Automated and robust organ segmentation for 3D-based internal dose calculation

PURPOSE

In this work, we address image segmentation in the scope of dosimetry using deep learning and make three main contributions: (a) to extend and optimize the architecture of an existing convolutional neural network (CNN) in order to obtain a fast, robust and accurate computed tomography (CT)-based organ segmentation method for kidneys and livers; (b) to train the CNN with an inhomogeneous set of CT scans and validate the CNN for daily dosimetry; and (c) to evaluate dosimetry results obtained using automated organ segmentation in comparison with manual segmentation done by two independent experts.

METHODS

We adapted a performant deep learning approach using CT-images to delineate organ boundaries with sufficiently high accuracy and adequate processing time. The segmented organs were consequently used as binary masks for further convolution with a point spread function to retrieve the activity values from quantitatively reconstructed SPECT images for "volumetric"/3D dosimetry. The resulting activities were used to perform dosimetry calculations with the kidneys as source organs.

RESULTS

The computational expense of the algorithm was sufficient for clinical daily routine, required minimum pre-processing and performed with acceptable accuracy a Dice coefficient of [Formula: see text] for liver segmentation and of [Formula: see text] for kidney segmentation, respectively. In addition, kidney self-absorbed doses calculated using automated segmentation differed by [Formula: see text] from dosimetry performed by two medical physicists in 8 patients.

CONCLUSION

The proposed approach may accelerate volumetric dosimetry of kidneys in molecular radiotherapy with 177Lu-labelled radiopharmaceuticals such as 177Lu-DOTATOC. However, even though a fully automated segmentation methodology based on CT images accelerates organ segmentation and performs with high accuracy, it does not remove the need for supervision and corrections by experts, mostly due to misalignments in the co-registration between SPECT and CT images. Trial registration EudraCT, 2016-001897-13. Registered 26.04.2016, www.clinicaltrialsregister.eu/ctr-search/search?query=2016-001897-13 .

A 3-D Method for Delineation of Activity Distributions and Assessment of Functional Organ Volumes with Spect

The distribution volume of an organ may have a clinical impact in many cases and various methods have been designed to make volume assessments. In this paper, we describe a new method for delineation of the distribution outline and volume determination. The method is based on smoothing, differentiation, image relaxation and voxel counting of single photon emission computer tomography (SPECT) image sets with 3-D operators. A special routine corrects for the inherent thickness of the voxel-based outline. Phantom experiments, using a SPECT system with LEGP-collimator and a 64×64 acquisition matrix with 6.3×6.3 mm2 pixel size, demonstrated good correlation between the measured and the true volumes. For volumes larger than 120 cc the correlation coefficient was 0.9999 with SE 1.0 cc and an average relative deviation of 0.49%. For volumes below 120 cc, the accuracy was impaired due to low resolution power. By improving the system spatial resolution with an LEHR-collimator and a smaller pixel-size (4.1×4.1 mm2), good accuracy was achieved also for volumes in the range from 3 to 120 cc. Measurements of 15 differently shaped phantoms of volumes between 3 and 104 cc demonstrated high correlation between measured and true volumes: R=0.9921 and SE=0.74 cc (5.3%). For volumes as small as 3 and 5 cc, the difference between the true and the assessed volume was 0.6 cc. The reproducibility of the method was within 3% for volumes above 120 cc and within 7% for volumes below. Due to this accuracy, we conclude that the method can be applied for various clinical routine and research applications using SPECT.

32nd International Austrian Winter Symposium : Zell am See, the Netherlands. 20-23 January 2016

A1 68Ga-PSMA PET/CT in staging and restaging of Prostate Cancer Patients: comparative study with 18F-Choline PET/CT

W Langsteger, A Rezaee, W Loidl, HS Geinitz, F Fitz, M Steinmair, G Broinger, L Pallwien-Prettner, M Beheshti

A2 F18 Choline PET – CT: an accurate diagnostic tool for the detection of parathyroid adenoma?

