A rapid dynamic in vivo near-infrared fluorescence imaging assay to track lung vascular permeability after acute radiation injury.

To develop a dynamic in vivo near-infrared (NIR) fluorescence imaging assay to quantify sequential changes in lung vascular permeability-surface area product (PS) in rodents. Dynamic NIR imaging methods for determining lung vascular permeability-surface area product were developed and tested on non-irradiated and 13 Gy irradiated rats with/without treatment with lisinopril, a radiation mitigator. A physiologically-based pharmacokinetic (PBPK) model of indocyanine green (ICG) pulmonary disposition was applied to in vivo imaging data and PS was estimated. In vivo results were validated by five accepted assays: ex vivo perfused lung imaging, endothelial filtration coefficient (Kf) measurement, pulmonary vascular resistance measurement, Evan's blue dye uptake, and histopathology. A PBPK model-derived measure of lung vascular permeability-surface area product increased from 2.60 ± 0.40 [CL: 2.42-2.78] mL/min in the non-irradiated group to 6.94 ± 8.25 [CL: 3.56-10.31] mL/min in 13 Gy group after 42 days. Lisinopril treatment lowered PS in the 13 Gy group to 4.76 ± 6.17 [CL: 2.12-7.40] mL/min. A much higher up to 5× change in PS values was observed in rats exhibiting severe radiation injury. Ex vivo Kf (mL/min/cm H2O/g dry lung weight), a measure of pulmonary vascular permeability, showed similar trends in lungs of irradiated rats (0.164 ± 0.081 [CL: 0.11-0.22]) as compared to non-irradiated controls (0.022 ± 0.003 [CL: 0.019-0.025]), with reduction to 0.070 ± 0.035 [CL: 0.045-0.096] for irradiated rats treated with lisinopril. Similar trends were observed for ex vivo pulmonary vascular resistance, Evan's blue uptake, and histopathology. Our results suggest that whole body dynamic NIR fluorescence imaging can replace current assays, which are all terminal. The imaging accurately tracks changes in PS and changes in lung interstitial transport in vivo in response to radiation injury.

Host genetic modifiers of nonproductive angiogenesis inhibit breast cancer.

PURPOSE: Multiple aspects of the tumor microenvironment (TME) impact breast cancer, yet the genetic modifiers of the TME are largely unknown, including those that modify tumor vascular formation and function. METHODS: To discover host TME modifiers, we developed a system called the Consomic/Congenic Xenograft Model (CXM). In CXM, human breast cancer cells are orthotopically implanted into genetically engineered consomic xenograft host strains that are derived from two parental strains with different susceptibilities to breast cancer. Because the genetic backgrounds of the xenograft host strains differ, whereas the inoculated tumor cells are the same, any phenotypic variation is due to TME-specific modifier(s) on the substituted chromosome (consomic) or subchromosomal region (congenic). Here, we assessed TME modifiers of growth, angiogenesis, and vascular function of tumors implanted in the SSIL2Rγ and SS.BN3IL2Rγ CXM strains. RESULTS: Breast cancer xenografts implanted in SS.BN3IL2Rγ (consomic) had significant tumor growth inhibition compared with SSIL2Rγ (parental control), despite a paradoxical increase in the density of blood vessels in the SS.BN3IL2Rγ tumors. We hypothesized that decreased growth of SS.BN3IL2Rγ tumors might be due to nonproductive angiogenesis. To test this possibility, SSIL2Rγ and SS.BN3IL2Rγ tumor vascular function was examined by dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), micro-computed tomography (micro-CT), and ex vivo analysis of primary blood endothelial cells, all of which revealed altered vascular function in SS.BN3IL2Rγ tumors compared with SSIL2Rγ. Gene expression analysis also showed a dysregulated vascular signaling network in SS.BN3IL2Rγ tumors, among which DLL4 was differentially expressed and co-localized to a host TME modifier locus (Chr3: 95-131 Mb) that was identified by congenic mapping. CONCLUSIONS: Collectively, these data suggest that host genetic modifier(s) on RNO3 induce nonproductive angiogenesis that inhibits tumor growth through the DLL4 pathway.

2nd Window NIR Imaging of Radiation Injury Mitigation Provided by Reduced Notch-Dll4 Expression on Vasculature.

