Erythropoiesis and estimated fluid volume regulation following initiation of ipragliflozin treatment in patients with type 2 diabetes mellitus and chronic kidney disease: A post-hoc analysis of the PROCEED trial.

AIMS: To analyse the changes in erythropoietic and estimated fluid volume parameters after the initiation of ipragliflozin, a sodium-glucose co-transporter 2 inhibitor, in patients with type 2 diabetes mellitus (T2DM) and chronic kidney disease (CKD). METHODS: This was a post-hoc analysis of the PROCEED trial, which evaluated the effect of 24-week ipragliflozin treatment on endothelial dysfunction in patients with T2DM and CKD. We evaluated the changes in erythropoietic and estimated fluid volume parameters from baseline to 24 weeks post-treatment in 53 patients who received ipragliflozin (ipragliflozin group) and 55 patients with T2DM and CKD without sodium-glucose co-transporter 2 inhibitors (control group), a full analysis set of the PROCEED trial. RESULTS: The increases in haemoglobin [estimated group difference, 0.5 g/dl; 95% confidence interval (CI), 0.3-0.8; p < .001], haematocrit (estimated group difference, 2.2%; 95% CI, 1.3-3.1; p < .001) and erythropoietin (estimated log-transformed group difference, 0.1; 95% CI, 0.01-0.3; p = .036) were significantly greater in the ipragliflozin group than those in the control group. Ipragliflozin treatment was significantly associated with an increase in erythropoietin, independent of the corresponding change in haemoglobin (ß = 0.253, p < .001) or haematocrit (ß = 0.278, p < .001). Reductions in estimated plasma volume (estimated group difference, -7.94%; 95% CI, -11.6 to -4.26%; p < .001) and estimated extracellular volume (estimated group difference, -181.6 ml; 95% CI, -275.7 to -87.48 ml; p < .001) were significantly greater in the ipragliflozin group than those in the control group. CONCLUSIONS: Erythropoiesis was enhanced and estimated fluid volumes were reduced by ipragliflozin in patients with T2DM and CKD. CLINICAL TRIAL: PROCEED trial (registration number: jRCTs071190054).

Hypothesis testing procedure for binary and multi-class F1 -scores in the paired design.

In modern medicine, medical tests are used for various purposes including diagnosis, disease screening, prognosis, and risk prediction. To quantify the performance of the binary medical test, we often use sensitivity, specificity, and negative and positive predictive values as measures. Additionally, the F 1 $$ {F}_1 $$ -score, which is defined as the harmonic mean of precision (positive predictive value) and recall (sensitivity), has come to be used in the medical field due to its favorable characteristics. The F 1 $$ {F}_1 $$ -score has been extended for multi-class classification, and two types of F 1 $$ {F}_1 $$ -scores have been proposed for multi-class classification: a micro-averaged F 1 $$ {F}_1 $$ -score and a macro-averaged F 1 $$ {F}_1 $$ -score. The micro-averaged F 1 $$ {F}_1 $$ -score pools per-sample classifications across classes and then calculates the overall F 1 $$ {F}_1 $$ -score, whereas the macro-averaged F 1 $$ {F}_1 $$ -score computes an arithmetic mean of the F 1 $$ {F}_1 $$ -scores for each class. Additionally, Sokolova and Lapalme 1 $$ {}^1 $$ gave an alternative definition of the macro-averaged F 1 $$ {F}_1 $$ -score as the harmonic mean of the arithmetic means of the precision and recall over classes. Although some statistical methods of inference for binary and multi-class F 1 $$ {F}_1 $$ -scores have been proposed, the methodology development of hypothesis testing procedure for them has not been fully progressing yet. Therefore, we aim to develop hypothesis testing procedure for comparing two F 1 $$ {F}_1 $$ -scores in paired study design based on the large sample multivariate central limit theorem.

Deep learning model with collage images for the segmentation of dedicated breast positron emission tomography images.

