End-to-End Semi-Supervised Opportunistic Osteoporosis Screening Using Computed Tomography.

Background: Osteoporosis is the most common metabolic bone disease and can cause fragility fractures. Despite this, screening utilization rates for osteoporosis remain low among populations at risk. Automated bone mineral density (BMD) estimation using computed tomography (CT) can help bridge this gap and serve as an alternative screening method to dual-energy X-ray absorptiometry (DXA). Methods: The feasibility of an opportunistic and population agnostic screening method for osteoporosis using abdominal CT scans without bone densitometry phantom-based calibration was investigated in this retrospective study. A total of 268 abdominal CT-DXA pairs and 99 abdominal CT studies without DXA scores were obtained from an oncology specialty clinic in the Republic of Korea. The center axial CT slices from the L1, L2, L3, and L4 lumbar vertebrae were annotated with the CT slice level and spine segmentation labels for each subject. Deep learning models were trained to localize the center axial slice from the CT scan of the torso, segment the vertebral bone, and estimate BMD for the top four lumbar vertebrae. Results: Automated vertebra-level DXA measurements showed a mean absolute error (MAE) of 0.079, Pearson's r of 0.852 (P<0.001), and R2 of 0.714. Subject-level predictions on the held-out test set had a MAE of 0.066, Pearson's r of 0.907 (P<0.001), and R2 of 0.781. Conclusion: CT scans collected during routine examinations without bone densitometry calibration can be used to generate DXA BMD predictions.

The Efficacy and Safety of an Indocyanine Green-Macroaggregated Albumin-Hyaluronic Acid Mixture (LuminoMark™) for Surgical Localization of Recurrent Thyroid Cancer.

BACKGROUND: The recurrence of thyroid cancer poses challenges compounded by postoperative fibrosis and anatomic changes. By overcoming the limitations of current localizing dye techniques, indocyanine green-macroaggregated albumin-hyaluronic acid (ICG-MAA-HA) mixture dye promises improved localization. This study aimed to evaluate the efficacy and safety of the dye for recurrent thyroid cancer. METHODS: The nine patients in this study underwent surgery and postoperative ultrasonography. The dye was injected into recurrent lesions in all the patients preoperatively. During surgery, the lesions were confirmed with an imaging system before and after excision. If the lesion was unidentifiable with the naked eye, surgical excision was performed under the corresponding fluorescent guide. Side effects related to the dye injection and completeness of the surgery were evaluated. RESULTS: No side effects such as bleeding, skin tattooing, or pain during or after the dye injection were reported, and no discoloration occurred that interfered with the surgical field of view during surgery. In three cases (33.3 %), because it was difficult to localize metastatic lesions with the naked eye, the operation was successfully completed using an imaging system. The completeness of the surgical resection was confirmed by ultrasonography after an average of 5 months postoperatively. CONCLUSION: The study found that ICG-MAA-HA dye effectively located metastatic and recurrent thyroid cancer and had favorable results in terms of minimal procedural side effects and potential for assisting the surgeon. A large-scale multi-institutional study is necessary to prove the clinical significance regarding patient survival and disease control.

Data-Efficient Computational Pathology Platform for Faster and Cheaper Breast Cancer Subtype Identifications: Development of a Deep Learning Model.