PURPOSE: Vascular endothelium plays a central role in the pathogenesis of acute and chronic radiation injuries, yet the mechanisms which promote sustained endothelial dysfunction and contribute to late responding organ failure are unclear. We employed 2nd window (> 1100 nm emission) Near-Infrared (NIR) imaging using indocyanine green (ICG) to track and define the role of the notch ligand Delta-like ligand 4 (Dll4) in mediating vascular injury in two late-responding radiosensitive organs: the lung and kidney. PROCEDURES: Consomic strains of female Salt Sensitive or SS (Dll4-high) and SS with 3rd chromosome inherited from Brown Norway, SS.BN3 (Dll4-low) rats at ages 11-12 weeks were used to demonstrate the impact of reduced Dll4 expression on long-term vascular integrity, renal function, and survival following high-dose 13 Gy partial body irradiation at 42- and 90 days post-radiation. 2nd window dynamic NIR fluorescence imaging with ICG was analyzed with physiology-based pharmacokinetic modeling and confirmed with assays of endothelial Dll4 expression to assess the role of endogenous Dll4 expression on radiation injury protection. RESULTS: We show that SS.BN3 (Dll4-low) rats are relatively protected from vascular permeability disruption compared to the SS (Dll4-high) strain. We further demonstrated that SS.BN3 (Dll4-low) rats have reduced radiation induced loss of CD31+ vascular endothelial cells, and increased Dll4 vascular expression is correlated with vascular dysfunction. CONCLUSIONS: Together, these data suggest Dll4 plays a key role in pathogenesis of radiation-induced vascular injury to the lung and kidney.

Probing multifractality in tissue refractive index: prospects for precancer detection.

Multiresolution analysis on the spatial refractive index inhomogeneities in the epithelium and connective tissue regions of a human cervix reveals a clear signature of multifractality. Importantly, the derived multifractal parameters, namely, the generalized Hurst exponent and the width of the singularity spectrum, derived via multifractal detrended fluctuation analysis, shows interesting differences between tissues having different grades of precancers. The refractive-index fluctuations are found to be more anticorrelated, and the strength of multifractality is observed to be considerably stronger in the higher grades of precancers. These observations on the multifractal nature of tissue refractive-index variations may prove to be valuable for developing light-scattering approaches for noninvasive diagnosis of precancer and early-stage cancer.

Dynamic NIR Fluorescence Imaging and Machine Learning Framework for Stratifying High vs. Low Notch-Dll4 Expressing Host Microenvironment in Triple-Negative Breast Cancer.

Delta like canonical notch ligand 4 (Dll4) expression levels in tumors are known to affect the efficacy of cancer therapies. This study aimed to develop a model to predict Dll4 expression levels in tumors using dynamic enhanced near-infrared (NIR) imaging with indocyanine green (ICG). Two rat-based consomic xenograft (CXM) strains of breast cancer with different Dll4 expression levels and eight congenic xenograft strains were studied. Principal component analysis (PCA) was used to visualize and segment tumors, and modified PCA techniques identified and analyzed tumor and normal regions of interest (ROIs). The average NIR intensity for each ROI was calculated from pixel brightness at each time interval, yielding easily interpretable features including the slope of initial ICG uptake, time to peak perfusion, and rate of ICG intensity change after reaching half-maximum intensity. Machine learning algorithms were applied to select discriminative features for classification, and model performance was evaluated with a confusion matrix, receiver operating characteristic curve, and area under the curve. The selected machine learning methods accurately identified host Dll4 expression alterations with sensitivity and specificity above 90%. This may enable stratification of patients for Dll4 targeted therapies. NIR imaging with ICG can noninvasively assess Dll4 expression levels in tumors and aid in effective decision making for cancer therapy.

Topological data analysis in medical imaging: current state of the art.

Machine learning, and especially deep learning, is rapidly gaining acceptance and clinical usage in a wide range of image analysis applications and is regarded as providing high performance in detecting anatomical structures and identification and classification of patterns of disease in medical images. However, there are many roadblocks to the widespread implementation of machine learning in clinical image analysis, including differences in data capture leading to different measurements, high dimensionality of imaging and other medical data, and the black-box nature of machine learning, with a lack of insight into relevant features. Techniques such as radiomics have been used in traditional machine learning approaches to model the mathematical relationships between adjacent pixels in an image and provide an explainable framework for clinicians and researchers. Newer paradigms, such as topological data analysis (TDA), have recently been adopted to design and develop innovative image analysis schemes that go beyond the abilities of pixel-to-pixel comparisons. TDA can automatically construct filtrations of topological shapes of image texture through a technique known as persistent homology (PH); these features can then be fed into machine learning models that provide explainable outputs and can distinguish different image classes in a computationally more efficient way, when compared to other currently used methods. The aim of this review is to introduce PH and its variants and to review TDA's recent successes in medical imaging studies.

