Predicting benefit from immune checkpoint inhibitors in patients with non-small-cell lung cancer by CT-based ensemble deep learning: a retrospective study.

BACKGROUND: Only around 20-30% of patients with non-small-cell lung cancer (NCSLC) have durable benefit from immune-checkpoint inhibitors. Although tissue-based biomarkers (eg, PD-L1) are limited by suboptimal performance, tissue availability, and tumour heterogeneity, radiographic images might holistically capture the underlying cancer biology. We aimed to investigate the application of deep learning on chest CT scans to derive an imaging signature of response to immune checkpoint inhibitors and evaluate its added value in the clinical context. METHODS: In this retrospective modelling study, 976 patients with metastatic, EGFR/ALK negative NSCLC treated with immune checkpoint inhibitors at MD Anderson and Stanford were enrolled from Jan 1, 2014, to Feb 29, 2020. We built and tested an ensemble deep learning model on pretreatment CTs (Deep-CT) to predict overall survival and progression-free survival after treatment with immune checkpoint inhibitors. We also evaluated the added predictive value of the Deep-CT model in the context of existing clinicopathological and radiological metrics. FINDINGS: Our Deep-CT model demonstrated robust stratification of patient survival of the MD Anderson testing set, which was validated in the external Stanford set. The performance of the Deep-CT model remained significant on subgroup analyses stratified by PD-L1, histology, age, sex, and race. In univariate analysis, Deep-CT outperformed the conventional risk factors, including histology, smoking status, and PD-L1 expression, and remained an independent predictor after multivariate adjustment. Integrating the Deep-CT model with conventional risk factors demonstrated significantly improved prediction performance, with overall survival C-index increases from 0·70 (clinical model) to 0·75 (composite model) during testing. On the other hand, the deep learning risk scores correlated with some radiomics features, but radiomics alone could not reach the performance level of deep learning, indicating that the deep learning model effectively captured additional imaging patterns beyond known radiomics features. INTERPRETATION: This proof-of-concept study shows that automated profiling of radiographic scans through deep learning can provide orthogonal information independent of existing clinicopathological biomarkers, bringing the goal of precision immunotherapy for patients with NSCLC closer. FUNDING: National Institutes of Health, Mark Foundation Damon Runyon Foundation Physician Scientist Award, MD Anderson Strategic Initiative Development Program, MD Anderson Lung Moon Shot Program, Andrea Mugnaini, and Edward L C Smith.

A model evaluation study for treatment planning of laser-induced thermal therapy.

A cross-validation analysis evaluating computer model prediction accuracy for a priori planning magnetic resonance-guided laser-induced thermal therapy (MRgLITT) procedures in treating focal diseased brain tissue is presented. Two mathematical models are considered. (1) A spectral element discretisation of the transient Pennes bioheat transfer equation is implemented to predict the laser-induced heating in perfused tissue. (2) A closed-form algorithm for predicting the steady-state heat transfer from a linear superposition of analytic point source heating functions is also considered. Prediction accuracy is retrospectively evaluated via leave-one-out cross-validation (LOOCV). Modelling predictions are quantitatively evaluated in terms of a Dice similarity coefficient (DSC) between the simulated thermal dose and thermal dose information contained within N = 22 MR thermometry datasets. During LOOCV analysis, the transient model's DSC mean and median are 0.7323 and 0.8001 respectively, with 15 of 22 DSC values exceeding the success criterion of DSC ≥ 0.7. The steady-state model's DSC mean and median are 0.6431 and 0.6770 respectively, with 10 of 22 passing. A one-sample, one-sided Wilcoxon signed-rank test indicates that the transient finite element method model achieves the prediction success criteria, DSC ≥ 0.7, at a statistically significant level.

(99m)Tc-labeled aminosilane-coated iron oxide nanoparticles for molecular imaging of ανß3-mediated tumor expression and feasibility for hyperthermia treatment.