BACKGROUND: Dedicated breast positron emission tomography (dbPET) has high contrast and resolution optimized for detecting small breast cancers, leading to its noisy characteristics. This study evaluated the application of deep learning to the automatic segmentation of abnormal uptakes on dbPET to facilitate the assessment of lesions. To address data scarcity in model training, we used collage images composed of cropped abnormal uptakes and normal breasts for data augmentation. METHODS: This retrospective study included 1598 examinations between April 2015 and August 2020. A U-Net-based model with an uptake shape classification head was trained using either the original or augmented dataset comprising collage images. The Dice score, which measures the pixel-wise agreement between a prediction and its ground truth, of the models was compared using the Wilcoxon signed-rank test. Moreover, the classification accuracies were evaluated. RESULTS: After applying the exclusion criteria, 662 breasts were included; among these, 217 breasts had abnormal uptakes (mean age: 58 ± 14 years). Abnormal uptakes on the cranio-caudal and mediolateral maximum intensity projection images of 217 breasts were annotated and labeled as focus, mass, or non-mass. The inclusion of collage images into the original dataset yielded a Dice score of 0.884 and classification accuracy of 91.5%. Improvement in the Dice score was observed across all subgroups, and the score of images without breast cancer improved significantly from 0.750 to 0.834 (effect size: 0.76, P = 0.02). CONCLUSIONS: Deep learning can be applied for the automatic segmentation of dbPET, and collage images can improve model performance.

Correlation between recurrence-free survival and overall survival after upfront surgery for resected colorectal liver metastases.

BACKGROUND: The role of recurrence-free survival (RFS) as a valid surrogate endpoint for overall survival (OS) in patients who underwent upfront surgery for colorectal liver metastases remains uncertain. The aim of the study was to compare the two survival measures in a nationwide cohort of upfront resected colorectal liver metastasis. METHODS: Data from patients with colorectal liver metastases without extrahepatic metastases who underwent curative surgery for liver metastases were retrieved from the Japanese nationwide database (data collection 2005-2007 and 2013-2014). RFS, OS, and survival after recurrence were estimated using the Kaplan-Meier method. The correlation (ρ) between RFS and OS was assessed using the rank correlation method combined with iterative multiple imputation, to account for censoring. As a secondary analysis, the correlation was evaluated according to adjuvant chemotherapy regimen. In sensitivity analysis, the pairwise correlation between RFS and OS was calculated. RESULTS: A total of 2385 patients with colorectal liver metastases were included. In the primary analysis, there was a moderately strong correlation between RFS and OS (ρ = 0.73, 95 per cent c.i. 0.70 to 0.76). The strength of the correlation was similar regardless of the adjuvant treatment regimen (oxaliplatin plus 5-fluorouracil: ρ = 0.72, 0.67 to 0.77; 5-fluorouracil alone: ρ = 0.72, 0.66 to 0.76; observation: ρ = 0.74, 0.69 to 0.78). The mean(s.d.) pairwise correlation coefficient between 3-year RFS and 5-year OS was 0.87(0.06). CONCLUSION: In surgically treated patients with colorectal liver metastases, there was a moderately strong correlation between RFS and OS, which was unaffected by the treatment regimen. Further validation using a trial-level analysis is required.

Upper gastrointestinal endoscopic findings in functional constipation and irritable bowel syndrome diagnosed using the Rome IV criteria: a cross-sectional survey during a medical check-up in Japan.

BACKGROUND: The Rome IV criteria have been established as an international standard for diagnosing disorders of gut-brain interaction. In this study, we aimed to examine the upper gastrointestinal (GI) endoscopic findings and symptoms of subjects with functional constipation (FC) and irritable bowel syndrome (IBS) of individuals undergoing a medical check-up. METHODS: A total of 13,729 subjects underwent a medical check-up at Osaka City University-affiliated clinic, MedCity21, between April 2018 and March 2019. Among the 5,840 subjects who underwent screening upper GI endoscopy and completed a questionnaire based on the Rome IV criteria, 5,402 subjects were consecutively enrolled after excluding subjects with a large amount of gastric residue (n = 6), those who had previously undergone partial or total gastrectomy (n = 40), or those with daily use of low-dose aspirin (n = 82), nonsteroidal anti-inflammatory drugs (n = 63), or acid secretion inhibitors (n = 308). RESULTS: Robust Poisson regression analyses adjusted for age, sex, Helicobacter pylori infection status, alcohol intake, and smoking habits showed a significant association between FC and corpus erosion (adjusted prevalence ratio [aPR], 2.93; 95% confidence interval [CI], 1.51-5.67; p < 0.01) and red streaks (aPR, 3.83; 95% CI, 2.53-5.79; p < 0.01), whereas IBS was significantly associated with erosive gastritis (aPR, 8.46; 95% CI, 4.89-14.67; p < 0.01) and duodenitis (aPR, 7.28; 95% CI, 3.64-14.59; p < 0.01). Red streaks tended to be associated with IBS (aPR, 1.96; 95% CI, 1.00-3.83; p = 0.05). Subjects with IBS were the most to complain of both upper and lower GI symptoms and psychological symptoms, followed by those with FC and controls. IBS subjects with erosive gastritis or duodenitis had significantly more complaints of stomachache and feeling stressed than those without erosive gastritis or duodenitis (54.5% vs. 18.8%; p = 0.03 and 66.7% vs. 25.0%; p = 0.01). CONCLUSIONS: Subjects with FC and IBS had a variety of upper GI and psychological symptoms. In the upper GI endoscopic findings, corpus erosion and red streaks were associated with FC, and erosive gastritis, duodenitis, and possibly red streaks were associated with IBS.