X-ray and MR Contrast Bearing Nanoparticles Enhance the Therapeutic Response of Image-Guided Radiation Therapy for Oral Cancer.

INTRODUCTION: Radiation therapy for head and neck squamous cell carcinoma is constrained by radiotoxicity to normal tissue. We demonstrate 100 nm theranostic nanoparticles for image-guided radiation therapy planning and enhancement in rat head and neck squamous cell carcinoma models. METHODS: PEG conjugated theranostic nanoparticles comprising of Au nanorods coated with Gadolinium oxide layers were tested for radiation therapy enhancement in 2D cultures of OSC-19-GFP-luc cells, and orthotopic tongue xenografts in male immunocompromised Salt sensitive or SS rats via both intratumoral and intravenous delivery. The radiation therapy enhancement mechanism was investigated. RESULTS: Theranostic nanoparticles demonstrated both X-ray/magnetic resonance contrast in a dose-dependent manner. Magnetic resonance images depicted optimal tumor-to-background uptake at 4 h post injection. Theranostic nanoparticle + Radiation treated rats experienced reduced tumor growth compared to controls, and reduction in lung metastasis. CONCLUSIONS: Theranostic nanoparticles enable preprocedure radiotherapy planning, as well as enhance radiation treatment efficacy for head and neck tumors.

Deep Learning Estimation of Median Nerve Volume Using Ultrasound Imaging in a Human Cadaver Model.

Median nerve swelling is one of the features of carpal tunnel syndrome (CTS), and ultrasound measurement of maximum median nerve cross-sectional area is commonly used to diagnose CTS. We hypothesized that volume might be a more sensitive measure than cross-sectional area for CTS diagnosis. We therefore assessed the accuracy and reliability of 3-D volume measurements of the median nerve in human cadavers, comparing direct measurements with ultrasound images interpreted using deep learning algorithms. Ultrasound images of a 10-cm segment of the median nerve were used to train the U-Net model, which achieved an average volume similarity of 0.89 and area under the curve of 0.90 from the threefold cross-validation. Correlation coefficients were calculated using the areas measured by each method. The intraclass correlation coefficient was 0.86. Pearson's correlation coefficient R between the estimated volume from the manually measured cross-sectional area and the estimated volume of deep learning was 0.85. In this study using deep learning to segment the median nerve longitudinally, estimated volume had high reliability. We plan to assess its clinical usefulness in future clinical studies. The volume of the median nerve may provide useful additional information on disease severity, beyond maximum cross-sectional area.

Automated measurement of total kidney volume from 3D ultrasound images of patients affected by polycystic kidney disease and comparison to MR measurements.

PURPOSE: Total kidney volume (TKV) is the most important imaging biomarker for quantifying the severity of autosomal-dominant polycystic kidney disease (ADPKD). 3D ultrasound (US) can accurately measure kidney volume compared to 2D US; however, manual segmentation is tedious and requires expert annotators. We investigated a deep learning-based approach for automated segmentation of TKV from 3D US in ADPKD patients. METHOD: We used axially acquired 3D US-kidney images in 22 ADPKD patients where each patient and each kidney were scanned three times, resulting in 132 scans that were manually segmented. We trained a convolutional neural network to segment the whole kidney and measure TKV. All patients were subsequently imaged with MRI for measurement comparison. RESULTS: Our method automatically segmented polycystic kidneys in 3D US images obtaining an average Dice coefficient of 0.80 on the test dataset. The kidney volume measurement compared with linear regression coefficient and bias from human tracing were R2 = 0.81, and - 4.42%, and between AI and reference standard were R2 = 0.93, and - 4.12%, respectively. MRI and US measured kidney volumes had R2 = 0.84 and a bias of 7.47%. CONCLUSION: This is the first study applying deep learning to 3D US in ADPKD. Our method shows promising performance for auto-segmentation of kidneys using 3D US to measure TKV, close to human tracing and MRI measurement. This imaging and analysis method may be useful in a number of settings, including pediatric imaging, clinical studies, and longitudinal tracking of patient disease progression.

Mitigating Bias in Radiology Machine Learning: 2. Model Development.