HYPOTHESIS: Dual-modality imaging agents, such as radiolabeled iron oxide nanoparticles (IO-NPs), are promising candidates for cancer diagnosis and therapy. We developed and evaluated aminosilane coated Fe3O4 (10±2nm) as a tumor imaging agent in nuclear medicine through 3-aminopropyltriethoxysilane (APTES) functionalization. We evaluated this multimeric system of targeted (99m)Tc-labeled nanoparticles (NPs) conjugated with a new RGD derivate (cRGDfK-Orn3-CGG), characterized as NPs-RGD as a potential thermal therapy delivery vehicle. EXPERIMENTS: Transmission Electron Microscopy (TEM) and spectroscopy techniques were used to characterize the IO-NPs indicating their functionalization with peptides. Radiolabeled IO-NPs (targeted, non-targeted) were evaluated with regard to their radiochemical, radiobiological and imaging characteristics. In vivo studies were performed in normal and ανß3-positive tumor (U87MG glioblastoma) bearing mice. We also demonstrated that this system could reach ablative temperatures in vivo. FINDINGS: Both radiolabeled IO-NPs were obtained in high radiochemical yield (>98%) and proved stable in vitro. The in vivo studies for both IO-NPs have shown significant liver and spleen uptake at all examined time points in normal and U87MG glioblastoma tumor-bearing mice, due to their colloidal nature. We have confirmed through in vivo biodistribution studies that the non-targeted (99m)Tc-NPs poorly internalized in the tumor, while the targeted (99m)Tc-NPs-RGD, present 9-fold higher tumor accumulation at 1h p.i. Accumulation of both IO-NPs in other organs was negligible. Blocking experiments indicated target specificity for integrin receptors in U87MG glioblastoma cells. The preliminary in vivo study of applied alternating magnetic field showed that the induced hyperthermia is feasible due to the aid of IO-NPs.

Estimating nanoparticle optical absorption with magnetic resonance temperature imaging and bioheat transfer simulation.

PURPOSE: Optically activated nanoparticle-mediated heating for thermal therapy applications is an area of intense research. The ability to characterise the spatio-temporal heating potential of these particles for use in modelling under various exposure conditions can aid in the exploration of new approaches for therapy as well as more quantitative prospective approaches to treatment planning. The purpose of this research was to investigate an inverse solution to the heat equation using magnetic resonance temperature imaging (MRTI) feedback, for providing optical characterisation of two types of nanoparticles (gold-silica nanoshells and gold nanorods). METHODS: The optical absorption of homogeneous nanoparticle-agar mixtures was measured during exposure to an 808 nm laser using real-time MRTI. A coupled finite element solution of heat transfer was registered with the data and used to solve the inverse problem. The L2 norm of the difference between the temperature increase in the model and MRTI was minimised using a pattern search algorithm by varying the absorption coefficient of the mixture. RESULTS: Absorption fractions were within 10% of literature values for similar nanoparticles. Comparison of temporal and spatial profiles demonstrated good qualitative agreement between the model and the MRTI. The weighted root mean square error was <1.5 σMRTI and the average Dice similarity coefficient for ΔT = 5 °C isotherms was >0.9 over the measured time interval. CONCLUSION: This research demonstrates the feasibility of using an indirect method for making minimally invasive estimates of nanoparticle absorption that might be expanded to analyse a variety of geometries and particles of interest.

Multiparametric fat-water separation method for fast chemical-shift imaging guidance of thermal therapies.

