A clinical evaluation study of cardiothoracic ratio measurement using artificial intelligence.

BACKGROUND: Artificial intelligence, particularly the deep learning (DL) model, can provide reliable results for automated cardiothoracic ratio (CTR) measurement on chest X-ray (CXR) images. In everyday clinical use, however, this technology is usually implemented in a non-automated (AI-assisted) capacity because it still requires approval from radiologists. We investigated the performance and efficiency of our recently proposed models for the AI-assisted method intended for clinical practice. METHODS: We validated four proposed DL models (AlbuNet, SegNet, VGG-11, and VGG-16) to find the best model for clinical implementation using a dataset of 7517 CXR images from manual operations. These models were investigated in single-model and combined-model modes to find the model with the highest percentage of results where the user could accept the results without further interaction (excellent grade), and with measurement variation within ± 1.8% of the human-operating range. The best model from the validation study was then tested on an evaluation dataset of 9386 CXR images using the AI-assisted method with two radiologists to measure the yield of excellent grade results, observer variation, and operating time. A Bland-Altman plot with coefficient of variation (CV) was employed to evaluate agreement between measurements. RESULTS: The VGG-16 gave the highest excellent grade result (68.9%) of any single-model mode with a CV comparable to manual operation (2.12% vs 2.13%). No DL model produced a failure-grade result. The combined-model mode of AlbuNet + VGG-11 model yielded excellent grades in 82.7% of images and a CV of 1.36%. Using the evaluation dataset, the AlbuNet + VGG-11 model produced excellent grade results in 77.8% of images, a CV of 1.55%, and reduced CTR measurement time by almost ten-fold (1.07 ± 2.62 s vs 10.6 ± 1.5 s) compared with manual operation. CONCLUSION: Due to its excellent accuracy and speed, the AlbuNet + VGG-11 model could be clinically implemented to assist radiologists with CTR measurement.

Cardiothoracic ratio measurement using artificial intelligence: observer and method validation studies.

BACKGROUND: Artificial Intelligence (AI) is a promising tool for cardiothoracic ratio (CTR) measurement that has been technically validated but not clinically evaluated on a large dataset. We observed and validated AI and manual methods for CTR measurement using a large dataset and investigated the clinical utility of the AI method. METHODS: Five thousand normal chest x-rays and 2,517 images with cardiomegaly and CTR values, were analyzed using manual, AI-assisted, and AI-only methods. AI-only methods obtained CTR values from a VGG-16 U-Net model. An in-house software was used to aid the manual and AI-assisted measurements and to record operating time. Intra and inter-observer experiments were performed on manual and AI-assisted methods and the averages were used in a method variation study. AI outcomes were graded in the AI-assisted method as excellent (accepted by both users independently), good (required adjustment), and poor (failed outcome). Bland-Altman plot with coefficient of variation (CV), and coefficient of determination (R-squared) were used to evaluate agreement and correlation between measurements. Finally, the performance of a cardiomegaly classification test was evaluated using a CTR cutoff at the standard (0.5), optimum, and maximum sensitivity. RESULTS: Manual CTR measurements on cardiomegaly data were comparable to previous radiologist reports (CV of 2.13% vs 2.04%). The observer and method variations from the AI-only method were about three times higher than from the manual method (CV of 5.78% vs 2.13%). AI assistance resulted in 40% excellent, 56% good, and 4% poor grading. AI assistance significantly improved agreement on inter-observer measurement compared to manual methods (CV; bias: 1.72%; - 0.61% vs 2.13%; - 1.62%) and was faster to perform (2.2 ± 2.4 secs vs 10.6 ± 1.5 secs). The R-squared and classification-test were not reliable indicators to verify that the AI-only method could replace manual operation. CONCLUSIONS: AI alone is not yet suitable to replace manual operations due to its high variation, but it is useful to assist the radiologist because it can reduce observer variation and operation time. Agreement of measurement should be used to compare AI and manual methods, rather than R-square or classification performance tests.

Evaluation of pulmonary metastases in children by non-contrast chest computed tomography.

BACKGROUND: Most of the metastatic lung lesions are relatively high contrast in comparison to the lung background and easily detected in non-contrast enhancement chest computed tomography alone (NECCT). Pediatric patients may get benefit from its minimal radiation dose and lack of adverse reaction from iodinated contrast agent. OBJECTIVE: To compare effectiveness of non-contrast enhancement chest computed tomography (NECCT) in detecting thoracic metastasis with full protocol chest computed tomography (FPCCT) (chest computed tomography with and without contrast) in non-hematologic extrathoracic malignancy in children. MATERIAL AND METHOD: Both NECCT and FPCCT were evaluated in 50 pediatric patients with non-hematologic extrathoracic malignancy retrospectively. Lung nodules, ground glass opacities, interlobular septal thickening, pleural effusion, pleural thickening, pericardial effusion, endobronchial lesion, and intravascular metastasis were evaluated separately on each CT protocol by two radiologists. RESULTS: Thirty boys and 20 girls were included in the present study (mean age = 10 years and 3 months). The lesions include nodule (333 detected by NECCT (median = 3), 336 detected by CECCT (median = 3)), ground glass opacity (12 detected by NECCT (median = 0), 15 detected by CECCT (median = 0)), interlobular septal thickening (12 detected by NECCT (median = 0), 11 detected by CECCT (median = 0)). There was 100 percent match of calcified nodules (n = 36), pleural effusion (n = 1), pleural thickening (n = 3), intravascular thrombus (n = 2), and mediastinal lymph node (n = 1) between NECCT and FPCCT studies. There was no statistically significant different in capability of demonstrating all lesions between NECCT and FPCCT. Most of the discrepancies between NECCT and FPCCT were from motion artifact, inadequate inspiration, and radiologist's opinion rather than effect of contrast agent administration itself CONCLUSION: NECCT is as effective as FPCCT in evaluation of pulmonary metastasis in non-hematologic extrathoracic malignancies. For evaluation of lung metastases in this population, NECCT alone is sufficient.