L Imamovic, M Beheshti, G Rendl, D Hackl, O Tsybrovsky, M Steinmair, K Emmanuel, F Moinfar, C Pirich, W Langsteger

A3 [18F]Fluoro-DOPA-PET/CT in the primary diagnosis of medullary thyroid carcinoma

A Bytyqi, G Karanikas, M Mayerhöfer, O Koperek, B Niederle, M Hartenbach

A4 Variations of clinical PET/MR operations: An international survey on the clinical utilization of PET/MRI

T Beyer, K Herrmann, J Czernin

A5 Standard Dixon-based attenuation correction in combined PET/MRI: Reproducibility and the possibility of Lean body mass estimation

I Rausch, P Rust, MD DiFranco, M Lassen, A Stadlbauer, ME Mayerhöfer, M Hartenbach, M Hacker, T Beyer

A6 High resolution digital FDG PET/MRI imaging for assessment of ACL graft viability

K Binzel, R Magnussen, W Wei, MU Knopp, DC Flanigan, C Kaeding, MV Knopp

A7 Using pre-existing hematotoxicity as predictor for severe side effects and number of treatment cycles of Xofigo therapy

A Leisser, M Nejabat, M Hartenbach, G Kramer, M Krainer, M Hacker, A Haug

A8 QDOSE – comprehensive software solution for internal dose assessment

Wencke Lehnert, Karl Schmidt, Sharok Kimiaei, Marcus Bronzel, Andreas Kluge

A9 Clinical impact of Time-of-Flight on next-generation digital PET imaging of Yttrium-90 radioactivity following liver radioembolization

CL Wright, K Binzel, J Zhang, Evan Wuthrick, Piotr Maniawski, MV Knopp

A10 Snakes in patients! Lessons learned from programming active contours for automated organ segmentation

M Blaickner, E Rados, A Huber, M Dulovits, H Kulkarni, S Wiessalla, C Schuchardt, RP Baum, B Knäusl, D Georg

A11 Influence of a genetic polymorphism on brain uptake of the dual ABCB1/ABCG2 substrate [11C]tariquidar

M Bauer, B Wulkersdorfer, W Wadsak, C Philippe, H Haslacher, M Zeitlinger, O Langer

A12 Outcome prediction of temporal lobe epilepsy surgery from P-glycoprotein activity. Pooled analysis of (R)-[11C]-verapamil PET data from two European centres

M Bauer, M Feldmann, R Karch, W Wadsak, M Zeitlinger, MJ Koepp, M-C Asselin, E Pataraia, O Langer

A13 In-vitro and in-vivo characterization of [18F]FE@SNAP and derivatives for the visualization of the melanin concentrating hormone receptor 1

M Zeilinger, C Philippe, M Dumanic, F Pichler, J Pilz, M Hacker, W Wadsak, M Mitterhauser

A14 Reducing time in quality control leads to higher specific radioactivity of short-lived radiotracers

L Nics, B Steiner, M Hacker, M Mitterhauser, W Wadsak

A15 In vitro 11C-erlotinib binding experiments in cancer cell lines with epidermal growth factor receptor mutations

A Traxl, Thomas Wanek, Kushtrim Kryeziu, Severin Mairinger, Johann Stanek, Walter Berger, Claudia Kuntner, Oliver Langer

A16 7-[11C]methyl-6-bromopurine, a PET tracer to measure brain Mrp1 function: radiosynthesis and first PET evaluation in mice

S Mairinger, T Wanek, A Traxl, M Krohn, J Stanek, T Filip, M Sauberer, C Kuntner, J Pahnke, O Langer

A17 18F labeled azidoglucose derivatives as “click” agents for pretargeted PET imaging

D Svatunek, C Denk, M Wilkovitsch, T Wanek, T Filip, C Kuntner-Hannes, J Fröhlich, H Mikula

A18 Bioorthogonal tools for PET imaging: development of radiolabeled 1,2,4,5-Tetrazines

C Denk, D Svatunek, T Wanek, S Mairinger, J Stanek, T Filip, J Fröhlich, H Mikula, C Kuntner-Hannes

A19 Preclinical evaluation of [18F]FE@SUPPY- a new PET-tracer for oncology

T Balber, J Singer, J Fazekas, C Rami-Mark, N Berroterán-Infante, E Jensen-Jarolim, W Wadsak, M Hacker, H Viernstein, M Mitterhauser

A20 Investigation of Small [18F]-Fluoroalkylazides for Rapid Radiolabeling and In Vivo Click Chemistry

C Denk, D Svatunek, B Sohr, H Mikula, J Fröhlich, T Wanek, C Kuntner-Hannes, T Filip

A21 Microfluidic 68Ga-radiolabeling of PSMA-HBED-CC using a flow-through reactor

S Pfaff, C Philippe, M Mitterhauser, M Hartenbach, M Hacker, W Wadsak

A22 Influence of 24-nor-ursodeoxycholic acid on hepatic disposition of [18F]ciprofloxacin measured with positron emission tomography

T Wanek, E Halilbasic, M Visentin, S Mairinger, B Stieger, C Kuntner, M Trauner, O Langer

A23 Automated 18F-flumazenil production using chemically resistant disposable cassettes

