PURPOSE: We developed predictive models using different programming languages and different computing platforms for machine learning (ML) and deep learning (DL) that classify clinical diagnoses in patients with epiphora. We evaluated the diagnostic performance of these models.METHODS: Between January 2016 and September 2017, 250 patients with epiphora who underwent dacryocystography (DCG) and lacrimal scintigraphy (LS) were included in the study. We developed five different predictive models using ML tools, Python-based TensorFlow, R, and Microsoft Azure Machine Learning Studio (MAMLS). A total of 27 clinical characteristics and parameters including variables related to epiphora (VE) and variables related to dacryocystography (VDCG) were used as input data. Apart from this, we developed two predictive convolutional neural network (CNN) models for diagnosing LS images. We conducted this study using supervised learning.RESULTS: Among 500 eyes of 250 patients, 59 eyes had anatomical obstruction, 338 eyes had functional obstruction, and the remaining 103 eyes were normal. For the data set that excluded VE and VDCG, the test accuracies in Python-based TensorFlow, R, multiclass logistic regression in MAMLS, multiclass neural network in MAMLS, and nuclear medicine physician were 81.70%, 80.60%, 81.70%, 73.10%, and 80.60%, respectively. The test accuracies of CNN models in three-class classification diagnosis and binary classification diagnosis were 72.00% and 77.42%, respectively.CONCLUSIONS: ML-based predictive models using different programming languages and different computing platforms were useful for classifying clinical diagnoses in patients with epiphora and were similar to a clinician's diagnostic ability.
Classification ; Dataset ; Diagnosis ; Humans ; Lacrimal Apparatus Diseases ; Learning ; Logistic Models ; Machine Learning ; Nuclear Medicine ; Programming Languages ; Radionuclide Imaging

PURPOSE: We developed predictive models using different programming languages and different computing platforms for machine learning (ML) and deep learning (DL) that classify clinical diagnoses in patients with epiphora. We evaluated the diagnostic performance of these models. METHODS: Between January 2016 and September 2017, 250 patients with epiphora who underwent dacryocystography (DCG) and lacrimal scintigraphy (LS) were included in the study. We developed five different predictive models using ML tools, Python-based TensorFlow, R, and Microsoft Azure Machine Learning Studio (MAMLS). A total of 27 clinical characteristics and parameters including variables related to epiphora (VE) and variables related to dacryocystography (VDCG) were used as input data. Apart from this, we developed two predictive convolutional neural network (CNN) models for diagnosing LS images. We conducted this study using supervised learning. RESULTS: Among 500 eyes of 250 patients, 59 eyes had anatomical obstruction, 338 eyes had functional obstruction, and the remaining 103 eyes were normal. For the data set that excluded VE and VDCG, the test accuracies in Python-based TensorFlow, R, multiclass logistic regression in MAMLS, multiclass neural network in MAMLS, and nuclear medicine physician were 81.70%, 80.60%, 81.70%, 73.10%, and 80.60%, respectively. The test accuracies of CNN models in three-class classification diagnosis and binary classification diagnosis were 72.00% and 77.42%, respectively. CONCLUSIONS ML-based predictive models using different programming languages and different computing platforms were useful for classifying clinical diagnoses in patients with epiphora and were similar to a clinician's diagnostic ability.

PURPOSE: We investigated the incidence, location, and clinical significance of focal ¹⁸F-FDG uptake of the spinal cord in patients with cancer.METHODS: We reviewed the medical records of 22,937 consecutive adult patients with known or suspicious malignancy who underwent ¹⁸F-FDG PET/CT. PET/CT scans with incidental focal spinal cord uptake were selected and retrospectively reviewed to determine the presence, location, number, and maximum standardized uptake value (SUV(max)) of any focal hypermetabolic lesions of the spinal cord. In subjects with focal spinal uptake, clinical characteristics and clinical follow-up results, including follow-up PET/CT, were reviewed.RESULTS: Incidental focal spinal cord uptake was observed in 69 of 22,937 adult patients (incidence = 0.3%; M:F = 31:38; age, 55.8 ± 14.7 years). Seventy-eight focal hypermetabolic lesions on spinal cord in the PET/CT scans of the 69 study subjects were analyzed. The most common sites of focal spinal cord uptake were the T12 vertebra (47/78; 60.3%) and L1 vertebra (20/78; 25.6%). Multifocal cord uptake was found in 8 of 69 patients (11.6%). The average SUV(max) for cord uptake was 2.5 ± 0.5 (range, 1.4∼3.9). There was no clinical or imaging evidence of abnormalities in the spinal cord, both at the time of PET/CT and during clinical follow-up.CONCLUSIONS: Although incidental focal ¹⁸F-FDG uptake of the spinal cord is rare in patients with cancer, it may be physiological or benign, but it should not be considered as malignant involvement. Common sites for the uptake were in the T12 and L1 spine levels.
Adult ; Diagnosis, Differential ; Follow-Up Studies ; Humans ; Incidence ; Medical Records ; Positron-Emission Tomography and Computed Tomography ; Retrospective Studies ; Spinal Cord ; Spine