BACKGROUND: Breast cancer subtyping is a crucial step in determining therapeutic options, but the molecular examination based on immunohistochemical staining is expensive and time-consuming. Deep learning opens up the possibility to predict the subtypes based on the morphological information from hematoxylin and eosin staining, a much cheaper and faster alternative. However, training the predictive model conventionally requires a large number of histology images, which is challenging to collect by a single institute. OBJECTIVE: We aimed to develop a data-efficient computational pathology platform, 3DHistoNet, which is capable of learning from z-stacked histology images to accurately predict breast cancer subtypes with a small sample size. METHODS: We retrospectively examined 401 cases of patients with primary breast carcinoma diagnosed between 2018 and 2020 at the Department of Pathology, National Cancer Center, South Korea. Pathology slides of the patients with breast carcinoma were prepared according to the standard protocols. Age, gender, histologic grade, hormone receptor (estrogen receptor [ER], progesterone receptor [PR], and androgen receptor [AR]) status, erb-B2 receptor tyrosine kinase 2 (HER2) status, and Ki-67 index were evaluated by reviewing medical charts and pathological records. RESULTS: The area under the receiver operating characteristic curve and decision curve were analyzed to evaluate the performance of our 3DHistoNet platform for predicting the ER, PR, AR, HER2, and Ki67 subtype biomarkers with 5-fold cross-validation. We demonstrated that 3DHistoNet can predict all clinically important biomarkers (ER, PR, AR, HER2, and Ki67) with performance exceeding the conventional multiple instance learning models by a considerable margin (area under the receiver operating characteristic curve: 0.75-0.91 vs 0.67-0.8). We further showed that our z-stack histology scanning method can make up for insufficient training data sets without any additional cost incurred. Finally, 3DHistoNet offered an additional capability to generate attention maps that reveal correlations between Ki67 and histomorphological features, which renders the hematoxylin and eosin image in higher fidelity to the pathologist. CONCLUSIONS: Our stand-alone, data-efficient pathology platform that can both generate z-stacked images and predict key biomarkers is an appealing tool for breast cancer diagnosis. Its development would encourage morphology-based diagnosis, which is faster, cheaper, and less error-prone compared to the protein quantification method based on immunohistochemical staining.

Development and Verification of Time-Series Deep Learning for Drug-Induced Liver Injury Detection in Patients Taking Angiotensin II Receptor Blockers: A Multicenter Distributed Research Network Approach.

OBJECTIVES: The objective of this study was to develop and validate a multicenter-based, multi-model, time-series deep learning model for predicting drug-induced liver injury (DILI) in patients taking angiotensin receptor blockers (ARBs). The study leveraged a national-level multicenter approach, utilizing electronic health records (EHRs) from six hospitals in Korea. METHODS: A retrospective cohort analysis was conducted using EHRs from six hospitals in Korea, comprising a total of 10,852 patients whose data were converted to the Common Data Model. The study assessed the incidence rate of DILI among patients taking ARBs and compared it to a control group. Temporal patterns of important variables were analyzed using an interpretable timeseries model. RESULTS: The overall incidence rate of DILI among patients taking ARBs was found to be 1.09%. The incidence rates varied for each specific ARB drug and institution, with valsartan having the highest rate (1.24%) and olmesartan having the lowest rate (0.83%). The DILI prediction models showed varying performance, measured by the average area under the receiver operating characteristic curve, with telmisartan (0.93), losartan (0.92), and irbesartan (0.90) exhibiting higher classification performance. The aggregated attention scores from the models highlighted the importance of variables such as hematocrit, albumin, prothrombin time, and lymphocytes in predicting DILI. CONCLUSIONS: Implementing a multicenter-based timeseries classification model provided evidence that could be valuable to clinicians regarding temporal patterns associated with DILI in ARB users. This information supports informed decisions regarding appropriate drug use and treatment strategies.

Progression of Low-Risk Papillary Thyroid Microcarcinoma During Active Surveillance: Interim Analysis of a Multicenter Prospective Cohort Study of Active Surveillance on Papillary Thyroid Microcarcinoma in Korea.

Background: Active surveillance (AS) is an alternative to thyroidectomy for the management of low-risk papillary thyroid microcarcinoma (PTMC). However, prospective AS data collected from diverse populations are needed. Methods: This multicenter prospective cohort study enrolled patients from three referral hospitals in Korea. The participants were self-assigned into two groups, AS or immediate surgery. All patients underwent neck ultrasound every 6-12 months to monitor for disease progression. Progression under AS was evaluated by a criterion of tumor size increment by 3 mm in one dimension (3 mm), 2 mm in two dimensions (2 × 2 mm), new extrathyroidal extension (ETE), or new lymph node metastasis (LNM), and a composite outcome was defined using all four criteria. Results: A total of 1177 eligible patients with PTMC (919 female, 78.1%) with a median age of 48 years (range 19-87) were enrolled; 755 (64.1%) patients chose AS and 422 (35.9%) underwent surgery. Among 755 patients under AS, 706 (female 537, 76.1%) underwent at least two ultrasound examinations and were analyzed. Over a follow-up period of 41.4 months (standard deviation, 16.0), 163 AS patients (23.1%) underwent surgery. Progression defined by the composite outcome was observed in 9.6% (68/706) of patients, and the 2- and 5-year progression estimates were 5.3% and 14.2%, respectively. The observed progression rates were 5.8% (41/706) and 5.4% (38/706) as defined by tumor size enlargement by 3 mm and 2 × 2 mm, respectively, and 1.3% (9/706) and 0.4% (3/706) for new LNM and ETE, respectively. No distant metastases developed during AS. In multivariate logistic regression analysis examining variables associated with progression under AS, age at diagnosis <30 years (odds ratio [OR], 2.86; 95% confidence interval [CI], 1.10 - 7.45), male sex (OR, 2.48; 95% CI, 1.47 - 4.20), and tumor size ≥6 mm (OR, 1.89; 95% CI, 1.09 - 3.27) were independently significant. Conclusions: The progression of low-risk PTMC during AS in the Korean population was low, but slightly higher than previously reported in other populations. Risk factors for disease progression under AS include younger age, male sex, and larger tumor size. Clinical trial registration: Clinicaltrials.gov NCT02938702.

