Efficacy of exponentiation method with a convolutional neural network for classifying lung nodules on CT images by malignancy level.

OBJECTIVES: The aim of this study was to examine the performance of a convolutional neural network (CNN) combined with exponentiating each pixel value in classifying benign and malignant lung nodules on computed tomography (CT) images. MATERIALS AND METHODS: Images in the Lung Image Database Consortium-Image Database Resource Initiative (LIDC-IDRI) were analyzed. Four CNN models were then constructed to classify the lung nodules by malignancy level (malignancy level 1 vs. 2, malignancy level 1 vs. 3, malignancy level 1 vs. 4, and malignancy level 1 vs. 5). The exponentiation method was applied for exponent values of 1.0 to 10.0 in increments of 0.5. Accuracy, sensitivity, specificity, and area under the curve of receiver operating characteristics (AUC-ROC) were calculated. These statistics were compared between an exponent value of 1.0 and all other exponent values in each model by the Mann-Whitney U-test. RESULTS: In malignancy 1 vs. 4, maximum test accuracy (MTA; exponent value = 2.0, 3.0, 3.5, 4.5, 5.0, 5.5, 6.0, 6.5, 7.0, 7.5, 8.0, 8.5, 9.0, 9.5, and 10.0) and specificity (6.5, 7.0, and 9.0) were improved by up to 0.012 and 0.037, respectively. In malignancy 1 vs. 5, MTA (6.5 and 7.0) and sensitivity (1.5) were improved by up to 0.030 and 0.0040, respectively. CONCLUSIONS: The exponentiation method improved the performance of the CNN in the task of classifying lung nodules on CT images as benign or malignant. The exponentiation method demonstrated two advantages: improved accuracy, and the ability to adjust sensitivity and specificity by selecting an appropriate exponent value. CLINICAL RELEVANCE STATEMENT: Adjustment of sensitivity and specificity by selecting an exponent value enables the construction of proper CNN models for screening, diagnosis, and treatment processes among patients with lung nodules. KEY POINTS: • The exponentiation method improved the performance of the convolutional neural network. • Contrast accentuation by the exponentiation method may derive features of lung nodules. • Sensitivity and specificity can be adjusted by selecting an exponent value.

Log file-based patient dose calculations of double-arc VMAT for head-and-neck radiotherapy.

PURPOSE: The log file-based method cannot display dosimetric changes due to linac component miscalibration because of the insensitivity of log files to linac component miscalibration. The purpose of this study was to supply dosimetric changes in log file-based patient dose calculations for double-arc volumetric-modulated arc therapy (VMAT) in head-and-neck cases. METHODS AND MATERIALS: Fifteen head-and-neck cases participated in this study. For each case, treatment planning system (TPS) doses were produced by double-arc and single-arc VMAT. Miscalibration-simulated log files were generated by inducing a leaf miscalibration of ±0.5 mm into the log files that were acquired during VMAT irradiation. Subsequently, patient doses were estimated using the miscalibration-simulated log files. RESULTS: For double-arc VMAT, regarding planning target volume (PTV), the change from TPS dose to miscalibration-simulated log file dose in Dmean was 0.9 Gy and that for tumor control probability was 1.4%. As for organ-at-risks (OARs), the change in Dmean was <0.7 Gy and normal tissue complication probability was <1.8%. A comparison between double-arc and single-arc VMAT for PTV showed statistically significant differences in the changes evaluated by Dmean and radiobiological metrics (P < 0.01), even though the magnitude of these differences was small. Similarly, for OARs, the magnitude of these changes was found to be small. CONCLUSIONS: Using the log file-based method for PTV and OARs, the log file-based method estimate of patient dose using the double-arc VMAT has accuracy comparable to that obtained using the single-arc VMAT.

Clinical impact of dosimetric changes for volumetric modulated arc therapy in log file-based patient dose calculations.

