Deep learning based identification of bone scintigraphies containing metastatic bone disease foci

PURPOSE

Metastatic bone disease (MBD) is the most common form of metastases, most frequently deriving from prostate cancer. MBD is screened with bone scintigraphy (BS), which have high sensitivity but low specificity for the diagnosis of MBD, often requiring further investigations. Deep learning (DL) - a machine learning technique designed to mimic human neuronal interactions- has shown promise in the field of medical imaging analysis for different purposes, including segmentation and classification of lesions. In this study, we aim to develop a DL algorithm that can classify areas of increased uptake on bone scintigraphy scans.

METHODS

We collected 2365 BS from three European medical centres. The model was trained and validated on 1203 and 164 BS scans respectively. Furthermore we evaluated its performance on an external testing set composed of 998 BS scans. We further aimed to enhance the explainability of our developed algorithm, using activation maps. We compared the performance of our algorithm to that of 6 nuclear medicine physicians.

RESULTS

The developed DL based algorithm is able to detect MBD on BSs, with high specificity and sensitivity (0.80 and 0.82 respectively on the external test set), in a shorter time compared to the nuclear medicine physicians (2.5 min for AI and 30 min for nuclear medicine physicians to classify 134 BSs). Further prospective validation is required before the algorithm can be used in the clinic.

Deconvolution of Tc-99m-Mercaptoacetyltriglycine Renograms with the Concomitant Use of a Sparse Legendre Polynomial Representation and the Moore-Penrose Pseudo-inverse

Objectives:

This study aimed to introduce an improved deconvolution technique for Tc-99m-mercaptoacetyltriglycine renograms based on the combination of a sparse Legendre polynomial representation and the Moore-Penrose inversion matrix (LG). This method reduces the effect of noise on the measurement of renal retention function transit time (TT).

Methods:

The stability and accuracy of the proposed method were tested using a renal database containing Monte Carlo-simulated studies and real adult patient data. Two clinical parameters, namely, split function (SF) and mean TT (meanTT), obtained with LG were compared with values calculated with the established method that combines matrix deconvolution and a three-point linear smoothing (F121) as recommended by the 2008 International Scientific Committee of Radionuclides in Nephrourology consensus on renal TT measurements.

Results:

For simulated data, the root mean square error (RMSE) between the theoretical non-noisy renal retention curve (RRC) and the results of the deconvolution methods applied to the noisy RRC were up to two times lower with LG (p<0.001). The RMSE of the reconvoluted renogram and the theoretical one was also lower for LG (p<0.001) and showed better preservation of the original signal. The SF was neither improved nor degraded by the proposed method. For patient data, no statistically significant difference was found between the SF for the LG method compared with the database values, and the meanTT better agreed with the physician’s diagnosis than the matrix or clinical software (Hermes) outputs. A visual improvement of the RRC was also observed.

Conclusion:

By combining the sparse Legendre representation of the renogram curves and the Moore-Penrose matrix inverse techniques, we obtained improved noise reduction in the deconvoluted data, leading to better elimination of non-physiological signals -as negative values- and the avoidance of the smear effect of conventional smoothing on the vascular peak, which both influenced the meanTT measurement.

Correlation analysis of [18F]fluorodeoxyglucose and [18F]fluoroazomycin arabinoside uptake distributions in lung tumours during radiation therapy

BACKGROUND

PET-guided dose painting (DP) aims to target radioresistant tumour regions in order to improve radiotherapy (RT) outcome. Besides the well-known [¹⁸F]fluorodeoxyglucose (FDG), the hypoxia positron emission tomography (PET) tracer [¹⁸F]fluoroazomycin arabinoside (FAZA) could provide further useful information to guide the radiation dose prescription. In this study, we compare the spatial distributions of FDG and FAZA PET uptakes in lung tumours.

MATERIAL AND METHODS

Fourteen patients with unresectable lung cancer underwent FDG and FAZA 4D-PET/CT on consecutive days at three time-points: prior to RT (pre), and during the second (w2), and the third (w3) weeks of RT. All PET/CT were reconstructed in their time-averaged midposition (MidP). The metabolic tumour volume (MTV: FDG standardised uptake value (SUV) > 50% SUV_max), and the hypoxic volume (HV: FAZA SUV > 1.4) were delineated within the gross tumour volume (GTV_CT). FDG and FAZA intratumoral PET uptake distributions were subsequently pairwise compared, using both volume-, and voxel-based analyses.