There are increasing concerns about the bias and fairness of artificial intelligence (AI) models as they are put into clinical practice. Among the steps for implementing machine learning tools into clinical workflow, model development is an important stage where different types of biases can occur. This report focuses on four aspects of model development where such bias may arise: data augmentation, model and loss function, optimizers, and transfer learning. This report emphasizes appropriate considerations and practices that can mitigate biases in radiology AI studies. Keywords: Model, Bias, Machine Learning, Deep Learning, Radiology © RSNA, 2022.

Mitigating Bias in Radiology Machine Learning: 3. Performance Metrics.

The increasing use of machine learning (ML) algorithms in clinical settings raises concerns about bias in ML models. Bias can arise at any step of ML creation, including data handling, model development, and performance evaluation. Potential biases in the ML model can be minimized by implementing these steps correctly. This report focuses on performance evaluation and discusses model fitness, as well as a set of performance evaluation toolboxes: namely, performance metrics, performance interpretation maps, and uncertainty quantification. By discussing the strengths and limitations of each toolbox, our report highlights strategies and considerations to mitigate and detect biases during performance evaluations of radiology artificial intelligence models. Keywords: Segmentation, Diagnosis, Convolutional Neural Network (CNN) © RSNA, 2022.

Algebraic topology-based machine learning using MRI predicts outcomes in primary sclerosing cholangitis.

BACKGROUND: Primary sclerosing cholangitis (PSC) is a chronic cholestatic liver disease that can lead to cirrhosis and hepatic decompensation. However, predicting future outcomes in patients with PSC is challenging. Our aim was to extract magnetic resonance imaging (MRI) features that predict the development of hepatic decompensation by applying algebraic topology-based machine learning (ML). METHODS: We conducted a retrospective multicenter study among adults with large duct PSC who underwent MRI. A topological data analysis-inspired nonlinear framework was used to predict the risk of hepatic decompensation, which was motivated by algebraic topology theory-based ML. The topological representations (persistence images) were employed as input for classification to predict who developed early hepatic decompensation within one year after their baseline MRI. RESULTS: We reviewed 590 patients; 298 were excluded due to poor image quality or inadequate liver coverage, leaving 292 potentially eligible subjects, of which 169 subjects were included in the study. We trained our model using contrast-enhanced delayed phase T1-weighted images on a single center derivation cohort consisting of 54 patients (hepatic decompensation, n = 21; no hepatic decompensation, n = 33) and a multicenter independent validation cohort of 115 individuals (hepatic decompensation, n = 31; no hepatic decompensation, n = 84). When our model was applied in the independent validation cohort, it remained predictive of early hepatic decompensation (area under the receiver operating characteristic curve = 0.84). CONCLUSIONS: Algebraic topology-based ML is a methodological approach that can predict outcomes in patients with PSC and has the potential for application in other chronic liver diseases.

Differing self-similarity in light scattering spectra: a potential tool for pre-cancer detection.

The fluctuations in the elastic light scattering spectra of normal and dysplastic human cervical tissues analyzed through wavelet transform based techniques reveal clear signatures of self-similar behavior in the spectral fluctuations. The values of the scaling exponent observed for these tissues indicate the differences in the self-similarity for dysplastic tissues and their normal counterparts. The strong dependence of the elastic light scattering on the size distribution of the scatterers manifests in the angular variation of the scaling exponent. Interestingly, the spectral fluctuations in both these tissues showed multi-fractality (non-stationarity in fluctuations), the degree of multi-fractality being marginally higher in the case of dysplastic tissues. These findings using the multi-resolution analysis capability of the discrete wavelet transform can contribute to the recent surge in the exploration for non-invasive optical tools for pre-cancer detection.

NIR-II window tracking of hyperglycemia induced intracerebral hemorrhage in cerebral cavernous malformation deficient mice.

Second near infrared (NIR-II) window fluorescence imaging between 1000 and 1700 nm with reduced scattering and autofluorescence and deep tissue light penetration allows early and non-invasive determination of vascular pathologies. Here, we demonstrate in vivo NIR-II imaging techniques for tracking hyperglycaemia-induced Intracerebral Hemorrhage (ICH) and Blood Brain Barrier (BBB) hyperpermeability in Cerebral Cavernous Malformation (CCM) deficient mice (CCM1+/-). We synthesised PEGylated Ag2S quantum dots (QDs) with a bright fluorescent emission peak centred at 1135 nm under an 808 nm NIR light for dynamic imaging of cerebral vasculature in mice and determined the development of ICH and BBB impairment in hyperglycaemic CCM1+/- mice. In vivo optical imaging was conducted with micro-CT (including k-mean cluster analysis) as well as in vivo permeability assays using FITC-dextran perfusion and IgG staining, respectively. The increased BBB permeability in CCM1+/- mice was further demonstrated to be associated with a high-glucose-caused decrease of CCM1 expressions. This study validates that deep-penetrating NIR-II QDs can be used for the tracking of ICH and BBB hyperpermeability in transgenic mice models of cerebral vascular anomalies.