PURPOSE: A log file-based method cannot detect dosimetric changes due to linac component miscalibration because log files are insensitive to miscalibration. Herein, clinical impacts of dosimetric changes on a log file-based method were determined. METHODS AND MATERIALS: Five head-and-neck and five prostate plans were applied. Miscalibration-simulated log files were generated by inducing a linac component miscalibration into the log file. Miscalibration magnitudes for leaf, gantry, and collimator at the general tolerance level were ±0.5mm, ±1°, and ±1°, respectively, and at a tighter tolerance level achievable on current linac were ±0.3mm, ±0.5°, and ±0.5°, respectively. Re-calculations were performed on patient anatomy using log file data. RESULTS: Changes in tumor control probability/normal tissue complication probability from treatment planning system dose to re-calculated dose at the general tolerance level was 1.8% on planning target volume (PTV) and 2.4% on organs at risk (OARs) in both plans. These changes at the tighter tolerance level were improved to 1.0% on PTV and to 1.5% on OARs, with a statistically significant difference. CONCLUSIONS: We determined the clinical impacts of dosimetric changes on a log file-based method using a general tolerance level and a tighter tolerance level for linac miscalibration and found that a tighter tolerance level significantly improved the accuracy of the log file-based method.

Patient-Specific Quality Assurance Using Monte Carlo Dose Calculation and Elekta Log Files for Prostate Volumetric-Modulated Arc Therapy.

Log file-based methods are attracting increasing interest owing to their ability to validate volumetric-modulated arc therapy outputs with high resolution in the leaf and gantry positions and in delivered dose. Cross-validation of these methods for comparison with measurement-based methods using the ionization chamber/ArcCHECK-3DVH software (version 3.2.0) under the same conditions of treatment anatomy and plan enables an efficient evaluation of this method. In this study, with the purpose of cross-validation, we evaluate the accuracy of a log file-based method using Elekta log files and an X-ray voxel Monte Carlo dose calculation technique in the case of leaf misalignment during prostate volumetric-modulated arc therapy. In this study, 10 prostate volumetric-modulated arc therapy plans were used. Systematic multileaf collimator leaf positional errors (±0.4 and ±0.8 mm for each single bank) were deliberately introduced into the optimized plans. Then, the delivered 3-dimensional doses to a phantom with a certain patient anatomy were estimated by our system. These doses were compared with the ionization chamber dose and the ArcCHECK-3DVH dose. For the given phantom and patient anatomy, the estimated dose strongly coincided with the ionization chamber/ArcCHECK-3DVH dose ( P < .01). In addition, good agreement between the estimated dose and the ionization chamber/ArcCHECK-3DVH dose was observed. The dose estimation accuracy of our system, which combines Elekta log files and X-ray voxel Monte Carlo dose calculation, was evaluated.

Quantification of residual dose estimation error on log file-based patient dose calculation.

PURPOSE: The log file-based patient dose estimation includes a residual dose estimation error caused by leaf miscalibration, which cannot be reflected on the estimated dose. The purpose of this study is to determine this residual dose estimation error. METHODS AND MATERIALS: Modified log files for seven head-and-neck and prostate volumetric modulated arc therapy (VMAT) plans simulating leaf miscalibration were generated by shifting both leaf banks (systematic leaf gap errors: ±2.0, ±1.0, and ±0.5mm in opposite directions and systematic leaf shifts: ±1.0mm in the same direction) using MATLAB-based (MathWorks, Natick, MA) in-house software. The generated modified and non-modified log files were imported back into the treatment planning system and recalculated. Subsequently, the generalized equivalent uniform dose (gEUD) was quantified for the definition of the planning target volume (PTV) and organs at risks. RESULTS: For MLC leaves calibrated within ±0.5mm, the quantified residual dose estimation errors that obtained from the slope of the linear regression of gEUD changes between non- and modified log file doses per leaf gap are in head-and-neck plans 1.32±0.27% and 0.82±0.17Gy for PTV and spinal cord, respectively, and in prostate plans 1.22±0.36%, 0.95±0.14Gy, and 0.45±0.08Gy for PTV, rectum, and bladder, respectively. CONCLUSIONS: In this work, we determine the residual dose estimation errors for VMAT delivery using the log file-based patient dose calculation according to the MLC calibration accuracy.

Blood flow changes of the anterior humeral circumflex artery decrease with the scapula in internal rotation.

