrations for tracking and quantifying biomolecules. We present a new paradigm to break the “no free lunch” limit in instrumentation by adding computation in the loop of system design, data collection, and image analysis. 

We have originated numerous computation-driven instruments and algorithms that deliver next-generation SRS microscopes with unprecedented spatial resolution (Nat. Methods, 2025), imaging speed (Light Sci. Appl. 2018; Nat. Commun. 2021), and molecular specificity (Sci. Adv. 2023; Adv. Sci. 2023; Theranostics 2024). Beyond SRS, we also work on novel label-free vibrational photothermal imaging modalities via stimulated Raman or overtone absorption (Sci. Adv. 2023; Nat. Commun. 2024) to push sensitivity and penetration. These technological breakthroughs have enabled collaborative biomedical applications across scales, including synthetic biology, molecular virology, histopathology, cell metabolism, neurodegeneration, and aging.

Research Area 1. Pushing the Resolution Limit of Chemical Imaging

Super-resolution imaging of cell metabolism is hindered by the incompatibility of small metabolites with fluorescent dyes and the limited resolution of imaging mass spectrometry. We developed ultrasensitive reweighted visible stimulated Raman scattering (URV-SRS), a label-free vibrational chemical imaging technique for multiplexed nanoscopy of intracellular metabolites. We developed a visible SRS microscope with extensive pulse chirping to improve the detection limit to ~4,000 molecules and introduced a self-supervised multi-agent denoiser to suppress non-independent noise in SRS by over 7.2 dB, resulting in a 50-fold sensitivity enhancement over near-infrared SRS. Leveraging the enhanced sensitivity, we employed Fourier reweighting to amplify sub-100-nm spatial frequencies that were previously overwhelmed by noise, reaching a lateral resolution of 86 nm in cell imaging. We visualized the reprogramming of metabolic nanostructures associated with virus replication in host cells and subcellular fatty acid synthesis in engineered bacteria, demonstrating its capability towards nanoscopic spatial metabolomics.

• Lin H, Seitz S, Tan Y, Lugagne JB, Wang L, Ding G, He H, Rauwolf TJ, Dunlop MJ, Connor JH, Porco JA Jr, Tian L, Cheng JX. Label-free nanoscopy of cell metabolism by ultrasensitive reweighted visible stimulated Raman scattering. Nat Methods. 2025 May;22(5):1040-1050. 

Research Area 2. Pushing the Speed Limit of Chemical Imaging

Vibrational hyperspectral imaging via stimulated Raman Scattering (SRS) provides spatially resolved spectroscopy to facilitate molecular fingerprinting in cells and tissues. One fundamental challenge is the slow acquisition speed of hyperspectral imaging, limiting its applications to dynamic or large-scale systems. We developed a series of "instrumentation + computation" approaches to achieve ultrafast hyperspectral SRS imaging. These include 1) compressive sampling and matrix completion to measure 20% pixels while retaining full spatial and spectral information; 2) ultrafast delay line tuning with a polygon scanner that acquired fingerprint SRS spectrum within 20 microseconds; 3) single-shot femtosecond SRS with deep learning to map chemically specific organellar maps at video rate. 

• Lin H, Lee HJ, Tague N, Lugagne JB, Zong C, Deng F, Shin J, Tian L, Wong W, Dunlop MJ, Cheng JX. Microsecond fingerprint stimulated Raman spectroscopic imaging by ultrafast tuning and spatial-spectral learning. Nat Commun. 2021 May 24;12(1):3052.

• Lin H, Liao CS, Wang P, Kong N, Cheng JX. Spectroscopic stimulated Raman scattering imaging of highly dynamic specimens through matrix completion. Light Sci Appl. 2018;7:17179.

• Zhang J*, Zhao J*, Lin H* (equal contribution), Tan Y, Cheng JX. High-Speed Chemical Imaging by Dense-Net Learning of Femtosecond Stimulated Raman Scattering. J Phys Chem Lett. 2020 Oct 15;11(20):8573-8578.

Research Area 3. Pushing the Specificity Limit of Chemical Imaging

Vibrational hyperspectral imaging offers comprehensive spectral signatures for high-content mapping of biomolecules such as proteins, lipids, nuelcic acids, etc. We develop machine learning algorithms augmented by domain knowledge to robustly translate the high-dimensional hyperspectral images into interpretable chemical concentration maps. Our contributions to this area include 1) pixel-wise least absolute shrinkage and selection operator (LASSO) for spectral unmixing of carbon-hydrogen window SRS images with minimal spectral crosstalk, achieving simultaneous of major metabolites in single cells, including protein, lipid, cholesterol, nucleic acid, and carbohydrate; 2) spectral selective sampling and LASSO spectral unmixing for 5-color chemical histology in tumor; 3) mass spectroscopy-augmented spectral unmixing for mapping chain length and unsaturation of biofuel in single engineered bacteria.