Plasma renin activity variation following admission predicts patient outcome in acute decompensated heart failure with reduced and mildly reduced ejection fraction.

Plasma renin activity (PRA) level at admission is reported to be a prognostic predictor of acute decompensated heart failure (ADHF) patients. Although PRA is affected during hospitalization by several factors including fluid volume and drug titration, whether the changes in PRA levels during hospitalization (ΔPRA) are associated with prognosis of ADHF patients are largely unknown. PURPOSE: Investigate the predictive impact of ΔPRA on the prognosis of ADHF patients with reduced ejection fraction (HFrEF) and mildly reduced ejection fraction (HFmrEF). METHODS: Retrospectively analyzed consecutive 116 HFrEF and HFmrEF patients admitted for ADHF. PRA measurements were acquired at admission and at discharge. The primary outcome was a composite of cardiovascular death and HF re-hospitalization. RESULTS: Out of 116 patients, 85 had PRA measurements both at admission and at discharge. Compared to admission, PRA level was significantly higher at discharge (0.8 (IQR 0.3-2.2) to 2.8 (IQR 1.0-7.2), p < 0.001). Tertiary groups ranked by PRA level on admission showed trend of poor prognosis in order of high, mid, and low PRA level (p = 0.07). On the contrary, PRA level at discharge significantly differentiated the prognosis and was poor in order of high, low, and mid (p = 0.026). Next, when the participants were divided into tertiary groups ranked by ΔPRA, prognosis worsened in the order of "minimal", "decreasing", and the "increasing" tier. Cubic splines analysis also indicate a similar tendency. CONCLUSIONS: In ADHF patients with HFrEF and HFmrEF, patients with minimal ΔPRA showed the better prognosis over the those with either increasing or decreasing.

Optimal threshold of a control parameter for tomotherapy respiratory tracking: A phantom study.

BACKGROUND: Radixact Synchrony® , a real-time motion tracking and compensating modality, is used for helical tomotherapy. Control parameters are used for the accurate application of irradiation. Radixact Synchrony® uses the potential difference, which is an index of the accuracy of the prediction model of target motion and is represented by a statistical prediction of the 3D distance error. Although there are several reports on Radixact Synchrony® , few have reported the appropriate settings of the potential difference threshold. PURPOSE: This study aims to determine the optimal threshold of the potential difference of Radixact Synchrony® during respiratory tumor-motion-tracking irradiation. METHODS: The relationship among the dosimetric accuracy, motion tracking accuracy, and control parameter was evaluated using a moving platform, a phantom with a basic respiratory model (the fourth power of a sinusoidal wave), and several irregular respiratory model waveforms. The dosimetric accuracy was evaluated by gamma analysis (3%, 1 mm, 10% dose threshold). The tracking accuracy was measured by the distance error of the difference between the tracked and driven positions of the phantom. The largest potential difference for 95% of treatment time was evaluated, and its correlation with the gamma-pass ratio and distance error was investigated. The optimal threshold of the potential difference was determined by receiver operating characteristic (ROC) analysis. RESULTS: A linear correlation was identified between the potential difference and the gamma-pass ratio (R = -0.704). A linear correlation was also identified between the potential difference and distance error (R = 0.827). However, as the potential difference increased, it tended to underestimate the distance error. The ROC analysis revealed that the appropriate cutoff value of the potential difference was 3.05 mm. CONCLUSION: The irradiation accuracy with motion tracking by Radixact Synchrony® could be predicted from the potential difference, and the threshold of the potential difference should be set to ∼3 mm.