PURPOSE: A k-means-based classification algorithm is investigated to assess suitability for rapidly separating and classifying fat/water spectral peaks from a fast chemical shift imaging technique for magnetic resonance temperature imaging. Algorithm testing is performed in simulated mathematical phantoms and agar gel phantoms containing mixed fat/water regions. METHODS: Proton resonance frequencies (PRFs), apparent spin-spin relaxation (T2*) times, and T1-weighted (T1-W) amplitude values were calculated for each voxel using a single-peak autoregressive moving average (ARMA) signal model. These parameters were then used as criteria for k-means sorting, with the results used to determine PRF ranges of each chemical species cluster for further classification. To detect the presence of secondary chemical species, spectral parameters were recalculated when needed using a two-peak ARMA signal model during the subsequent classification steps. Mathematical phantom simulations involved the modulation of signal-to-noise ratios (SNR), maximum PRF shift (MPS) values, analysis window sizes, and frequency expansion factor sizes in order to characterize the algorithm performance across a variety of conditions. In agar, images were collected on a 1.5T clinical MR scanner using acquisition parameters close to simulation, and algorithm performance was assessed by comparing classification results to manually segmented maps of the fat/water regions. RESULTS: Performance was characterized quantitatively using the Dice Similarity Coefficient (DSC), sensitivity, and specificity. The simulated mathematical phantom experiments demonstrated good fat/water separation depending on conditions, specifically high SNR, moderate MPS value, small analysis window size, and low but nonzero frequency expansion factor size. Physical phantom results demonstrated good identification for both water (0.997 ± 0.001, 0.999 ± 0.001, and 0.986 ± 0.001 for DSC, sensitivity, and specificity, respectively) and fat (0.763 ± 0.006, 0.980 ± 0.004, and 0.941 ± 0.002 for DSC, sensitivity, and specificity, respectively). Temperature uncertainties, based on PRF uncertainties from a 5 × 5-voxel ROI, were 0.342 and 0.351°C for pure and mixed fat/water regions, respectively. Algorithm speed was tested using 25 × 25-voxel and whole image ROIs containing both fat and water, resulting in average processing times per acquisition of 2.00 ± 0.07 s and 146 ± 1 s, respectively, using uncompiled MATLAB scripts running on a shared CPU server with eight Intel Xeon(TM) E5640 quad-core processors (2.66 GHz, 12 MB cache) and 12 GB RAM. CONCLUSIONS: Results from both the mathematical and physical phantom suggest the k-means-based classification algorithm could be useful for rapid, dynamic imaging in an ROI for thermal interventions. Successful separation of fat/water information would aid in reducing errors from the nontemperature sensitive fat PRF, as well as potentially facilitate using fat as an internal reference for PRF shift thermometry when appropriate. Additionally, the T1-W or R2* signals may be used for monitoring temperature in surrounding adipose tissue.

Generalised polynomial chaos-based uncertainty quantification for planning MRgLITT procedures.

PURPOSE: A generalised polynomial chaos (gPC) method is used to incorporate constitutive parameter uncertainties within the Pennes representation of bioheat transfer phenomena. The stochastic temperature predictions of the mathematical model are critically evaluated against MR thermometry data for planning MR-guided laser-induced thermal therapies (MRgLITT). METHODS: The Pennes bioheat transfer model coupled with a diffusion theory approximation of laser tissue interaction was implemented as the underlying deterministic kernel. A probabilistic sensitivity study was used to identify parameters that provide the most variance in temperature output. Confidence intervals of the temperature predictions are compared to MR temperature imaging (MRTI) obtained during phantom and in vivo canine (n = 4) MRgLITT experiments. The gPC predictions were quantitatively compared to MRTI data using probabilistic linear and temporal profiles as well as 2-D 60 °C isotherms. RESULTS: Optical parameters provided the highest variance in the model output (peak standard deviation: anisotropy 3.51 °C, absorption 2.94 °C, scattering 1.84 °C, conductivity 1.43 °C, and perfusion 0.94 °C). Further, within the statistical sense considered, a non-linear model of the temperature and damage-dependent perfusion, absorption, and scattering is captured within the confidence intervals of the linear gPC method. Multivariate stochastic model predictions using parameters with the dominant sensitivities show good agreement with experimental MRTI data. CONCLUSIONS: Given parameter uncertainties and mathematical modelling approximations of the Pennes bioheat model, the statistical framework demonstrates conservative estimates of the therapeutic heating and has potential for use as a computational prediction tool for thermal therapy planning.

Quantitative comparison of thermal dose models in normal canine brain.