P Lam, M Aistleitner, R Eichinger, C Artner

A24 Similarities and differences in the synthesis and quality control of 177Lu-DOTA-TATE, 177Lu -HA-DOTA-TATE and 177Lu-DOTA-PSMA (PSMA-617)

H Eidherr, C Vraka, A Haug, M Mitterhauser, L Nics, M Hartenbach, M Hacker, W Wadsak

A25 68Ga- and 177Lu-labelling of PSMA-617

H Kvaternik, R Müller, D Hausberger, C Zink, RM Aigner

A26 Radiolabelling of liposomes with 67Ga and biodistribution studies after administration by an aerosol inhalation system

U Cossío, M Asensio, A Montes, S Akhtar, Y te Welscher, R van Nostrum, V Gómez-Vallejo, J Llop

A27 Fully automated quantification of DaTscan SPECT: Integration of age and gender differences

F VandeVyver, T Barclay, N Lippens, M Troch

A28 Lesion-to-background ratio in co-registered 18F-FET PET/MR imaging – is it a valuable tool to differentiate between low grade and high grade brain tumor?

L Hehenwarter, B Egger, J Holzmannhofer, M Rodrigues-Radischat, C Pirich

A29 [11C]-methionine PET in gliomas - a retrospective data analysis of 166 patients

N Pötsch, I Rausch, D Wilhelm, M Weber, J Furtner, G Karanikas, A Wöhrer, M Mitterhauser, M Hacker, T Traub-Weidinger

A30 18F-Fluorocholine versus 18F-Fluorodeoxyglucose for PET/CT imaging in patients with relapsed or progressive multiple myeloma: a pilot study

T Cassou-Mounat, S Balogova, V Nataf, M Calzada, V Huchet, K Kerrou, J-Y Devaux, M Mohty, L Garderet, J-N Talbot

A31 Prognostic benefit of additional SPECT/CT in sentinel lymph node mapping of breast cancer patients

S Stanzel, G Pregartner, T Schwarz, V Bjelic-Radisic, B Liegl-Atzwanger, R Aigner

A32 Evaluation of diagnostic value of TOF-18F-FDG PET/CT in patients with suspected pancreatic cancer

S Stanzel, F Quehenberger, RM Aigner

A33 New quantification method for diagnosis of primary hyperpatahyroidism lesions and differential diagnosis vs thyropid nodular disease in dynamic scintigraphy

A Koljević Marković, Milica Janković, V Miler Jerković, M Paskaš, G Pupić, R Džodić, D Popović

A34 A rare case of diffuse pancreatic involvement in patient with merkel cell carcinoma detected by 18F-FDG

MC Fornito, D Familiari

A35 TSH-stimulated 18F-FDG PET/CT in the diagnosis of recurrent/metastatic radioiodine-negative differentiated thyroid carcinomas in patients with various thyroglobuline levels

P Koranda, H Polzerová, I Metelková, L Henzlová, R Formánek, E Buriánková, M Kamínek

A36 Breast Dose from lactation following I131 treatment

WH Thomson, C Lewis

A37 A new concept for performing SeHCAT studies with the gamma camera

WH Thomson, J O’Brien, G James, A Notghi

A38 Whole body F-18-FDG-PET and tuberculosis: sensitivity compared to x-ray-CT

H Huber, I Stelzmüller, R Wunn, M Mandl, F Fellner, B Lamprecht, M Gabriel

A39 Emerging role 18F-FDG PET-CT in the diagnosis and follow-up of the infection in heartware ventricular assist system (HVAD)

MC Fornito, G Leonardi

A40 Validation of Poisson resampling software

WH Thomson, J O’Brien, G James

A41 Protection of PET nuclear medicine personnel: problems in satisfying dose limit requirements

J Hudzietzová, J Sabol, M Fülöp

Concentration of 123I-metaiodobenzylguanidine in left and right liver lobes. Findings indicate regional differences in function in the normal liver

PURPOSE

Normal radioactivity in the liver is often shown to be higher in the left lobe than in the right lobe at clinical examination by single photon emission computed tomography (SPECT) with 123I-metaiodobenzylguanidine (123I-MIBG). Our objective was to determine whether this represents a pathological condition.

MATERIAL AND METHODS

The distribution of 123I-MIBG in the liver was retrospectively studied in clinical patients who had normal CT examinations of the liver and spleen. In the SPECT sections of 27 123I-MIBG examinations, we determined the activity ratios between the left and right lobes. The control group comprised 33 examinations with (111)In-pentetreotide (OctreoScan) and 23 examinations with 99mTc-antigranulocyte antibody.