Proposal to Improve the Image Quality of Short-Acquisition Time-Dedicated Breast Positron Emission Tomography Using the Pix2pix Generative Adversarial Network.

This study aimed to evaluate the ability of the pix2pix generative adversarial network (GAN) to improve the image quality of low-count dedicated breast positron emission tomography (dbPET). Pairs of full- and low-count dbPET images were collected from 49 breasts. An image synthesis model was constructed using pix2pix GAN for each acquisition time with training (3776 pairs from 16 breasts) and validation data (1652 pairs from 7 breasts). Test data included dbPET images synthesized by our model from 26 breasts with short acquisition times. Two breast radiologists visually compared the overall image quality of the original and synthesized images derived from the short-acquisition time data (scores of 1−5). Further quantitative evaluation was performed using a peak signal-to-noise ratio (PSNR) and structural similarity (SSIM). In the visual evaluation, both readers revealed an average score of >3 for all images. The quantitative evaluation revealed significantly higher SSIM (p < 0.01) and PSNR (p < 0.01) for 26 s synthetic images and higher PSNR for 52 s images (p < 0.01) than for the original images. Our model improved the quality of low-count time dbPET synthetic images, with a more significant effect on images with lower counts.

Incremental medical cost of delirium in elderly patients with cognitive impairment: analysis of a nationwide administrative database in Japan.

OBJECTIVES: Delirium is a neuropsychiatric disorder that commonly occurs in elderly patients with cognitive impairment. The economic burden of delirium in Japan has not been well characterised. In this study, we assessed incremental medical costs of delirium in hospitalised elderly Japanese patients with cognitive impairment. DESIGN: Retrospective, cross-sectional, observational study. SETTING: Administrative data collected from acute care hospitals in Japan between April 2012 and September 2020. PARTICIPANTS: Hospitalised patients ≥65 years old with cognitive impairment were categorised into groups-with and without delirium. Delirium was identified using a delirium identification algorithm based on the International Classification of Diseases 10th Revision codes or antipsychotic prescriptions. OUTCOME MEASURES: Total medical costs during hospitalisation were compared between the groups using a generalised linear model. RESULTS: The study identified 297 600 hospitalised patients ≥65 years of age with cognitive impairment: 39 836 had delirium and 257 764 did not. Patient characteristics such as age, sex, inpatient department and comorbidities were similar between groups. Mean (SD) unadjusted total medical cost during hospitalisation was 979 907.7 (871 366.4) yen for patients with delirium and 816 137.0 (794 745.9) yen for patients without delirium. Adjusted total medical cost was significantly greater for patients with delirium compared with those without delirium (cost ratio=1.09, 95% CI: 1.09 to 1.10; p<0.001). Subgroup analyses revealed significantly higher total medical costs for patients with delirium compared with those without delirium in most subgroups except patients with hemiplegia or paraplegia. CONCLUSIONS: Medical costs during hospitalisation were significantly higher for patients with delirium compared with those without delirium in elderly Japanese patients with cognitive impairment, regardless of patient subgroups such as age, sex, intensive care unit admission and most comorbidities. These findings suggest that delirium prevention strategies are critical to reducing the economic burden as well as psychological/physiological burden in cognitively impaired elderly patients in Japan.

Demographic and clinical characteristics of patients with delirium: analysis of a nationwide Japanese medical database.

OBJECTIVES: Delirium commonly occurs during hospitalisation and is associated with increased mortality, especially in elderly patients. This study aimed to determine the demographic and clinical characteristics of patients with delirium in the Japanese real-world clinical setting using a nationwide database comprising claims and discharge abstract data. DESIGN: This was an observational, cross-sectional, retrospective study in hospitalised patients with an incident delirium identified by a diagnosis based on International Classification of Diseases, 10th Revision codes or initiating antipsychotics recommended for delirium treatment in Japan during their hospitalisation. SETTING: Patients from the Medical Data Vision database including more than 400 acute care hospitals in Japan were evaluated from admission to discharge. PARTICIPANTS: Of the 32 910 227 patients who were included in the database between April 2012 and September 2020, a total of 145 219 patients met the criteria for delirium. PRIMARY AND SECONDARY OUTCOME MEASURES: Demographic and baseline characteristics, comorbidities, clinical profiles and pharmacological treatments were evaluated in patients with delirium. RESULTS: The mean (SD) patient age was 76.5 (13.8) years. More than half of the patients (n=82 159; 56.6%) were male. The most frequent comorbidities were circulatory system diseases, observed in 81 954 (56.4%) patients. Potentially inappropriate medications (PIMs) with risk of delirium including benzodiazepines and opioids were prescribed to 76 798 (52.9%) patients. Approximately three-fourths of these patients (56 949; 74.2%) were prescribed ≥4 PIMs. The most prescribed treatment for delirium was injectable haloperidol (n=82 490; 56.8%). Mean (SD) length of hospitalisation was 16.0 (12.1) days. CONCLUSIONS: The study results provide comprehensive details of the clinical characteristics of patients with delirium and treatment patterns with antipsychotics in the Japanese acute care setting. In this patient population, the prescription rate of injectable haloperidol and PIMs was high, suggesting the need for improved understanding among healthcare providers about the appropriate management of delirium, which may benefit patients.