OBJECTIVE: We investigated the clinical significance of incidental diffuse thyroid uptake (DTU) on 18F-FDG PET in subjects without a history of cancer. MATERIALS AND METHODS: This study included 2062 studies from adults who underwent 18F-FDG PET as a cancer screening program. Subjects were divided into the following two groups: with (group I) or without (group II) DTU. The presence of DTU and the thyroid visual grading score were compared with thyroid function tests, serum anti-microsomal antibody (AMA) levels, and the presence of diffuse parenchymal change (DPC) on ultrasonography (USG). RESULTS: DTU was found in 6.6% of the scans (137/2062). Serum thyroid stimulating hormone (TSH) and AMA levels were significantly higher in group I than in group II. Increased AMA level (55.1%) and DPC (48.7%) were more frequently found in group I (p < 0.001). The proportion of subjects with any abnormal results in serum free thyroxine, triiodothyronine, TSH, or AMA levels or DPC on USG was significantly higher in group I than in group II (71.5% vs. 10.6%, p < 0.001), and was significantly and gradually increased according to the visual grading score group (0 vs. 1-2 vs. 3-4 = 10.6% vs. 58.5% vs. 90.9%, p < 0.001). TSH and is AMA levels were significantly increased according to the visual grading score. CONCLUSION: The presence or degree of incidental DTU on 18F-FDG PET is closely correlated with increased serum AMA and TSH levels, and the presence of DPC on USG. Therefore, the most plausible pathological cause of DTU may be cell damage by an autoimmune mechanism.
Adult ; Aged ; Aged, 80 and over ; Antibodies/blood ; Female ; Fluorodeoxyglucose F18/*diagnostic use/pharmacokinetics ; Humans ; *Incidental Findings ; Male ; Microsomes/immunology ; Middle Aged ; Neoplasms ; Positron-Emission Tomography/methods ; Radiopharmaceuticals/*diagnostic use/pharmacokinetics ; Retrospective Studies ; Thyroid Gland/metabolism/*radionuclide imaging/ultrasonography ; Thyrotropin/blood ; Young Adult

PURPOSE: We evaluated the incidence and malignant risk of focal breast lesions incidentally detected by 18F-FDG PET/CT. Various PET/CT findings of the breast lesions were also analyzed to improve the differentiation between benign from malignant focal breast lesions. MATERIALS AND METHODS: The subjects were 3,768 consecutive 18F-FDG PET/CT exams performed in adult females without a history of breast cancer. A focal breast lesion was defined as a focal 18F-FDG uptake or a focal nodular lesion on CT image irrespective of 18F-FDG uptake in the breasts. The maximum SUV and CT pattern of focal breast lesions were evaluated, and were compared with final diagnosis. RESULTS: The incidence of focal breast lesions on PET/CT in adult female subjects was 1.4% (58 lesions in 53 subjects). In finally confirmed 53 lesions of 48 subjects, 11 lesions of 8 subjects (20.8%) were proven to be malignant. When the PET/CT patterns suggesting benignancy (maximum attenuation value > 75 HU or < 30 HU; standard deviation of mean attenuation > 20) were added as diagnostic criteria of PET/CT to differentiate benign from malignant breast lesions along with maximum SUV, the area under ROC curve of PET/CT was significantly increased compared with maximum SUV alone (0.680+/-0.093 vs. 0.786+/-0.076, p<0.05). CONCLUSION: The malignant risk of focal breast lesions incidentally found on 18F-FDG PET/CT is not low, deserving further diagnostic confirmation. Image interpretation considering both 18F-FDG uptake and PET/CT pattern may be helpful to improve the differentiation from malignant and benign focal breast lesion.
Adult ; Breast ; Breast Neoplasms ; Early Detection of Cancer ; Female ; Fluorodeoxyglucose F18 ; Humans ; Incidence ; ROC Curve