RESULTS

The mean activity ratio between left and right lobes for 123I-MIBG was 1.24, and for (111)In-pentetreotide 0.87. These figures differed significantly both from each other and from 1.00. The mean ratio for 99mTc-antibody at the early examinations was 0.99, and at the late examinations 0.95. These figures differed significantly from the values for 123I-MIBG and (111)In-pentetreotide.

CONCLUSION

The increased concentration of 123I-MIBG in the left lobe of the liver compared to the right lobe is a normal finding. The reason for this cannot be explained. The difference in the way in which the left and right lobes deal with the three radiopharmaceuticals is a new finding and indicates a regional difference in liver function that has not been reported previously.

Optimal design of planar-concave collimators for SPECT--an analytical approach

In this paper we present an analytical tool for the design and optimization of planar-concave collimators for SPECT. We conclude that a single general planar-concave collimator that eliminates the non-isotropic blurring for all SPECT applications does not exist. On the other hand, it is possible to achieve pseudo-optimal collimators for different clinical applications by a careful choice of the design parameters. By classifying the clinical applications into two groups, for instance body and brain studies, the non-isotropic blurring for most SPECT situations may be dramatically reduced by means of planar-concave collimators. The results based on Monte Carlo simulations show that the shape of the point source response function of the reconstructed image obtained from the planar-concave collimator is more isotropic than that obtained from a conventional parallel-hole collimator. Specifically, the ratio of the FWHM of the reconstructed point response function in the radial and tangential direction is increased from 0.5 for the parallel-hole to 0.85 for the planar-concave collimator for sources 200 mm away from the centre of rotation.

Resting state rCBF mapping with single-photon emission tomography and positron emission tomography: magnitude and origin of differences

Single-photon emission tomography (SPET), using technetium-99m hexamethylpropylene amine oxime, and positron emission tomography (PET), using oxygen-15 butanol were compared in six healthy male volunteers with regard to the mapping of resting state regional cerebral blood flow (rCBF). A computerized brain atlas was utilized for 3D regional analyses and comparison of 64 selected and normalized volumes of interest (VOIs). The normalized mean rCBF values in SPET, as compared to PET, were higher in most of the Brodmann areas in the frontal and parietal lobes (4.8% and 8.7% respectively). The average differences were small in the temporal (2. 3%) and occipital (1.1%) lobes. PET values were clearly higher in small VOIs like the thalamus (12.3%), hippocampus (12.3%) and basal ganglia (9.9%). A resolution phantom study showed that the in-plane SPET/PET system resolution was 11.0/7.5 mm. In conclusion, SPET and PET data demonstrated a fairly good agreement despite the superior spatial resolution of PET. The differences between SPET and PET rCBF are mainly due to physiological and physical factors, the data processing, normalization and co-registration methods. In order to further improve mapping of rCBF with SPET it is imperative not only to improve the spatial resolution but also to apply accurate correction techniques for scatter, attenuation and non-linear extraction.

Evaluation of optimally designed planar-concave collimators in single-photon emission tomography

The imaging properties of optimally designed planar-concave (PC) collimators were evaluated by means of Monte Carlo simulations. The evaluation was done with respect to total system spatial resolution and the overall image noise distribution in single-photon emission tomography. The results showed that the non-isotropy with PC collimators, assessed by the ratio of the full-width at half-maximum in the radial and tangential directions, was reduced by about 60% as compared with a conventional parallel-hole collimator for sources located 200 mm away from the centre of rotation. Furthermore, the image noise distribution along the object radius became more uniform when the curved collimator was used. The maximum increase in noise due to use of the curved collimator was about 45% close to the edge of the phantom, where the hole length was about 3 times longer. We also showed with Monte Carlo simulations that the spatial resolution of the lateral cortex when using the curved collimator was significantly improved due to improved radial resolution.

Effect of food intake on liver and spleen volume: assessment with single photon emission computed tomography

RATIONALE AND OBJECTIVES

The authors evaluated the influence of food intake on liver and spleen size. The rationale was that since a meal induces a marked increase in splanchnic blood flow, organ volume might also be affected. This could influence conclusions made at clinical examination.

METHODS

Single photon emission computed tomography (SPECT) was performed in 20 healthy volunteers after administration of technetium-99m albumin colloid before and after ingestion of a standardized meal. A semiautomatic technique was used to determine liver and spleen size.

RESULTS

There was a nonsignificant reduction in liver volume of 1.5% (P = .063) and a significant reduction in spleen volume of 3.2% (P = .00033) after food intake.

CONCLUSION

Liver and spleen volume are marginally affected by food intake, and the changes are within the methodologic error of radiologic modalities.

Collimator design for improved spatial resolution in SPECT and planar scintigraphy

This paper evaluates the design of a new planar-concave collimator with nonuniform response that better matches the body shape than conventional collimators.