PURPOSE: Poor posture is strongly related to scapular kinematics and shoulder diseases. The purpose of this study was to clarify the blood flow changes of the anterior and posterior humeral circumflex arteries in reference to the scapular position and to clarify anatomical features related to the changes in blood flow. METHODS: Cadaveric shoulders were dissected to elucidate the anatomical difference between the two arteries. The blood flow of these arteries in healthy volunteers (22-48 years, 113 shoulders) was evaluated by MRI with changing the position of the scapula (internal and external rotation). RESULTS: The anterior humeral circumflex artery was attached to the subscapularis tendon and located beneath the subdeltoid bursa. The posterior humeral circumflex artery was located between the deltoid muscle and the bursa, which allowed free movement during scapular motion. There were no direct extraosseous anastomoses between the two arteries. Compared with the posterior humeral circumflex artery, the anterior humeral circumflex artery decreased blood flow with the scapula in internal rotation. CONCLUSIONS: No direct extraosseous anastomoses between the two arteries were observed. Internal rotation of the scapula decreased blood flow of the anterior humeral circumflex artery, which might be related to various pathologies of the shoulder. LEVEL OF EVIDENCE: Diagnostic studies, Level III.

High-resolution magnetic resonance imaging of rotator cuff tears using a microscopy coil: noninvasive detection without intraarticular contrast material.

PURPOSE: The aim of this study was to evaluate the feasibility of high-resolution magnetic resonance imaging (MRI) using a microscopy coil for the diagnosis of rotator cuff tears by comparing the method to conventional MRI and MRI arthrography. MATERIALS AND METHODS: A total of 68 shoulders were prospectively studied using a 1.5-T MRI unit. Conventional MRI scans were obtained with a surface coil and high-resolution MRI scans with a microscopy coil. MRI arthrography was performed in 28 shoulders using a surface coil. MRI evaluation of tears of rotator cuff tendons was compared with arthroscopic findings and surgical results. RESULTS: The surgery revealed 40 full-thickness tears, 13 partial-thickness tears, and 15 intact cuffs. In all, 35 (88%) full-thickness tears were correctly diagnosed on conventional MRI and 40 (100%) on high-resolution MRI. MR arthrography delineated 11 of 12 (92%) full-thickness tears. Altogether, 5 (38%) of the partial-thickness tears were detected on conventional MRI, and 12 (92%) were clearly demonstrated on high-resolution MRI. MRI arthrography depicted three (60%) of five partial-thickness tears. High-resolution MRI showed higher sensitivity than conventional MRI (P < 0.05) and had values equivalent to those of MRI arthrography for diagnosing partial-thickness tears. CONCLUSION: High-resolution MRI with a microscopy coil is a feasible, noninvasive technique for diagnosing rotator cuff tears.

Assessment of coronary artery calcification in hemodialysis patients using multi-detector spiral CT scan.

Cardiovascular disease in association with coronary artery calcification (CAC) is the leading cause of death in patients with end-stage renal disease (ESRD). The evaluation of CAC has been performed by electron beam CT scan. The purpose of the present study was to assess CAC using multi-detector spiral CT (MDCT) and to evaluate contributors to CAC in these patients. Fifty-three patients on chronic hemodialysis participated in this study. Their mean age was 61.0+/-9.6 years, and the mean duration of dialysis therapy was 6.7+/-5.4 years. We used an automatic device to measure arterial pulse wave velocity (PWV) as an index of arterial wall stiffness. The aortic calcification index (ACI) was quantified morphometrically by CT scan. The CAC score correlated positively with ACI score (r =0.863, p <0.0001). Linear regression analysis indicated that the CAC scores correlated positively with age (r =0.406, p =0.0023), C-reactive protein (r =0.38, p =0.0047) and PWV (r =0.303, p =0.0271). Stepwise regression analysis indicated that ACI (beta-coefficient=0.862, p <0.0001) and arterial PWV (beta-coefficient=0.303, p <0.0001) were independently associated with CAC score. The mean CAC score of patients with cardiac events (2,568.5+/-2,575.1 mm3) was significantly higher than that (258.0+/-409.2 mm3) of patients without cardiac events. In conclusion, our results showed clearly that assessment of CAC score using MDCT may be predictive for detecting the presence of coronary artery disease. CAC is indirectly associated with increased arterial stiffness and the extent of aortic calcification in hemodialysis patients. We did not find a significant correlation between CAC score and parameters of mineral metabolism, including serum levels of calcium, phosphorus and parathyroid hormone. A longitudinal prospective study is required to assess the predictive value of this technique in determining cardiac events in large numbers of hemodialysis patients.