PURPOSE: This study was designed to investigate the cellular uptake of various tumor imaging radiopharmaceuticals in human breast cancer cells before and after paclitaxel exposure considering viable cell number. MATERIALS AND METHODS: F-18-fluorodeoxyglucose, C-11-methionine, Tl-201, Tc-99m-MIBI, and Tc-99m-tetrofosmin were used to evaluate the cellular uptake in MCF-7 cells. MCF-7 cells were cultured in multi-well plates. Wells were divided into DMSO exposure control group, and paclitaxel exposure group. The exposure durations of paclitaxel with 10 nM or 100 nM were 2 h, 6 h, 12 h, 24 h, and 48 h. RESULTS: Viable cell fraction was reduced as the concentration and exposure time of paclitaxel increased. After 10 nM paclitaxel exposure, the cellular uptake of all 5 radiopharmaceuticals was not reduced significantly, irrespective of exposure time and viable cell fraction. After 100 nM paclitaxel exposure, the cellular uptake of all 5 radiopharmaceuticals was enhanced significantly irrespective of viable cell fraction. The peak uptake was observed in experimental groups with paclitaxel exposure for 6 to 48 h according the type of radiopharmaceutical. When the cellular uptake was adjusted for the viable cell fraction and cell count, the peak cellular uptake was observed in experimental groups with paclitaxel exposure for 48 h, irrespective of the type of radiopharmaceutical. CONCLUSION: The cellular uptake of F-18-fluorodeoxyglucose, C-11-methionine, Tl-201, Tc-99m-MIBI, and Tc-99m-tetrofosmin did not reflect viable cell number in MCF-7 cells after paclitaxel exposure for up to 48 h.
Breast Neoplasms* ; Breast* ; Cell Count ; Cell Line* ; Cell Survival ; Dimethyl Sulfoxide ; Humans* ; MCF-7 Cells ; Paclitaxel* ; Radiopharmaceuticals

PURPOSE: It was reported that CT-based measured attenuation correction (CT-MAC) produced radioactivity concentration values significantly higher than 68Ge-based segmented attenuation correction (Ge-SAC) in PET images. However, it was unknown whether the radioactivity concentration difference resulted from different sources (CT vs. Ge) or types (MAC vs. SAC) of attenuation correction (AC). We evaluated the influences of the source and type of AC on the radioactivity concentration differences between reconstructed PET images in normal subjects and patients. MATERIAL AND METHODS: Five normal subjects and 35 patients with a known or suspected cancer underwent 18F-FDG PET/CT. In each subject, attenuation corrected PET images using OSEM algorithm (28 subsets, 2 iterations) were reconstructed by 4 methods: CT-MAC, CT-SAC, Ge-MAC, and Ge-SAC. The physiological uptake in normal subjects and pathological uptake in patients were quantitatively compared between the PET images according to the source and type of AC. RESULTS: The SUVs of physiological uptake measured in CT-MAC PET images were significantly higher than other 3 differently corrected PET images. Maximum SUVs of the 145 foci with abnormal FDG uptake in CT-MAC images were significantly highest among 4 differently corrected PET images with a difference of 2.4% to 5.1% (p<0.001). The SUVs of pathological uptake in Ge-MAC images were significantly higher than those in CT-SAC and Ge-MAC PET images (p<0.001). CONCLUSION: Quantitative radioactivity values were highest in CT-MAC PET images. The adoption of MAC may make a more contribution than the adoption of CT attenuation map to such differences.
Fluorodeoxyglucose F18* ; Humans ; Positron-Emission Tomography and Computed Tomography ; Radioactivity

PURPOSE: Neuroreceptor PET studies require 60-120 minutes to complete and head motion of the subject during the PET scan increases the uncertainty in measured activity. In this study, we investigated the effects of the data-driven head motion correction on the evaluation of endogenous dopamine release (DAR) in the striatum during the motor task which might have caused significant head motion artifact. MATERIALS AND METHODS: [11C]raclopride PET scans on 4 normal volunteers acquired with bolus plus constant infusion protocol were retrospectively analyzed. Following the 50 min resting period, the participants played a video game with a monetary reward for 40 min. Dynamic frames acquired during the equilibrium condition (pre-task: 30-50 min, task: 70-90 min, post-task: 110-120 min) were realigned to the first frame in pre-task condition. Intra-condition registrations between the frames were performed, and average image for each condition was created and registered to the pre-task image (inter-condition registration). Pre-task PET image was then co-registered to own MRI of each participant and transformation parameters were reapplied to the others. Volumes of interest (VOI) for dorsal putamen (PU) and caudate (CA), ventral striatum (VS), and cerebellum were defined on the MRI. Binding potential (BP) was measured and DAR was calculated as the percent change of BP during and after the task. SPM analyses on the BP parametric images were also performed to explore the regional difference in the effects of head motion on BP and DAR estimation. RESULTS: Changes in position and orientation of the striatum during the PET scans were observed before the head motion correction. BP values at pre-task condition were not changed significantly after the intra-condition registration. However, the BP values during and after the task and DAR were significantly changed after the correction. SPM analysis also showed that the extent and significance of the BP differences were significantly changed by the head motion correction and such changes were prominent in periphery of the striatum. CONCLUSION: The results suggest that misalignment of MRI-based VOI and the striatum in PET images and incorrect DAR estimation due to the head motion during the PET activation study were significant, but could be remedied by the data-driven head motion correction.
Artifacts ; Basal Ganglia ; Brain* ; Cerebellum ; Dopamine* ; Head ; Healthy Volunteers ; Magnetic Resonance Imaging ; Positron-Emission Tomography ; Putamen ; Retrospective Studies ; Reward ; Sensory Receptor Cells ; Uncertainty ; Video Games

