PET Imaging of Tumor Perfusion: A Potential Cancer Biomarker?

Technical and functional design considerations for a real-world interpretable AI solution for NIR perfusion analysis (including cancer)

Near infrared (NIR) analysis of tissue perfusion via indocyanine green fluorescence assessment is performed clinically during surgery for a range of indications. Its usefulness can potentially be further enhanced through the application of interpretable artificial intelligence (AI) methods to improve dynamic interpretation accuracy in these and also open new applications. While its main use currently is for perfusion assessment as a tissue health check prior to performing an anastomosis, there is increasing interest in using fluorophores for cancer detection during surgical interventions with most research being based on the paradigm of static imaging for fluorophore uptake hours after preoperative dosing. Although some image boosting and relative estimation of fluorescence signals is already inbuilt into commercial NIR systems, fuller implementation of AI methods can enable actionable predictions especially when applied during the dynamic, early inflow-outflow phase that occurs seconds to minutes after ICG (or indeed other fluorophore) administration. Already research has shown that such methods can accurately differentiate cancer from benign tissue in the operating theatre in real time in principle based on their differential signalling and could be useful for tissue perfusion classification more generally. This can be achieved through the generation of fluorescence intensity curves from an intra-operative NIR video stream. These curves are processed to adjust for image disturbances and curve features known to be influential in tissue characterisation are extracted. Existing machine learning based classifiers can then use these features to classify the tissue in question according to prior training sets. The use of this interpretable methodology enables accurate classification algorithms to be built with modest training sets in comparison to those required for deep learning modelling in addition to achieving compliance with medical device regulations. Integration of the multiple algorithms required to achieve this classification into a desktop application or medical device could make the use of this method accessible and useful to (as well as useable by) surgeons without prior training in computer technology. This document details some technical and functional design considerations underlying such a novel recommender system to advance the foundational concept and methodology as software as medical device for in situ cancer characterisation with relevance more broadly also to other tissue perfusion applications.

Whole-body parametric mapping of tumour perfusion in metastatic prostate cancer using long axial field-of-view [15O]H2O PET

PURPOSE

Tumour perfusion is a nutrient-agnostic biomarker for cancer metabolic rate. Use of tumour perfusion for cancer growth assessment has been limited by complicated image acquisition, image analysis and limited field-of-view scanners. Long axial field-of-view (LAFOV) PET scan using [15O]H2O, allows quantitative assessment of whole-body tumour perfusion. We created a tool for automated creation of quantitative parametric whole-body tumour perfusion images in metastatic cancer.

METHODS

Ten metastatic prostate cancer patients underwent dynamic LAFOV [15O]H2O PET (Siemens, Quadra) followed by [18F]PSMA-1007 PET. Perfusion was measured as [15O]H2O K1 (mL/min/mL) with a single-tissue compartment model and an automatically captured cardiac image-derived input function. Parametric perfusion images were automatically calculated using the basis-function method with initial voxel-wise delay estimation and a leading-edge approach. Subsequently, perfusion of volumes-of-interest (VOI) can be directly extracted from the parametric images. We used a [18F]PSMA-1007 SUV 4 fixed threshold for tumour delineation and transferred these VOIs to the perfusion map.

RESULTS

For 8 primary tumours, 64 lymph node metastases, and 85 bone metastases, median tumour perfusion were 0.19 (0.15-0.27) mL/min/mL, 0.16 (0.13-0.27) mL/min/mL, and 0.26 (0.21-0.39), respectively. The correlation between calculated perfusion from time-activity-curves and parametric images was excellent (r = 0.99, p < 0.0001).

CONCLUSION

LAFOV PET imaging using [15O]H2O enables truly quantitative parametric images of whole-body tumour perfusion, a potential biomarker for guiding personalized treatment and monitoring treatment response.