Effect of Music in Reducing Pain during Hemodialysis Access Cannulation: A Crossover Randomized Controlled Trial.

BACKGROUND AND OBJECTIVES: Pain during cannulation for vascular access is a considerable problem for patients with kidney disease who are undergoing hemodialysis. We examined whether listening to music can reduce cannulation pain in these patients. DESIGN, SETTING, PARTICIPANTS, & MEASUREMENTS: We conducted a multicenter, single-blind, crossover, randomized trial of 121 patients who reported pain during cannulation for hemodialysis. We compared participants listening to "Sonata for Two Pianos in D Major, K.448" or white noise as control while undergoing the cannulation procedure. The cannulation operator was blinded to the intervention, and the hypothesized superiority of music over white noise was concealed during explanations to the participants. The primary end point was the visual analog scale score for cannulation pain independently evaluated by participants. RESULTS: The primary analysis was on the basis of the modified intention-to-treat principle. The median baseline visual analog scale pain score was 24.7 mm (interquartile range, 16.5-42.3). Median change of the visual analog scale pain score from the "no sound" to the music period was -2.7 mm (interquartile range, -9.2 to 3.6), whereas it was -0.3 mm (interquartile range, -5.8 to 4.5) from "no sound" to white noise. The visual analog scale pain score decreased when listening to music compared with white noise. (Adjusted difference of visual analog scale pain score: -12%; 95% confidence interval, -21 to -2; P=0.02.) There were no significant differences in the secondary outcomes of anxiety, BP, or stress assessed by salivary amylase (adjusted difference of visual analog scale anxiety score -8%, 95% confidence interval, -18 to 4; P=0.17). No intervention-related adverse events were reported. CONCLUSIONS: Listening to music reduced cannulation pain in patients on hemodialysis, although there was no significant effect on anxiety, BP, or stress markers.

Possible role of insulin resistance in activation of plasma xanthine oxidoreductase in health check-up examinees.

We previously found an association of insulin resistance (IR) with plasma xanthine oxidoreductase (XOR) activity in a cross-sectional study. However, whether IR induces increased XOR activity has not been elucidated. This retrospective longitudinal observational study included 347 participants (173 males, 174 females) who underwent annual health examinations and were medication naïve. Homeostasis model assessment of IR (HOMA-IR) index, and physical and laboratory measurements were determined at the baseline. At baseline and 12-month follow-up examinations, plasma XOR activity was determined using our novel assay based on [13C2,15N2] xanthine and liquid chromatography/triple quadrupole mass spectrometry. Subjects with IR, defined as HOMA-IR index ≥ 1.7 (n = 92), exhibited significantly (p < 0.001) higher plasma XOR activity levels than those without IR (n = 255), with an increase in that activity seen in 180 (51.9%) after 12 months. Multivariable linear and logistic regression analyses showed that IR, but not BMI or waist circumference, at baseline was significantly associated with plasma XOR activity (ß = 0.094, p = 0.033) and increased plasma XOR activity over the 12-month period (odds ratio, 1.986; 95% confidence interval, 1.048-3.761; p = 0.035), after adjustments for various clinical parameters, including plasma XOR activity at baseline. These results suggest that IR induces increased plasma XOR activity in a manner independent of adiposity.

A questionnaire-based survey on the etiopathogenesis of chronic constipation during a medical check-up in Japan.