RESULTS

Volume-based analysis showed large overlap between MTV and HV: median overlapping fraction was 0.90, 0.94 and 0.94, at the pre, w2 and w3 time-points, respectively. Voxel-wise analysis between FDG and FAZA intratumoral PET uptake distributions showed high correlation: median Spearman's rank correlation coefficient was 0.76, 0.77 and 0.76, at the pre, w2 and w3 time-points, respectively. Interestingly, tumours with high FAZA uptake tended to show more similarity between FDG and FAZA intratumoral uptake distributions than those with low FAZA uptake.

CONCLUSIONS

In unresectable lung carcinomas, FDG and FAZA PET uptake distributions displayed unexpectedly strong similarity, despite the distinct pathways targeted by these tracers. Hypoxia PET with FAZA brought very little added value over FDG from the perspective of DP in this population.

Evolution of [18F]fluorodeoxyglucose and [18F]fluoroazomycin arabinoside PET uptake distributions in lung tumours during radiation therapy

BACKGROUND

Dose painting (DP) aims to improve radiation therapy (RT) outcome by targeting radioresistant tumour regions identified through functional imaging, e.g., positron emission tomography (PET). Importantly, the expected benefit of DP relies on the ability of PET imaging to identify tumour areas which could be consistently targeted throughout the treatment. In this study, we analysed the spatial stability of two potential DP targets in lung cancer patients undergoing RT: the tumour burden surrogate [¹⁸F]fluorodeoxyglucose (FDG) and the hypoxia surrogate [¹⁸F]fluoroazomycin arabinoside (FAZA).

MATERIALS AND METHODS

Thirteen patients with unresectable lung tumours underwent FDG and FAZA 4D-PET/CT before (pre), and during the second (w2) and third (w3) weeks of RT. All PET/CT were reconstructed in their time-averaged midposition (MidP) for further analysis. The metabolic tumour volume (MTV: FDG standardised uptake value (SUV) > 50% SUV_max) and the hypoxic volume (HV: FAZA SUV >1.4) were delineated within the gross tumour volume (GTV_CT). The stability of FDG and FAZA PET uptake distributions during RT was subsequently assessed through volume-overlap analysis and voxel-based correlation analysis.

RESULTS

The volume-overlap analysis yielded median overlapping fraction (OF) of 0.86 between MTV_pre and MTV_w2 and 0.82 between MTV_pre and MTV_w3. In patients with a detectable HV, median OF was 0.82 between HV_pre and HV_w2 and 0.90 between HV_pre and HV_w3. The voxel-based correlation analysis yielded median Spearman's correlation coefficient (r_S) of 0.87 between FDG_pre and FDG_w2 and 0.83 between FDG_pre and FDG_w3. Median r_S was 0.78 between FAZA_pre and FAZA_w2 and 0.79 between FAZA_pre and FAZA_w3.

CONCLUSIONS

FDG and FAZA PET uptake distributions were spatially stable during the 3 first weeks of RT in patients with unresectable lung cancer, both based on volume- and voxel-based indicators. This might allow for a consistent targeting of high FDG or FAZA PET uptake regions as part of a DP strategy.

αVβ3 integrin-targeted microSPECT/CT imaging of inflamed atherosclerotic plaques in mice

Background

α_Vβ₃-integrin is expressed by activated endothelial cells and macrophages in atherosclerotic plaques and may represent a valuable marker of high-risk plaques. We evaluated ^99mTc-maraciclatide, an integrin-specific tracer, for imaging vascular inflammation in atherosclerotic lesions in mice.

Methods

Apolipoprotein E-negative (ApoE^−/−) mice on a Western diet (n = 10) and normally fed adult C57BL/6 control mice (n = 4) were injected with ^99mTc-maraciclatide (51.8 ± 3.7 MBq). A blocking peptide was infused in three ApoE^−/− mice; this condition served as another control. After 90 min, the animals were imaged via single-photon emission computed tomography (SPECT). While maintained in the same position, the mice were transferred to computed tomography (CT) to obtain contrast-enhanced images of the aortic arch. Images from both modalities were fused, and signal was quantified in the aortic arch and in the vena cava for subtraction of blood-pool activity. The aorta was carefully dissected after imaging for gamma counting, autoradiography, and histology.