Assessment of Brain Tumour Perfusion Using Early-Phase 18F-FET PET: Comparison with Perfusion-Weighted MRI

PURPOSE

Morphological imaging using MRI is essential for brain tumour diagnostics. Dynamic susceptibility contrast (DSC) perfusion-weighted MRI (PWI), as well as amino acid PET, may provide additional information in ambiguous cases. Since PWI is often unavailable in patients referred for amino acid PET, we explored whether maps of relative cerebral blood volume (rCBV) in brain tumours can be extracted from the early phase of PET using O-(2-18F-fluoroethyl)-L-tyrosine (18F-FET).

PROCEDURE

Using a hybrid brain PET/MRI scanner, PWI and dynamic 18F-FET PET were performed in 33 patients with cerebral glioma and four patients with highly vascularized meningioma. The time interval from 0 to 2 min p.i. was selected to best reflect the blood pool phase in 18F-FET PET. For each patient, maps of MR-rCBV, early 18F-FET PET (0-2 min p.i.) and late 18F-FET PET (20-40 min p.i.) were generated and coregistered. Volumes of interest were placed on the tumour (VOI-TU) and normal-appearing brain (VOI-REF). The correlation between tumour-to-brain ratios (TBR) of the different parameters was analysed. In addition, three independent observers evaluated MR-rCBV and early 18F-FET maps (18F-FET-rCBV) for concordance in signal intensity, tumour extent and intratumoural distribution.

RESULTS

TBRs calculated from MR-rCBV and 18F-FET-rCBV showed a significant correlation (r = 0.89, p < 0.001), while there was no correlation between late 18F-FET PET and MR-rCBV (r = 0.24, p = 0.16) and 18F-FET-rCBV (r = 0.27, p = 0.11). Visual rating yielded widely agreeing findings or only minor differences between MR-rCBV maps and 18F-FET-rCBV maps in 93 % of the tumours (range of three independent raters 91-94%, kappa among raters 0.78-1.0).

CONCLUSION

Early 18F-FET maps (0-2 min p.i.) in gliomas provide similar information to MR-rCBV maps and may be helpful when PWI is not possible or available. Further studies in gliomas are needed to evaluate whether 18F-FET-rCBV provides the same clinical information as MR-rCBV.

Extracardiac findings with increased perfusion during clinical O-15-H2O PET/CT myocardial perfusion imaging: A case series

BACKGROUND

Coincidental extracardiac findings with increased perfusion were reported during myocardial perfusion imaging (MPI) with various retention radiotracers. Clinical parametric O-15-H2O PET MPI yielding quantitative measures of myocardial blood flow (MBF) was recently implemented at our facility. We aim to explore whether similar extracardiac findings are observed using O-15-H2O.

METHODS AND RESULTS

All patients (2963) were scanned with O-15-H2O PET MPI according to international guidelines and extracardiac findings were collected. In contrast to parametric O-15-H2O MBF images, extracardiac perfusion was assessed using summed images. Biopsy histopathology and other imaging modalities served as reference standards. Various malignant lesions with increased perfusion were detected, including lymphomas, large-celled neuroendocrine tumour, breast, and lung cancer plus metastases from colonic and renal cell carcinomas. Furthermore, inflammatory and hyperplastic benign conditions with increased perfusion were observed: rib fractures, gynecomastia, atelectasis, sarcoidosis, pneumonia, chronic lung inflammation and fibrosis, benign lung nodule, chronic diffuse lung infiltrates, pleural plaques and COVID-19 infiltrates.

CONCLUSIONS

Malignant and benign extracardiac coincidental findings with increased perfusion are readily visible and frequently seen on O-15-H2O PET MPI. We recommend evaluating the summed O-15-H2O PET images in addition to the low-dose CT attenuation images.

Novel computational biology modeling system can accurately forecast response to neoadjuvant therapy in early breast cancer

BACKGROUND

Generalizable population-based studies are unable to account for individual tumor heterogeneity that contributes to variability in a patient's response to physician-chosen therapy. Although molecular characterization of tumors has advanced precision medicine, in early-stage and locally advanced breast cancer patients, predicting a patient's response to neoadjuvant therapy (NAT) remains a gap in current clinical practice. Here, we perform a study in an independent cohort of early-stage and locally advanced breast cancer patients to forecast tumor response to NAT and assess the stability of a previously validated biophysical simulation platform.