The study group of the Japanese Society of Gastroenterology released evidence-based clinical practice guidelines for chronic constipation (CC) in 2017, and irritable bowel syndrome (IBS) was treated as one of the causes of CC. We examined the differences in characteristics between IBS and non-IBS subjects with CC who underwent a medical check-up in Japan. A total of 10,658 subjects participated in this study, and we focused on 467 subjects who fulfilled the diagnostic criteria of CC using a questionnaire survey. The number of IBS subjects was 21, and they had sleep disorders, were more symptomatic (e.g., abdominal pain, abdominal bloating/distension, feeling stressed, annoyance, lack of motivation, fatigue upon waking, and feeling depressed), and had more episodes of sensation of incomplete evacuation and anorectal obstruction/blockage during defecation than non-IBS subjects. Furthermore, stool frequency of IBS subjects was significantly different from non-IBS subjects. Multivariate ordinal logistic regression analysis revealed that the factors associated with a higher stool frequency were IBS [odds ratio (OR), 2.46; 95% confidence interval (CI), 1.00-6.05; p = 0.049], male sex (OR, 1.97; 95% CI, 1.20-3.23; p = 0.007), and regular exercise (OR, 1.80; 95% CI, 1.05-3.07; p = 0.033). These findings suggest that IBS has unique characteristics in subjects with CC.

Confidence interval for micro-averaged F 1 and macro-averaged F 1 scores.

A binary classification problem is common in medical field, and we often use sensitivity, specificity, accuracy, negative and positive predictive values as measures of performance of a binary predictor. In computer science, a classifier is usually evaluated with precision (positive predictive value) and recall (sensitivity). As a single summary measure of a classifier's performance, F 1 score, defined as the harmonic mean of precision and recall, is widely used in the context of information retrieval and information extraction evaluation since it possesses favorable characteristics, especially when the prevalence is low. Some statistical methods for inference have been developed for the F 1 score in binary classification problems; however, they have not been extended to the problem of multi-class classification. There are three types of F 1 scores, and statistical properties of these F 1 scores have hardly ever been discussed. We propose methods based on the large sample multivariate central limit theorem for estimating F 1 scores with confidence intervals.

Deep learning method with a convolutional neural network for image classification of normal and metastatic axillary lymph nodes on breast ultrasonography.

PURPOSE: To investigate the ability of deep learning (DL) using convolutional neural networks (CNNs) for distinguishing between normal and metastatic axillary lymph nodes on ultrasound images by comparing the diagnostic performance of radiologists. MATERIALS AND METHODS: We retrospectively gathered 300 images of normal and 328 images of axillary lymph nodes with breast cancer metastases for training. A DL model using the CNN architecture Xception was developed to analyze test data of 50 normal and 50 metastatic lymph nodes. A board-certified radiologist with 12 years' experience. (Reader 1) and two residents with 3- and 1-year experience (Readers 2, 3), respectively, scored these test data with and without the assistance of the DL system for the possibility of metastasis. The sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) were calculated. RESULTS: Our DL model had a sensitivity of 94%, a specificity of 88%, and an AUC of 0.966. The AUC of the DL model was not significantly different from that of Reader 1 (0.969; p = 0.881) and higher than that of Reader 2 (0.913; p = 0.101) and Reader 3 (0.810; p < 0.001). With the DL support, the AUCs of Readers 2 and 3 increased to 0.960 and 0.937, respectively, which were comparable to those of Reader 1 (p = 0.138 and 0.700, respectively). CONCLUSION: Our DL model demonstrated great diagnostic performance for differentiating benign from malignant axillary lymph nodes on breast ultrasound and for potentially providing effective diagnostic support to residents.

Validation of Continuous Noninvasive Blood Pressure Monitoring Using Error Grid Analysis.