PURPOSE: We investigated prospectively whether the interpretation considering the patterns of FDG uptake and the findings of unenhanced CT for attenuation correction can improve the diagnostic accuracy for assessing malignant lymph node (LN) and N stage in non-small cell lung cancer (NSCLC) using CT-corrected FDG-PET (PET/CT). MATERIALS AND METHODS: Subjects were 91 NSCLC patients (M/F: 62/29, age: 60+/-9 yr) who underwent PET/CT before LN dissection. We evaluated the maximum SUV ( (max) SUV), patterns of FDG uptake, short axis diameter, and calcification of LN showing abnormally increased FDG uptake. Then we investigated criteria improving the diagnostic accuracy and correlated results with postoperative pathology. In step 1, LN was classified as benign or malignant based on (max) SUV only. In step 2, LN was regarded as benign if it had lower (max) SUV than the cut-off value of step 1 or it had calcification irrespective of its (max) SUV. In step 3, LN regarded as malignant in step 2 was classified as benign if they had indiscrete margin of FDG uptake. RESULTS: Among 432 LN groups surgically resected (28 malignant, 404 benign), 71 showed abnormally increased FDG uptake. We determined the cut-off as (max) SUV = 3.5 using ROC curve analysis. The sensitivity, specificity, and accuracy for assessing malignant LN were 64.3%, 86.9%, 85.4% in step 1, 64.3%, 95.0%, 93.1% in step 2, and 57.1%, 98.0%, 95.4% in step3, respectively. The accuracy for assessing N stage was 64.8% in step 1, 80.2% in step 2, and 85.7% in step 3. CONCLUSION: Interpreting PET/CT, consideration of calcification and shape of the FDG uptake margin along with (max) SUV can improve the diagnostic accuracy for assessing malignant involvement and N stage of hilar and mediastinal LNs in NSCLC.
Axis, Cervical Vertebra ; Carcinoma, Non-Small-Cell Lung* ; Humans ; Lung ; Lymph Nodes ; Pathology ; Positron-Emission Tomography and Computed Tomography ; Prospective Studies ; ROC Curve ; Sensitivity and Specificity

PURPOSE: Since I-125 emits low energy (27-35 keV) radiation, thinner crystal and collimator could be employed and, hence, it is favorable to obtain high quality images. The purpose of this study was to derive the optimized parameters of I-125 SPECT using a new simulation tool, GATE (Geant4 Application for Tomographic Emission). MATERIALS AND METHODS: To validate the simulation method, gamma camera developed by Weisenberger et al. was modeled. NaI (Tl) plate crystal was used and its thickness was determined by calculating detection efficiency. Spatial resolution and sensitivity curves were estimated by changing variable parameters for parallel-hole and pinhole collimator. Performances of I-125 SPECT equipped with the optimal collimator were also estimated. RESULTS: In the validation study, simulations were found to agree well with experimental measurements in spatial resolution (4%) and sensitivity (3%). In order to acquire 98% gamma ray detection efficiency, NaI (Tl) thickness was determined to be 1 mm. Hole diameter (mm), length (mm) and shape were chosen to be 0.2: 5: square and 0.5: 10: hexagonal for high resolution (HR) and general purpose (GP) parallel-hole collimator, respectively. Hole diameter, channel height and acceptance angle of pinhole (PH) collimator were determined to be 0.25 mm, 0.1 mm and 90 degree. The spatial resolutions of reconstructed image of the I-125 SPECT employing HR: GP: PH were 1.2: 1.7: 0.8 mm. The sensitivities of HR: GP: PH were 39.7: 71.9: 5.5 cps/MBq. CONCLUSION: The optimal crystal and collimator parameters for I-125 imaging were derived by simulation using GATE. The results indicate that excellent resolution and sensitivity imaging is feasible using I-125 SPECT.
Animals* ; Computer Simulation* ; Gamma Cameras ; Gamma Rays ; Hydrogen-Ion Concentration ; Tomography, Emission-Computed, Single-Photon*