Results

Tracer uptake was significantly higher in ApoE^−/− mice than in both groups of control mice (1.56 ± 0.33 vs. 0.82 ± 0.24 vs. 0.98 ± 0.11, respectively; P = 0.006). Furthermore, higher tracer activity was detected via gamma counting in the aorta of hypercholesterolemic mice than in both groups of control mice (1.52 ± 0.43 vs. 0.78 ± 0.19 vs. 0.47 ± 0.31 ^99mTc-maraciclatide %ID/g, respectively; P = 0.018). Autoradiography showed significantly higher tracer uptake in the atherosclerotic aorta than in the control aorta (P = 0.026). Finally, in the atherosclerotic aorta, immunostaining indicated that the integrin signal came predominantly from macrophages and was correlated with the macrophage CD68 immunomarker (r = 0.73).

Conclusions

^99mTc-maraciclatide allows in vivo detection of inflamed atherosclerotic plaques in mice and may represent a non-invasive approach for identifying high-risk plaques in patients.

Methodology for adaptive and robust FDG-PET escalated dose painting by numbers in head and neck tumors

OBJECTIVE

To develop a methodology for using FDG PET/CT in adaptive dose painting by numbers (DPBN) in head and neck squamous cell carcinoma (HNSCC) patients. Issues related to noise in PET and treatment robustness against geometric errors are addressed.

METHODS

Five patients with locally advanced HNSCC scheduled for chemo-radiotherapy were imaged with FDG-PET/CT at baseline and 2-3 times during radiotherapy (RT). The GTVPET was segmented with a gradient-based method. A double median filter reduces the impact of noise in the PET uptake-to-dose conversion. Filtered FDG uptake values were linearly converted into a voxel-by-voxel prescription from 70 (median uptake) to 86 Gy (highest uptake). A PTVPET was obtained by applying a dilation of 2.5 mm to the entire prescription. Seven iso-uptake thresholds led to seven sub-levels compatible with the Tomotherapy HiArt(®) Treatment Planning System. Planning aimed to deliver a median dose of 56 Gy and 70 Gy in 35 fractions on the elective and therapeutic PTVs, respectively. Plan quality was assessed with quality volume histogram (QVH). At each time point, plans were generated with a total of 3-4 plans for each patient. Deformable image registration was used for automatic contour propagation and dose summation of the 3 or 4 treatment plans (MIMvista(®)).

RESULTS

GTVPET segmentations were performed successfully until week 2 of RT but failed in two patients at week 3. QVH analysis showed high conformity for all plans (mean VQ = 0.95 93%; mean VQ = 1.05 3.9%; mean QF 2.2%). Good OAR sparing was achieved while keeping high plan quality.

CONCLUSION

Our results show that adaptive FDG-PET-based escalated dose painting in patients with locally advanced HNSCC is feasible while respecting strict dose constraints to organs at risk. Clinical studies must be conducted to evaluate toxicities and tumor response of such a strategy.

PET imaging biomarkers in head and neck cancer

In locally advanced head and neck squamous cell carcinoma (HNSCC), the role of imaging becomes more and more critical in the management process. In this framework, molecular imaging techniques such as PET allow noninvasive assessment of a range of tumour biomarkers such as metabolism, hypoxia and proliferation, which can serve different purposes. First, in a pretreatment setting they can influence therapy selection strategies and target delineation for radiation therapy. Second, their predictive and/or prognostic value could help enhance the therapeutic ratio in the management of HNSCC. Third, treatment modification can be performed through the generation of a molecular-based heterogeneous dose distribution with dose escalation to the most resistant parts of the tumour, a concept known as dose painting. Fourth, they are increasingly becoming a tool for monitoring response to therapy. In this review, PET imaging biomarkers used in the routine management of HNSCC or under investigation are discussed.

A randomized trial on the optimization of 18F-FDG myocardial uptake suppression: implications for vulnerable coronary plaque imaging

(18)F-FDG PET/CT can be used to detect arterial atherosclerotic plaque inflammation. However, avid myocardial glucose uptake may preclude its use for visualizing coronary plaques. Fatty acid loading or calcium channel blockers could decrease myocardial (18)F-FDG uptake, thus assisting coronary plaque inflammation identification. The present prospective randomized trial compared the efficacies of different interventions for suppressing myocardial (18)F-FDG uptake. We also investigated whether circulating free fatty acid (cFFA) levels predicted the magnitude of myocardial (18)F-FDG uptake.