Effects of cyclic intermittent etidronate therapy on coronary artery calcification in patients receiving long-term hemodialysis.

BACKGROUND: Coronary artery calcification (CAC) is thought to be associated with greater cardiovascular mortality in patients with end-stage renal disease than in nonuremic persons. The purpose of the present study is to assess the effects of etidronate, a synthetic analogue of pyrophosphate, on progression of CAC score. METHODS: The extent of CAC was evaluated by using multidetector spiral computed tomography. Repeated CAC score estimation was possible in 35 patients (29 men, 6 women). Bone mineral density (BMD) was measured using dual-energy X-ray absorptiometry. Serum osteoprotegerin (OPG) was measured by using enzyme-linked immunoassay. Serum etidronate was measured by means of the gas spectrometry technique using deuterium-labeled etidronate as internal standard. RESULTS: Mean patient age was 63.2 +/- 8.2 (SD) years, and mean duration of dialysis therapy was 7.4 +/- 5.5 years. CAC score was estimated 3 times in each patient. After the second CAC score estimation, 35 patients were administered etidronate, 200 mg/d, for 14 days. This cycle was repeated 3 times every 90 days. CAC progression was significantly less pronounced during treatment with etidronate compared with the period before treatment was initiated. The median annualized absolute increase in calcified volume was 195.0 mm3 without treatment compared with -490.0 mm3 during treatment ( P < 0.01). Patients were divided into 2 groups based on changes in CAC score during etidronate treatment. Responders (n = 26) were patients whose CAC score decreased during therapy, and nonresponders (n = 9) were patients whose CAC score increased, even after etidronate therapy. Serum C-reactive protein values (0.18 +/- 0.13 mg/dL) in the responder group were greater than those (0.14 +/- 0.08 mg/dL) in the nonresponder group ( P = 0.013). Serum OPG levels decreased significantly during etidronate therapy (256.8 +/- 93.8 versus 245.0 +/- 83.0 pg/mL; P = 0.0161). Etidronate was well tolerated during the study. BMD values during etidronate therapy were not significantly changed from 0.941 +/- 0.125 to 0.968 +/- 0.246 g/cm2. CONCLUSION: Results of the present study suggest that the extent of CAC may be suppressed by etidronate in association with a reduction in chronic inflammatory responses. They also suggest that a decrease in serum OPG concentrations by means of etidronate may be associated with changes in vascular calcification in dialysis patients.

Adrenal adenomas: relationship between histologic lipid-rich cells and CT attenuation number.

OBJECTIVE: To evaluate the relationship between lipid-rich cells of the adrenal adenoma and precontrast computed tomographic (CT) attenuation numbers in three clinical groups. MATERIALS AND METHODS: Thirty-five surgically resected adrenal adenomas were used. The clinical diagnoses of the patients included 13 cases of primary aldosteronism, 15 cases of Cushing's syndrome, and 7 non-functioning tumors. The number of lipid-rich clear cells was counted using a microscopic eyepiece grid that contained 100 squares. The results were expressed as the percentages of lipid-rich areas. RESULTS: There was a strong inverse linear relationship between the percentage of lipid-rich cells and the precontrast CT attenuation number (R(2)=0.724, P<0.0001). There were significantly more lipid-rich cells in the primary aldosteronism and non-functioning tumor cases compared to cases of Cushing's syndrome (P=0.007 and 0.015, respectively). The CT attenuation numbers of the primary aldosteronism cases were significantly lower than those of Cushing's syndrome (P=0.0052). Furthermore, the CT attenuation numbers of the non-functioning tumor cases were lower than those of Cushing's syndrome cases. CONCLUSION: We showed that adrenal adenomas in primary aldosteronism and non-functioning tumors contain significantly more lipid-rich cells than those in Cushing's syndrome. They also showed significantly lower attenuation than that in Cushing's syndrome on CT scans. Our results suggest that precontrast CT attenuation numbers may be helpful in the differentiation of adenomas from non-adenomatous lesions, which include malignancies.