PURPOSE: Minimally invasive thermal ablative therapies as alternatives to conventional surgical management of solid tumors and other pathologies is increasing owing to the potential benefits of performing these procedures in an outpatient setting with reduced complications and comorbidity. Magnetic resonance temperature imaging (MRTI) measurement allows existing thermal dose models to use the spatiotemporal temperature history to estimate the thermal damage to tissue. However, the various thermal dose models presented in the literature employ different parameters and thresholds, affecting the reliability of thermal dosimetry. In this study, the authors quantitatively compared three thermal dose models (Arrhenius rate process, CEM43, and threshold temperature) using the dice similarity coefficient (DSC). METHODS: The DSC was used to compare the spatial overlap between the region of thermal damage as predicted by the models for in vivo normal canine brain during thermal therapy to the region of thermal damage as revealed by contrast-enhanced T1-weighted images acquired immediately after therapy (< 20 min). The outer edge of the hyperintense rim of the ablation region was used as the surrogate marker for the limits of thermal coagulation. The DSC was also used to investigate the impact of varying the thresholds on each models' ability to predict the zone of thermal necrosis. RESULTS: At previously reported thresholds, the authors found that all three models showed good agreement (defined as DSC > 0.7) with post-treatment imaging. All three models examined across the range of commonly applied thresholds consistently showed highly accurate spatial overlap, low variability, and little dependence on temperature uncertainty. DSC values corresponding to cited thresholds were not significantly different from peak DSC values. CONCLUSIONS: Thus, the authors conclude that the all three thermal dose models can be used as a reliable surrogate for postcontrast tissue damage verification imaging in rapid ablation procedures and can also be used to enhance the capability of MRTI to control thermal therapy in real time.

Use of gold nanoshells to constrain and enhance laser thermal therapy of metastatic liver tumours.

PURPOSE: To investigate the impact of intravenously injected gold nanoparticles on interstitially delivered laser induced thermal therapy (LITT) in the liver. METHODS: 3D finite element modelling, ex vivo canine liver tissue containing gold nanoparticles absorbing at 800 nm, and agar gel phantoms were used to simulate the presence of nanoparticles in the liver during LITT. Real-time magnetic resonance temperature imaging (MRTI) based on the temperature sensitivity of the proton resonance frequency shift (PRFS) was used to map the spatiotemporal distribution of heating in the experiments and validate the predictions of 3D finite element simulations of heating. RESULTS: Experimental results show good agreement with both the simulation and the ex vivo experiments. Average discrepancy between simulation and experiment was shown to be 1.6 degrees C or less with the maximum difference being 3.8 degrees C due to a small offset in laser positioning. CONCLUSION: A high nanoshell concentration in the surrounding liver parenchyma, such as that which would be expected from an intravenous injection of gold nanoshells ( approximately 120 nm) acts as both a beam stop for the laser and secondary heat source for the treatment, helping to better heat the lesions and confine the treatment to the lesion. This indicates a potential to use nanoparticles to enhance both the safety and efficacy of LITT procedures in the liver.

Modulation of in vivo tumor radiation response via gold nanoshell-mediated vascular-focused hyperthermia: characterizing an integrated antihypoxic and localized vascular disrupting targeting strategy.

We report noninvasive modulation of in vivo tumor radiation response using gold nanoshells. Mild-temperature hyperthermia generated by near-infrared illumination of gold nanoshell-laden tumors, noninvasively quantified by magnetic resonance temperature imaging, causes an early increase in tumor perfusion that reduces the hypoxic fraction of tumors. A subsequent radiation dose induces vascular disruption with extensive tumor necrosis. Gold nanoshells sequestered in the perivascular space mediate these two tumor vasculature-focused effects to improve radiation response of tumors. This novel integrated antihypoxic and localized vascular disrupting therapy can potentially be combined with other conventional antitumor therapies.

Partially parallel imaging with phase-sensitive data: Increased temporal resolution for magnetic resonance temperature imaging.

Magnetic resonance temperature imaging can be used to monitor the progress of thermal ablation therapies, increasing treatment efficacy and improving patient safety. High temporal resolution is important when therapies rapidly heat tissue, but many approaches to faster image acquisition compromise image resolution, slice coverage, or phase sensitivity. Partially parallel imaging techniques offer the potential for improved temporal resolution without forcing such concessions. Although these techniques perturb image phase, relative phase changes between dynamically acquired phase-sensitive images, such as those acquired for MR temperature imaging, can be reliably measured through partially parallel imaging techniques using reconstruction filters that remain constant across the series. Partially parallel and non-accelerated phase-difference-sensitive data can be obtained through arrays of surface coils using this method. Average phase differences measured through partially parallel and fully Fourier encoded images are virtually identical, while phase noise increases with g(sqrt)L as in standard partially parallel image acquisitions..