METHODS

Thirty-six volunteers ate a high-fat low-carbohydrate meal, followed by a 12-h fasting period. They were then randomized to 1 of 4 intervention groups. Group 1 received no additional preparation and served as a reference. Groups 2 and 3, respectively, received a commercial high-fat solution containing 43.8 g of lipids or 50 mL of olive oil 1 h before (18)F-FDG injection to evaluate the impact of fatty acid loading on myocardial (18)F-FDG uptake. Group 4 received verapamil to evaluate the effect of calcium channel blockers. Cardiac PET/CT was performed after administration of 370 MBq of (18)F-FDG. Myocardial uptake suppression was assessed using a qualitative visual scale and by measuring the myocardial maximum standardized uptake value (SUV(max)). Insulin, glucose, and cFFA were serially measured.

RESULTS

The qualitative visual scale showed good myocardial (18)F-FDG uptake suppression in 8 of 9, 5 of 9, 4 of 9, and 8 of 9 subjects of groups 1, 2, 3, and 4, respectively (P = 0.09). SUV(max) did not significantly differ between groups (P = 0.17). Interestingly, cFFA levels were higher in volunteers with good suppression (0.80 ± 0.31 mmol/L) than in those with poor suppression (0.53 ± 0.15 mmol/L; P = 0.011). We found an inverse correlation between cFFA level (measured at (18)F-FDG injection) and the SUV(max) (R = 0.61). Receiver-operating-characteristic curve analysis identified 0.65 mmol/L cFFA as the best cutoff value to predict adequate (18)F-FDG uptake suppression (positive predictive value, 89%).

CONCLUSION

A high-fat low-carbohydrate meal followed by a 12-h fasting period effectively suppressed myocardial (18)F-FDG uptake in most subjects. Neither complementary fatty acid loading nor verapamil administered 1 h before (18)F-FDG injection conferred any additional benefit. Myocardial (18)F-FDG uptake was inversely correlated with cFFA level, representing an interesting way to predict myocardial (18)F-FDG uptake suppression.

Advances in the management of squamous cell carcinoma of the head and neck

Squamous cell carcinoma of the head and neck (SCCHN) is the sixth most common cancer worldwide. The main risk factors for cancers of the oral cavity, larynx, oropharynx, and hypopharynx are alcohol and tobacco use. In addition, the human papillomavirus (HPV) is an established cause of oropharyngeal cancer. An experienced multidisciplinary team is necessary for adequate management and optimal outcome. The treatment of locally advanced disease generally requires various combinations of radiotherapy, surgery, and systemic therapy, but despite this aggressive multimodal treatment, 40% to 60% of the patients will relapse. In this report, we will discuss recent advances in the management of SCCHN, including new developments in molecular biology, imaging, and treatment.

Tumour control probability derived from dose distribution in homogeneous and heterogeneous models: assuming similar pharmacokinetics, (125)Sn-(177)Lu is superior to (90)Y-(177)Lu in peptide receptor radiotherapy

Clinical trials on (177)Lu-(90)Y therapy used empirical activity ratios. Radionuclides (RN) with larger beta maximal range could favourably replace (90)Y. Our aim is to provide RN dose-deposition kernels and to compare the tumour control probability (TCP) of RN combinations. Dose kernels were derived by integration of the mono-energetic beta-ray dose distributions (computed using Monte Carlo) weighted by their respective beta spectrum. Nine homogeneous spherical tumours (1-25 mm in diameter) and four spherical tumours including a lattice of cold, but alive, spheres (1, 3, 5, 7 mm in diameter) were modelled. The TCP for (93)Y, (90)Y and (125)Sn in combination with (177)Lu in variable proportions (that kept constant the renal cortex biological effective dose) were derived by 3D dose kernel convolution. For a mean tumour-absorbed dose of 180 Gy, 2 mm homogeneous tumours and tumours including 3 mm diameter cold alive spheres were both well controlled (TCP > 0.9) using a 75-25% combination of (177)Lu and (90)Y activity. However, (125)Sn-(177)Lu achieved a significantly better result by controlling 1 mm-homogeneous tumour simultaneously with tumours including 5 mm diameter cold alive spheres. Clinical trials using RN combinations should use RN proportions tuned to the patient dosimetry. (125)Sn production and its coupling to somatostatin analogue appear feasible. Assuming similar pharmacokinetics (125)Sn is the best RN for combination with (177)Lu in peptide receptor radiotherapy justifying pharmacokinetics studies in rodent of (125)Sn-labelled somatostatin analogues.