Vertebral body compression fracture: discriminating benign from malignant causes by diffusion-weighted MR imaging and apparent diffusion coefficient value.

OBJECTIVE: To evaluate the diagnostic performance of apparent diffusion coefficient (ADC) value in discriminating benign from malignant vertebral compression fracture. MATERIAL AND METHOD: 22 symptomatic patients with compression fracture of vertebra referred for conventional MRI spines during January 2009-March 2010 underwent additional diffusion weighted MR techniques. Evaluation of diffusion weighted MR imaging and quantified ADC value from reconstructed ADC map were performed. The accuracy, sensitivity, specificity, positive predictive value and negative predictive value of apparent diffusion coefficient (ADC) value were calculated. RESULTS: A total of 39 vertebral fractures; 7 malignant compression fractures and 32 benign compression fractures were evaluated. The difference between ADC values of malignant, benign compression fracture and normal vertebrae were statistically significant (p < 0.0001). The accuracy, sensitivity and specificity were 89.7%, 85.7% and 90.6% respectively with the ADC threshold of 0.89 to discriminate malignancy. CONCLUSION: The ADC promises to be an effective implement for characterization of vertebral body compression fracture in differentiating benign and malignant compression fractures.

C-reactive protein as a monitor of chemotherapy response in advanced non-small cell lung cancer (CML study).

OBJECTIVE: The aim of the present prospective study was to evaluate the correlation between the change of serum c-reactive protein (CRP) levels and response to chemotherapy in patients with locally advanced or metastatic non-small cell lung cancer. MATERIAL AND METHOD: Patients with locally advanced or metastatic non-small cell lung cancer who received the first line chemotherapy were measured serum CRP levels prior to treatment. Chemotherapy regimen was given to patients according to physicians and radiologic imaging was evaluated after two or three cycles of treatment. Serum CRP levels were measured first time at pre-treatment and second time in patients who had pre-treatment serum CRP levels greater than normal range (3 mg/l) at the time of response assessment or clinical progression. The primary endpoint was the correlation between change of serum CRP levels and radiologic response. The secondary endpoint was the prevalence of elevated CRP levels in advanced NSCLC patients and correlation between initial CRP levels and progression free survival (PFS). RESULTS: Fifty four patients were enrolled. Prevalence of elevated CRP levels in advanced NSCLC was 76%. Thirty patients had serial serum CRP measured. There was correlation between change in serum CRP levels and response to treatment (r = 0.43, p = 0.018, spearman rank). There was significant correlation between response to treatment and decrease in CRP levels greater than 50% (p = 0.009, Fisher's exact test). In contrast there was no correlation between progression and increase in CRP levels (p = 0.640, Fisher's exact test). All patients with serial CRP levels decreased to normal range (< 3 mg/ l) had response to chemotherapy. High pre-treatment CRP levels (> 100 mg/l) correlated with poor PFS. Median PFS for patients with pre-treatment CRP levels of 3-30 mg/l, 30-100 mg/l and >100 mg/l was 23.0 weeks, 13.0 weeks and 6.3 weeks, respectively. Patients with serial CRP levels less than 3 mg/l had greater PFS than patients with serial CRP levels higher than 3 mg/l (p = 0.026, log rank test). CONCLUSION: The present study suggested that high levels of pre-treatment serum CRP and persistent CRP in serum was a poor prognostic factor. The decrease in CRP levels greater than 50% was a simple method to predict the response to treatment in patients with locally advanced or metastatic non-small lung cancer.

Reliability study of white matter rating scale for the dementia and disability in Thai Elderly Project.

OBJECTIVE: To study the reliability of white matter rating scale on magnetic resonance imaging (MRI) of elderly patients in the urban community of Bangkok. MATERIAL AND METHOD: One hundred elderly with clinical diagnosis of cognitive impairment in the urban community around Siriraj Hospital underwent cranial MRI according to the Dementia and Disability in Thai Elderly Project. The axial T1wi, T2wi, and fluid attenuated inversion recovery (FLAIR) were separately assessed by two neuroradiologists. The assessment included white matter change by using Scheltens' rating scale, atrophy, and evidence of infarction. The inter-rater agreements were analyzed. RESULTS: The inter-rater agreement of periventricular hyperintensities, white matter, basal ganglion and infratentorial foci of hyperintensities were very good (ICC = 0.89-0.98). The agreement was good for central atrophy (Kw = 0.66) and moderate for cortical atrophy (Kw = 0.49). The silent infarction was found in the study population and divided into cortical (15%), subcortical (26%), brainstem (3%), and infratentorial infarction (8%). CONCLUSION: White matter hyperintensities was an important radiological criteria for diagnosis of vascular dementia. Appropriate rating scale is necessary especially in research study. The authors found that Scheltens rating scale needed some training and slightly time consuming at the beginning but was a good reliable tool.