Time-related changes of radiofrequency ablation lesion in the normal rabbit liver: findings of magnetic resonance imaging and histopathology.

RATIONALE AND OBJECTIVES: To study time-related magnetic resonance imaging (MRI) and histopathological findings after radiofrequency ablation (RFA) in the normal liver. METHODS: Under ultrasound guidance, RFA was performed in 22 rabbits with a 15-gauge expandable probe inserted into the liver. The RF power was applied at 10 W initially with a 5-W increment every minute until RF power "roll-off" occurred. The rabbits subjected to postmortem MRI at 3 days (n = 6), 2 weeks (n = 6), 4 weeks (n = 6), and 12 weeks (n = 4) after RFA. Spin-echo (SE) T(1)-weighted image (WI) (TR/TE=500/12) and fast SE T2-WI (TR/TE= 3500/99) were obtained. The liver specimens were excised and processed for conventional histologic study. The MRI and pathologic findings were compared qualitatively and quantitatively. RESULTS: The lesion diameters measured with MRI and liver specimen were well correlated (P < 0.05). The coagulated regions presented 2 of 4 laminar patterns on T1/T2-WI 3 days after RFA, and 2 of 3 laminar patterns on T1/T2-WI 2 to 4 weeks after RFA. 12 weeks after RFA, the signal intensity in coagulated regions increased on T1-WI with a 2-laminar pattern still visible on T2-WI, the same as that of 2 to 4 weeks. The observed pathologic changes (enlarged sinusoids, marginal fibrous tissue and hepatocyte degeneration) could be responsible for the MRI laminar patterns and signal intensity changes. CONCLUSION: The time-related pathologic changes of RFA lesions in rabbit livers can be reflected as laminar patterns on MR images. Understanding of the zone structure of the lesions is useful for the evaluation of follow-up MRI.

[Computed tomography appearance of high-dose-irradiated hepatic parenchyma surrounding small tumors after stereotactic radiotherapy].

INTRODUCTION: The purpose of this study was to demonstrate the computed tomography (CT) appearance of high-dose-irradiated hepatic parenchyma surrounding small tumors after stereotactic radiotherapy (SRT). MATERIALS AND METHODS: Thirteen patients were enrolled between September 1999 and August 2001. We treated 15 lesions [six hepatocellular carcinomas (HCC) and nine liver metastases] with SRT using a linear accelerator. We used two SRT protocols of 45 Gy/3 fractions or 60 Gy/8 fractions. Hepatic CT examinations were performed at 3- to 4-month intervals after SRT. RESULTS: We classified the appearance of high-dose-irradiated hepatic parenchyma on CT into three types: Type 1: eight lesions (53.3%) showed low attenuation areas on non-contrast CT and high attenuation areas on contrast CT; Type 2: five lesions (33.3%) showed low attenuation areas on non-contrast and contrast CT; Type 3: two lesions (13.3%) showed no change. The CT appearance of Type 1 after SRT was similar to that of pre-irradiated HCC, but the other types were easily differentiated from the primary tumors. These changes were observed and continued for 6 to 22 months after treatment. In two cases of Type 1, we performed histological examinations of the regions of high attenuation on contrast CT. These specimens revealed only radiation-induced hepatic injury. CONCLUSION: This study showed three types of CT appearance in high-dose-irradiated hepatic parenchyma. We suggest that other examinations be considered to distinguish between radiation hepatitis and local recurrences for HCC after SRT.

Two cases of adrenal hemangioma: CT and MRI findings with pathological correlations.

We present two cases of adrenal hemangioma, a rare non-functioning adrenal tumor, and correlate the CT and MR imaging findings with the pathological findings. Peripheral spotty contrast enhancement with centripetal enhancement was noted in one case. This pattern of enhancement is crucial for diagnosing adrenal hemangioma. The other case showed only thin rim enhancement without centripetal enhancement. This finding can be seen in other adrenal tumors, making diagnosis difficult. Marked hyperintensity on T2-weighted images and focal hyperintensity on T1-weighted images that showed focal hemorrhage and calcification were noted. These findings can also be seen in other adrenal tumors and are not pathognomonic of adrenal hemangioma. However, observing the combination of these findings can lead to the correct diagnosis.