In vivo longitudinal imaging of RNA interference-induced endocrine therapy resistance in breast cancer.

Endocrine therapy resistance in breast cancer is a major obstacle in the treatment of patients with estrogen receptor-positive (ER+) tumors. Herein, we demonstrate the feasibility of longitudinal, noninvasive and semiquantitative in vivo molecular imaging of resistance to three endocrine therapies by using an inducible fluorescence-labeled short hairpin RNA (shRNA) system in orthotopic mice xenograft tumors. We employed a dual fluorescent doxycycline (Dox)-regulated lentiviral inducer system to transfect ER+ MCF7L breast cancer cells, with green fluorescent protein (GFP) expression as a marker of transfection and red fluorescent protein (RFP) expression as a surrogate marker of Dox-induced tumor suppressor phosphatase and tensin homolog deleted on chromosome 10 (PTEN) knockdown. Xenografted MCF7L tumor-bearing nude mice were randomized to therapies comprising estrogen deprivation, tamoxifen or an ER degrader (fulvestrant) and an estrogen-treated control group. Longitudinal imaging was performed by a home-built multispectral imaging system based on a cooled image intensified charge coupled device camera. The GFP signal, which corresponds to number of viable tumor cells, exhibited excellent correlation to caliper-measured tumor size (P << .05). RFP expression was substantially higher in mice exhibiting therapy resistance and strongly and significantly (P < 1e-7) correlated with the tumor size progression for the mice with shRNA-induced PTEN knockdown. PTEN loss was strongly correlated with resistance to estrogen deprivation, tamoxifen and fulvestrant therapies.

Localized and triggered release of oxaliplatin for the treatment of colorectal liver metastasis.

Purpose: The aim of this study was to develop and evaluate a liposome formulation that deliver oxaliplatin under magnetic field stimulus in high concentration to alleviate the off-target effects in a rat model of colorectal liver metastases (CRLM). Materials and Methods: Hybrid liposome-magnetic nanoparticles loaded with Cy5.5 dye and oxaliplatin (L-NIR- Fe3O4/OX) were synthesized by using thermal decomposition method. CRLM (CC-531) cell viability was assessed and rats orthotopically implanted with CC-531 cells were treated with L-NIR-Fe3O4/OX or by drug alone via different routes, up to 3 cycles of alternating magnetic field (AMF). Optical and MR imaging was performed to assess the targeted delivery. Biodistribution and histology was performed to determine the distribution of oxaliplatin. Results: L-NIR-Fe3O4/OX presented a significant increase of oxaliplatin release (~18%) and lower cell viability after AMF exposure (p<0.001). Optical imaging showed a significant release of oxaliplatin among mesenteric vein injected (MV) group of animals. MR imaging on MV injected animals showed R2* changes in the tumor regions at the same regions immediately after infusion compared to the surrounding liver (p<0.001). Biodistribution analysis showed significantly higher levels of oxaliplatin in liver tissues compared to lungs (p<0.001) and intestines (p<0.001) in the MV animals that received AMF after L-NIR- Fe3O4/OX administration. Large tumor necrotic zones and significant improvement in the survival rates were noted in the MV animals treated with AMF. Conclusion: AMF triggers site selective delivery of oxaliplatin at high concentrations and improves survival outcomes in colorectal liver metastasis tumor bearing rats.

Heritable modifiers of the tumor microenvironment influence nanoparticle uptake, distribution and response to photothermal therapy.