METHODS

The collimator properties are evaluated and assessed by means of both a stimulation program and an experimental test using a prototype planar-concave collimator.

RESULTS

The results, for points located 150 mm from the axis of rotation, demonstrate that the ratio of radial and tangential spatial resolution in SPECT with the new collimator decreased by 40%, as compared to SPECT with a standard collimator. In planar scintigraphy, the spatial resolution improved correspondingly from 10.9 mm (FWHM) to 7.8 mm in lateral areas of the body.

CONCLUSION

The new collimator reduces nonisotropic blurring in SPECT in addition to improving spatial resolution in both planar scintigraphy and in SPECT.

A 3-D method for delineation of activity distributions and assessment of functional organ volumes with SPECT

The distribution volume of an organ may have a clinical impact in many cases and various methods have been designed to make volume assessments. In this paper, we describe a new method for delineation of the distribution outline and volume determination. The method is based on smoothing, differentiation, image relaxation and voxel counting of single photon emission computer tomography (SPECT) image sets with 3-D operators. A special routine corrects for the inherent thickness of the voxel-based outline. Phantom experiments, using a SPECT system with LEGP-collimator and a 64 x 64 acquisition matrix with 6.3 x 6.3 mm2 pixel size, demonstrated good correlation between the measured and the true volumes. For volumes larger than 120 cc the correlation coefficient was 0.9999 with SE 1.0 cc and an average relative deviation of 0.49%. For volumes below 120 cc, the accuracy was impaired due to low resolution power. By improving the system spatial resolution with an LEHR-collimator and a smaller pixel-size (4.1 x 4.1 mm2), good accuracy was achieved also for volumes in the range from 3 to 120 cc. Measurements of 15 differently shaped phantoms of volumes between 3 and 104 cc demonstrated high correlation between measured and true volumes: R = 0.9921 and SE = 0.74 cc (5.3%). For volumes as small as 3 and 5 cc, the difference between the true and the assessed volume was 0.6 cc. The reproducibility of the method was within 3% for volumes above 120 cc and within 7% for volumes below.(ABSTRACT TRUNCATED AT 250 WORDS)

Change in size, shape and radiocolloid uptake of the alcoholic liver during alcohol withdrawal, as demonstrated by single photon emission computed tomography

The volume of the total liver and separate right and left lobes was studied before and after 1 week of alcohol withdrawal in 16 consecutive alcoholics by means of single photon emission computed tomography after intravenous injection of 99Tcm-human albumin colloid; the relative tissue distribution of radioactivity was also followed. The left liver lobe increased in volume more than the right lobe during drinking and decreased more rapidly after alcohol withdrawal. Median volume reductions during 1 week of alcohol withdrawal were: total liver 12%, left lobe 26%, and right lobe 8%, indicating that half of the reduction to values of a control group was achieved during this first week. The volume of the right but not of the left lobe was significantly correlated to body size in alcoholics and in controls. The left lobe had a lower capacity to concentrate the radiocolloid than the right lobe in alcoholics and in controls. The liver/spleen, liver/bone marrow and liver/background radioactivity concentration ratios in the alcoholics increased during alcohol withdrawal. We conclude that heavy drinking causes both an increased total liver volume and a change in liver shape, with a relatively more enlarged left than right lobe, as well as a decreased capacity to concentrate radiocolloid. These changes are rapidly reversible during abstinence from alcohol.

A better target-to-background activity ratio using a large-sized colloid compared to a nano-sized colloid for scintigraphy of the peripheral bone marrow. A study in mice and humans

Small-sized radiocolloids with a diameter less than 100 nm (nanocolloids), reduce the confounding liver and spleen uptake compared to large-sized colloids at bone marrow scintigraphy. Consequently, nanocolloids are today generally used for this purpose. Such colloids, though, seem to produce considerable background activity, especially from the large vessels and the urinary tract. This has been studied in mice and humans using a nanocolloid and a large-sized colloid, both composed of human albumin. In mice, the findings confirm a lower relative and absolute uptake of the nanocolloid in the liver and spleen compared to the large-sized colloid, whereas the bone marrow activity relative to the activity of kidneys, heart, lungs, intestines and peripheral blood is higher for the large-sized colloid than for the nanocolloid. The latter was confirmed in humans who were examined by single photon emission computed tomography corrected for attenuation and scattering of photons. It is concluded that a large-sized colloid should be used for spot examination of a specific marrow region outside the liver and spleen, while a nanocolloid must be used when examining structures close to the liver and spleen as well as for depiction of the entire bone marrow.