METHODS

A single-blinded study was performed using a retrospective database from a single institution (9/2014-12/2020). Patients included: ≥ 18 years with breast cancer who completed NAT, with pre-treatment dynamic contrast enhanced magnetic resonance imaging. Demographics, chemotherapy, baseline (pre-treatment) MRI and pathologic data were input into the TumorScope Predict (TS) biophysical simulation platform to generate predictions. Primary outcomes included predictions of pathological complete response (pCR) versus residual disease (RD) and final volume for each tumor. For validation, post-NAT predicted pCR and tumor volumes were compared to actual pathological assessment and MRI-assessed volumes. Predicted pCR was pre-defined as residual tumor volume ≤ 0.01 cm³ (≥ 99.9% reduction).

RESULTS

The cohort consisted of eighty patients; 36 Caucasian and 40 African American. Most tumors were high-grade (54.4% grade 3) invasive ductal carcinomas (90.0%). Receptor subtypes included hormone receptor positive (HR+)/human epidermal growth factor receptor 2 positive (HER2+, 30%), HR+/HER2- (35%), HR-/HER2+ (12.5%) and triple negative breast cancer (TNBC, 22.5%). Simulated tumor volume was significantly correlated with post-treatment radiographic MRI calculated volumes (r = 0.53, p = 1.3 × 10^-7, mean absolute error of 6.57%). TS prediction of pCR compared favorably to pathological assessment (pCR: TS n = 28; Path n = 27; RD: TS n = 52; Path n = 53), for an overall accuracy of 91.2% (95% CI: 82.8% - 96.4%; Clopper-Pearson interval). Five-year risk of recurrence demonstrated similar prognostic performance between TS predictions (Hazard ratio (HR): - 1.99; 95% CI [- 3.96, - 0.02]; p = 0.043) and clinically assessed pCR (HR: - 1.76; 95% CI [- 3.75, 0.23]; p = 0.054).

CONCLUSION

We demonstrated TS ability to simulate and model tumor in vivo conditions in silico and forecast volume response to NAT across breast tumor subtypes.

Potential synergy between PSMA uptake and tumour blood flow for prediction of human prostate cancer aggressiveness

BACKGROUND

Both prostate-specific membrane antigen (PSMA) uptake and tumour blood flow (TBF) correlate with International Society of Urological Pathology (ISUP) Grade Group (GG) and hence prostate cancer (PCa) aggressiveness. The aim of the present study was to evaluate the potential synergistic benefit of combining the two physiologic parameters for separating significant PCa from insignificant findings.

METHODS

From previous studies of [82Rb]Rb positron emission tomography (PET) TBF in PCa, the 43 patients that underwent clinical [68Ga]Ga-PSMA-11 PET were selected for this retrospective study. Tumours were delineated on [68Ga]Ga-PSMA-11 PET or magnetic resonance imaging. ISUP GG was recorded from 52 lesions.

RESULTS

[68Ga]Ga-PSMA-11 maximum standardized uptake value (SUVmax) and [82Rb]Rb SUVmax correlated moderately with ISUP GG (rho = 0.59 and rho = 0.56, both p < 0.001) and with each other (r = 0.65, p < 0.001). A combined model of [68Ga]Ga-PSMA-11 and [82Rb]Rb SUVmax separated ISUP GG > 2 from ISUP GG 1-2 and benign with an area-under-the-curve of 0.85, 96% sensitivity, 74% specificity, and 95% negative predictive value. The combined model performed significantly better than either tracer alone did (p < 0.001), primarily by reducing false negatives from five or six to one (p ≤ 0.025).

CONCLUSION

PSMA uptake and TBF provide complementary information about tumour aggressiveness. We suggest that a combined analysis of PSMA uptake and TBF could significantly improve the negative predictive value and allow non-invasive separation of significant from insignificant PCa.