We report the impact of notch-DLL4-based hereditary vascular heterogeneities on the enhanced permeation and retention (EPR) effect and plasmonic photothermal therapy response in tumors. Methods: We generated two consomic rat strains with differing DLL4 expression on 3rd chromosome. These strains were based on immunocompromised Salt-sensitive or SSIL2Rγ- (DLL4-high) and SS.BN3IL2Rγ- (DLL4-low) rats with 3rd chromosome substituted from Brown Norway rat. We further constructed three novel SS.BN3IL2Rγ- congenic strains by introgressing varying segments of BN chromosome 3 into the parental SSIL2Rγ- strain to localize the role of SSIL2Rγ- DLL4 on tumor EPR effect with precision. We synthesized multimodal theranostic nanoparticles (TNPs) based on Au-nanorods which provide magnetic resonance imaging (MRI), X-ray, and optical contrasts to assess image guided PTT response and quantify host specific therapy response differences in tumors orthotopically xenografted in DLL4-high and -low strains. We tested recovery of therapy sensitivity of PTT resistant strains by employing anti-DLL4 conjugated TNPs in two triple negative breast cancer tumor xenografts. Results: Host strains with high DLL4 allele demonstrated slightly increased tumor nanoparticle uptake but consistently developed photothermal therapy resistance compared to tumors in host strains with low DLL4 allele. Tumor micro-environment with low DLL4 expression altered the geographic distribution of nanoparticles towards closer proximity with vasculature which improved efficacy of PTT in spite of lower overall TNP uptake. Targeting TNPs to tumor endothelium via anti-DLL4 antibody conjugation improved therapy sensitivity in high DLL4 allele hosts for two triple negative human breast cancer xenografts. Conclusions: Inherited DLL4 expression modulates EPR effects in tumors, and molecular targeting of endothelial DLL4 via nanoparticles is an effective personalized nanomedicine strategy.

Erratum: Methods for detecting host genetic modifiers of tumor vascular function using dynamic near-infrared fluorescence imaging: errata.

[This corrects the article on p. 543 in vol. 9, PMID: 29552392.].

Methods for detecting host genetic modifiers of tumor vascular function using dynamic near-infrared fluorescence imaging.

Vascular supply is a critical component of the tumor microenvironment (TME) and is essential for tumor growth and metastasis, yet the endogenous genetic modifiers that impact vascular function in the TME are largely unknown. To identify the host TME modifiers of tumor vascular function, we combined a novel genetic mapping strategy [Consomic Xenograft Model] with near-infrared (NIR) fluorescence imaging and multiparametric analysis of pharmacokinetic modeling. To detect vascular flow, an intensified cooled camera based dynamic NIR imaging system with 785 nm laser diode based excitation was used to image the whole-body fluorescence emission of intravenously injected indocyanine green dye. Principal component analysis was used to extract the spatial segmentation information for the lungs, liver, and tumor regions-of-interest. Vascular function was then quantified by pK modeling of the imaging data, which revealed significantly altered tissue perfusion and vascular permeability that were caused by host genetic modifiers in the TME. Collectively, these data demonstrate that NIR fluorescent imaging can be used as a non-invasive means for characterizing host TME modifiers of vascular function that have been linked with tumor risk, progression, and response to therapy.

Vascular Interventional Radiology-Guided Photothermal Therapy of Colorectal Cancer Liver Metastasis with Theranostic Gold Nanorods.

We report sub-100 nm optical/magnetic resonance (MR)/X-ray contrast-bearing theranostic nanoparticles (TNPs) for interventional image-guided photothermal therapy (PTT) of solid tumors. TNPs were composed of Au@Gd2O3:Ln (Ln = Yb/Er) with X-ray contrast (∼486 HU; 1014 NPs/mL, 0.167 nM) and MR contrast (∼1.1 × 108 mM-1 S-1 at 9.4 T field strength). Although TNPs are deposited in tumors following systemic administration via enhanced permeation and retention effect, the delivered dose to tumors is typically low; this can adversely impact the efficacy of PTT. To overcome this limitation, we investigated the feasibility of site-selective hepatic image-guided delivery of TNPs in rats bearing colorectal liver metastasis (CRLM). The mesenteric vein of tumor-bearing rats was catheterized, and TNPs were infused into the liver by accessing the portal vein for site-selective delivery. The uptake of TNPs with hepatic delivery was compared with systemic administration. MR imaging confirmed that delivery via the hepatic portal vein can double the CRLM tumor-to-liver contrast compared with systemic administration. Photothermal ablation was performed by inserting a 100 µm fiber-optic carrying 808 nm light via a JB1, 3-French catheter for 3 min under DynaCT image guidance. Histological analysis revealed that the thermal damage was largely confined to the tumor region with minimal damage to the adjacent liver tissue. Transmission electron microscopy imaging validated the stability of core-shell structure of TNPs in vivo pre- and post-PTT. TNPs comprising Gd-shell-coated Au nanorods can be effectively employed for the site-directed PTT of CRLM by leveraging interventional radiology methods.