• Ni H*, Dessai CP*, Lin H* (equal contribution), Wang W, Chen S, Yuan Y, Ge X, Ao J, Vild N, Cheng JX. High-content stimulated Raman histology of human breast cancer. Theranostics. 2024;14(4):1361-1370.

• Tan Y*, Lin H* (equal contribution), Cheng JX. Profiling single cancer cell metabolism via high-content SRS imaging with chemical sparsity. Sci Adv. 2023 Aug 18;9(33):eadg6061.

• Tague N*, Lin H* (equal contribution), Lugagne JB, O'Connor OM, Burman D, Wong WW, Cheng JX, Dunlop MJ. Longitudinal Single-Cell Imaging of Engineered Strains with Stimulated Raman Scattering to Characterize Heterogeneity in Fatty Acid Production. Adv Sci. 2023 Jul;10(20):e2206519.

Research Area 4. Pushing the Sensitivity Limit by Model-based Deep Learning

An fundamental challenge of vibrational chemical imaging is the limited millimolar sensitivity, making it challenging to study a wide range of intracellular biomolecules at micromolar concentrations. In addition to physically enhancing the signal via advanced instrumentation, e.g., electronic pre-resonance (Adv Sci 2021) or plasmonic enhancement (Sci Adv 2026), the detection sensitivity can be synergistically improved by computationally suppressing noise. We developed a series of deep learning denoising networks that are specialized for handling vibrational chemical imaging, which have unique noise statistics and physical priors. These include 1) Spatial-Spectral Residual Net (SS-ResNet), a 3D U-Net with a 2+1 CNN kernel for denoising hyperspectral SRS imaging with high training efficiency. 2) Noisy-As-Clean with Consensus Equilibrium (NACE), a self-supervised 2D denoiser that is trained on a series of noisier-noisy image pairs and achieves SNR-matched denoising via consensus equilibrium. 3) Self-permutation Noise2Noise Denoiser (SPEND), a self-supervised 3D denoiser that requires only single noisy hyperspectral images for training and denoising.

• Lin H, Lee HJ, Tague N, Lugagne JB, Zong C, Deng F, Shin J, Tian L, Wong W, Dunlop MJ, Cheng JX. Microsecond fingerprint stimulated Raman spectroscopic imaging by ultrafast tuning and spatial-spectral learning. Nat Commun. 2021 May 24;12(1):3052.

• Lin H, Seitz S, Tan Y, Lugagne JB, Wang L, Ding G, He H, Rauwolf TJ, Dunlop MJ, Connor JH, Porco JA Jr, Tian L, Cheng JX. Label-free nanoscopy of cell metabolism by ultrasensitive reweighted visible stimulated Raman scattering. Nat Methods. 2025 May;22(5):1040-1050. 

• Ding G, Liu C, Yin J, Teng X, Tan Y, He H, Lin H# (Co-Corresponding), Tian L#, Cheng JX#. Self-Supervised Elimination of Non-Independent Noise in Hyperspectral Imaging. Newton. 2025 Aug 4;1(6). doi: 10.1016/j.newton.2025.100195.

Research Area 5. Novel Mechanisms for Chemical Imaging 

Chemically-selective photothermal imaging indirectly measures the local refractive index change due to vibrational absorption of biomolecules, which provides new avenues to expand the physical limits of chemical imaging. We have explored novel contrast mechanisms in vibrational photothermal imaging, including 1) stimulated Raman photothermal imaging that measures SRS-induced heating for enhanced sensitivity; 2) Overtone photothermal imaging that used visible light to probe the overtone absorption-induced heating for enhanced resolution:

• Wang L*, Lin H* (equal contribution), Zhu Y, Ge X, Li M, Liu J, Chen F, Zhang M, Cheng JX. Overtone photothermal microscopy for high-resolution and high-sensitivity vibrational imaging. Nat Commun. 2024 Jun 25;15(1):5374. 

• Zhu Y, Ge X, Ni H, Yin J, Lin H, Wang L, Tan Y, Prabhu Dessai CV, Li Y, Teng X, Cheng JX. Stimulated Raman photothermal microscopy toward ultrasensitive chemical imaging. Sci Adv. 2023 Oct 27;9(43):eadi2181.

Research Area 6. Biomedical Applications Enabled by CCI

With next-gen capabilities of CCI, we have a broad spectrum of biomedical applications:

• Synthetic biology (Nature Communications, 2021; Advanced Science, 2023; Advanced Science, 2022)

• Cancer metabolism (Nature Communications, 2021; Science Advances, 2023)

• Tissue histology (Nature Communications, 2021; Theranostics, 2024)

• Molecular virology (Nature Methods, 2025)