Effect of interferon-alpha treatment on [68Ga-DOTA,Tyr3,Thre8]octreotide uptake in CA20948 tumors: a small-animal PET study

In peptide receptor radionuclide therapy of neuroendocrine tumors, improvements have been made by increasing the affinity for receptors and by protecting critical organs (e.g., kidneys). However, tumor parameters involved in radiopeptide uptake are still under investigation. Interferon-α (IFNα) is used as biotherapy for neuroendocrine tumors. Several mechanisms of action are described, but the potential effect of IFNα on tumor uptake of labeled peptide has not been studied in vivo yet.

METHODS

Twenty-six male CA20948 tumor-bearing Lewis rats were imaged before and during IFNα treatment using quantitative small-animal PET with [(68)Ga-DOTA,Tyr(3),Thre(8)]octreotide. Imaging was performed at days 0, 3, and 7. Animals were divided into 3 groups according to the treatment: control (injected daily with saline), half (4 d of IFNα treatment from day 0 to day 3, then saline), and full (7 d of IFNα). A daily dose of IFNα (1.5 mIU) was administered subcutaneously. Quantitative PET results are expressed as percentage injected dose per cm(3) and normalized to baseline (day 0) values. Tumor size was monitored by PET and caliper measurements.

RESULTS

Gross tumor uptake and tumor volumes increased in all groups over the 7-d period. On day 3, mean ± SD ratios to day 0 were 1.2 ± 0.2, 1.3 ± 0.5, and 1.2 ± 0.4, respectively, for control, half, and full groups. On day 7, respective values were 1.1 ± 0.2, 1.3 ± 0.6, and 1.5 ± 0.4. At day 3, a comparison among groups showed no statistically significant difference. At day 7, the full group showed a significantly higher ratio in activity concentration than the control group (P = 0.021). A good correlation was found between tumor volumes assessed by small-animal PET and caliper measurements (R = 0.89, P < 0.0001).

CONCLUSION

As expected, over a period of 7 d, both tumor volumes and radiopeptide uptake increased in all animals. However, the activity concentration increased significantly more at day 7 in animals treated for 7 d with IFNα, compared with controls. This is the first, to our knowledge, in vivo indication that IFNα is able to increase tumor uptake of the labeled analog in a small-animal model of neuroendocrine tumors. The mechanisms underlying this effect (flow, vascular permeability, receptor upregulation) remain unknown and need to be further investigated.

Tumor uptake of 68Ga-DOTA-Tyr3-octreotate: animal PET studies of tumor flow and acute somatostatin receptor modulation in the CA20948 rat model

INTRODUCTION

Factors determining the in vivo uptake of radiolabeled somatostatin analogs by neuroendocrine tumors are poorly known. The aim is to evaluate in vivo tumor perfusion and regulation of somatostatin receptors (sstr) following acute exposure to octreotide, in an animal model of neuroendocrine tumor.

METHODS

H(2)(15)O flow studies were performed in 8 CA20948 tumor-bearing rats and another 36 rats underwent three [(68)Ga]-DOTA-Tyr(3)-octreotate imaging sessions at 24-h intervals. After baseline (Day 0) imaging, scanning was repeated on Day 1 after octreotide injection (175 microg/kg), with a variable delay: no injection (controls, n=7), coinjection (n=6), and octreotide injection 20 min (n=7), 2 h (n=8) and 4 h (n=8) before imaging. Repeat images without octreotide was performed at Day 2 followed by sacrifice and tumor counting.

RESULTS

H(2)(15)O studies failed to measure quantitative tumor perfusion in this model. On Day 1, ratio of tumor uptake to Day 0 was 1.2+/-0.3 in controls; 0.6+/-0.2 in the coinjection group; 0.9+/-0.2, 1.1+/-0.1 and 1.2+/-0.2 in the other groups, respectively. Uptake in the coinjection group showed a statistically significant reduction of tumor uptake (P<.0001). All groups showed increased uptake on Day 2, without statistical differences between groups. In vivo tumor counts showed good correlation with ex vivo countings (R(2)=0.946).

CONCLUSION

Acute exposure to unlabeled octreotide in this neuroendocrine tumor model results in a rapid recycling or resynthesis of sstr. Positron emission tomography (PET) allowed to reliably assess quantitative uptake of [(68)Ga]-DOTA-Tyr(3)-octreotate over time in the same animal, but failed in this model to measure tumor perfusion.