Thermal therapy of canine cerebral tumors using a 980 nm diode laser with MR temperature-sensitive imaging feedback.

BACKGROUND AND PURPOSE: The laser-induced thermal therapy (LITT) of cerebral tumors has conventionally been performed using Nd:YAG lasers and is associated with a risk of high focal temperatures potentially followed by cavitation that could result in boiling and/or explosive char. We have developed small diffusing laser fiber tips to better distribute the energy deposition and a computer controlled feedback system to monitor therapy and prevent excess temperature buildup. In this study, we evaluated the feasibility of using magnetic resonance temperature imaging (MRTI)-based feedback system for the thermal treatment of experimental intracerebral tumors using 980 nm laser irradiation delivered through these diffusing tips. STUDY DESIGN/MATERIALS AND METHODS: Transmissible venereal tumors (TVTs) were grown via inoculation in the right cerebral hemisphere of seven canines. The laser fiber tips were inserted into a total of 10 independent TVT-suspected regions in the seven animals. Margins for the target area in each animal were prescribed on the basis of pretreatment MR images. MRTI-based feedback software was used to measure and regulate both temperature and the delivered thermal dose to achieve the desired thermal ablation and prevent excess heating. The effects of treatment were verified by results of histologic analyses. RESULTS: Treatments resulted in contiguous areas of thermal necrosis in tumors and adjacent brain margin. The feedback software successfully cut off the laser power once the desired treatment volume was achieved, and prevented focal temperatures from exceeding predefined thresholds. Follow-up MRI studies showed 1.4- to 2.9-fold LITT-induced lesion expansion within 1-6 days after treatment. CONCLUSIONS: Targeted thermal coagulation of small intracerebral tumors is feasible using MRTI-based feedback and diffused 980 nm diode laser light.

MR thermometry-based feedback control of laser interstitial thermal therapy at 980 nm.

BACKGROUND AND OBJECTIVES: The goal of this study was to explore the feasibility of magnetic resonance thermal imaging (MRTI)-based feedback control of intracerebral laser interstitial thermal therapy (LITT), using a computer workstation and 980-nm diode laser interfaced to an MR scanner. STUDY DESIGN/MATERIALS AND METHODS: A computer-controlled laser thermal therapy system was used to produce 12 ex vivo lesions in 3 canine and porcine brains and 16 in vivo lesions in 6 canines with diffusing tip fiberoptic applicators and energies from 54 to 900 J. MRTI predictions of thermal damage were correlated with histopathologic analysis. RESULTS: Under feedback control, no carbonization, vaporization, or applicator damage was observed. MRTI-based prediction of thermal dose was not significantly different from histological evaluation of achieved thermal necrosis. CONCLUSIONS: The computer-controlled thermal therapy system was effective at regulating heating, eliminating carbonization and vaporization, and protecting fiberoptic applicators. MRTI estimation of thermal dose accurately predicted achieved thermal necrosis.

Magnetic resonance imaging-guided focused ultrasound thermal therapy in experimental animal models: correlation of ablation volumes with pathology in rabbit muscle and VX2 tumors.

PURPOSE: To further investigate the use of magnetic resonance-guided focused ultrasound therapy (MRgFUS) as a noninvasive alternative to surgery in the local control of soft-tissue tumors by ablating prescribed volumes of VX2 rabbit tumors and comparing with ablation of normal tissue volumes. MATERIALS AND METHODS: Small, ellipsoidal ablations at shallow depth were created using 5- to 15-second sonication pulses at radio frequency (RF) powers of 50-125 W using a spherical, air-backed transducer operating at 1.463 MHz under MR guidance in a 1.5-T clinical scanner. RESULTS: Excellent correlation was observed between prescribed treatment volumes, MR thermal dosimetry, post-treatment verification MRI, and histopathology. Multifocal ablations of VX2 tumors in rabbits at depths of up to 2.5 cm resulted in complete ablation of the prescribed treatment volume. CONCLUSION: MRgFUS is an effective technique for treating tumors in vivo. Techniques developed for treatments in homogeneous tissue volumes are applicable in the more complicated tumor environment if MR temperature feedback is available to modify treatment delivery parameters.