BACKGROUND: Error grid analysis was recently proposed to compare blood pressure obtained by 2 measurement methods. This study aimed to compare continuous noninvasive blood pressure (CNBP) with invasive blood pressure (IBP) using the error grid analysis and investigate the confounding risk factors attributable to the differences between CNBP and IBP. METHODS: Sixty adult patients undergoing general anesthesia were prospectively enrolled. Simultaneous comparative data regarding CNBP and IBP were collected. The Bland-Altman analysis was conducted to compare CNBP and IBP for systolic blood pressure (SBP) and mean blood pressure (MBP; acceptable accuracy: mean bias <5 mm Hg; standard deviation <8 mm Hg). The clinical relevance of the discrepancies between CNBP and IBP was evaluated by the error grid analysis, which classifies the differences into 5 zones from "no risk" (A) to "dangerous risk" (E). Additionally, an ordinal logistic regression analysis was performed to evaluate the relationship between the risk zones for MBP, classified by the error grid analysis and covariates of interest. RESULTS: A total of 10,663 pairs of CNBP/IBP were finally analyzed. The Bland-Altman analysis showed an acceptable accuracy with a bias of -3.3 ± 5.6 mm Hg for MBP but a poor accuracy with a bias of 5.4 ± 10.5 mm Hg for SBP. The error grid analysis showed the proportions of zones A to E as 96.7%, 3.2%, 0.1%, 0%, and 0% for SBP, respectively, and 72.0%, 27.9%, 0.1%, 0%, and 0% for MBP, respectively. The finger cuff missed 23.9% of epochs when SBP <90 mm Hg and 55.3% of epochs when MBP <65 mm Hg. The ordinal logistic regression analysis revealed that older age (adjusted odds ratio for decade: 1.54, 95% confidence interval [CI], 1.15-2.08; P = .004) and length of time from the initiation of finger cuff inflation (adjusted odds ratio for 60 minutes: 1.40, 95% CI, 1.13-1.73; P = .002) were significant factors of being in a more dangerous zone of the error grid. CONCLUSIONS: The error grid analysis revealed the larger clinical discrepancy between CNBP and IBP in MBP compared with that in SBP. Old age and longer finger cuff inflation time were significant factors of being in a more dangerous zone of the error grid, which could affect the hemodynamic management during surgery.

Deep Learning Using Multiple Degrees of Maximum-Intensity Projection for PET/CT Image Classification in Breast Cancer.

Deep learning (DL) has become a remarkably powerful tool for image processing recently. However, the usefulness of DL in positron emission tomography (PET)/computed tomography (CT) for breast cancer (BC) has been insufficiently studied. This study investigated whether a DL model using images with multiple degrees of PET maximum-intensity projection (MIP) images contributes to increase diagnostic accuracy for PET/CT image classification in BC. We retrospectively gathered 400 images of 200 BC and 200 non-BC patients for training data. For each image, we obtained PET MIP images with four different degrees (0°, 30°, 60°, 90°) and made two DL models using Xception. One DL model diagnosed BC with only 0-degree MIP and the other used four different degrees. After training phases, our DL models analyzed test data including 50 BC and 50 non-BC patients. Five radiologists interpreted these test data. Sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) were calculated. Our 4-degree model, 0-degree model, and radiologists had a sensitivity of 96%, 82%, and 80-98% and a specificity of 80%, 88%, and 76-92%, respectively. Our 4-degree model had equal or better diagnostic performance compared with that of the radiologists (AUC = 0.936 and 0.872-0.967, p = 0.036-0.405). A DL model similar to our 4-degree model may lead to help radiologists in their diagnostic work in the future.

Deep learning for image classification in dedicated breast positron emission tomography (dbPET).

OBJECTIVE: This study aimed to investigate and determine the best deep learning (DL) model to predict breast cancer (BC) with dedicated breast positron emission tomography (dbPET) images. METHODS: Of the 1598 women who underwent dbPET examination between April 2015 and August 2020, a total of 618 breasts on 309 examinations for 284 women who were diagnosed with BC or non-BC were analyzed in this retrospective study. The Xception-based DL model was trained to predict BC or non-BC using dbPET images from 458 breasts of 109 BCs and 349 non-BCs, which consisted of mediallateral and craniocaudal maximum intensity projection images, respectively. It was tested using dbPET images from 160 breasts of 43 BC and 117 non-BC. Two expert radiologists and two radiology residents also interpreted them. Sensitivity, specificity, and area under the receiver operating characteristic curves (AUCs) were calculated. RESULTS: Our DL model had a sensitivity and specificity of 93% and 93%, respectively, while radiologists had a sensitivity and specificity of 77-89% and 79-100%, respectively. Diagnostic performance of our model (AUC = 0.937) tended to be superior to that of residents (AUC = 0.876 and 0.868, p = 0.073 and 0.073), although not significantly different. Moreover, no significant differences were found between the model and experts (AUC = 0.983 and 0.941, p = 0.095 and 0.907). CONCLUSIONS: Our DL model could be applied to dbPET and achieve the same diagnostic ability as that of experts.