Effect of a Wireless Vital Sign Monitoring System on the Rapid Response System in the General Ward.

While wireless vital sign monitoring is expected to reduce the vital sign measurement time (thus reducing the nursing workload), its impact on the rapid response system is unclear. This study compared the time from vital sign measurement to recording and rapid response system activation between wireless and conventional vital sign monitoring in the general ward, to investigate the impact of wireless vital sign monitoring system on the rapid response system. The study divided 249 patients (age > 18 years; female: 47, male: 202) admitted to the general ward into non-wireless (n = 101) and wireless (n = 148) groups. Intervals from vital sign measurement to recording and from vital sign measurement to rapid response system activation were recorded. Effects of wireless system implementation for vital sign measurement on the nursing workload were surveyed in 30 nurses. The interval from vital sign measurement to recording was significantly shorter in the wireless group than in the non-wireless group (4.3 ± 2.9 vs. 44.7 ± 14.4 min, P < 0.001). The interval from vital sign measurement to rapid response system activation was also significantly lesser in the wireless group than in the non-wireless group (27.5 ± 12.9 vs. 41.8 ± 19.6 min, P = 0.029). The nursing workload related to vital sign measurement significantly decreased from 3 ± 0.87 to 2.4 ± 9.7 (P = 0.021) with wireless system implementation. Wireless vital sign monitoring significantly reduced the time to rapid response system activation by shortening the time required to measure the vital signs. It also significantly reduced the nursing workload.

Long-term clinical outcomes of oral antidiabetic drugs as fixed-dose combinations: A nationwide retrospective cohort study.

AIM: To compare treatment patterns and clinical outcomes of single-pill fixed-dose combination (FDC) and two-pill combination (TPC) therapies using real-world data. METHODS: We conducted a nationwide retrospective cohort study using South Korea's healthcare database (2002-2015). We identified two cohorts of incident patients with type 2 diabetes who initiated FDC or TPC therapy within 4 months of their first prescription for metformin or sulphonylurea. We examined persistence and adherence patterns and the clinical outcome of a composite endpoint of death or hospitalization for acute myocardial infarction, heart failure or stroke and compared the differences in treatment patterns and clinical outcomes using Cox models. RESULTS: Of 5143 and 10 973 patients who initiated FDC and TPC therapy, respectively, we identified 5143 patient pairs after propensity score matching. The FDC group exhibited greater median time to treatment discontinuation (163 vs. 146 days), and proportion of days covered at 12 months (mean 0.60 vs. 0.57, P < .0001) and at 24 months (0.53 vs. 0.51, P = .014) than the TPC group. The FDC group, compared with the TPC group, had reduced risks of the composite clinical outcome (hazard ratio 0.86, 95% confidence intervals 0.77-0.97) and hospitalization for stroke (0.80, 0.67-0.96). CONCLUSION: FDC therapy may provide favourable cardiovascular benefits, especially reducing the risk of hospitalization for stroke, and has better medication adherence among patients with type 2 diabetes.

A Cross-Sectional Survey of Patient Treatment Choice in a Multicenter Prospective Cohort Study on Active Surveillance of Papillary Thyroid Microcarcinoma (MAeSTro).

Background: Active surveillance (AS) is offered as a choice to patients with low-risk papillary thyroid microcarcinoma (PTMC). This study aimed to identify patient and physician factors associated with the choice of AS. Methods: We conducted a cross-sectional survey of patients with low-risk PTMC who were enrolled in a prospective study comparing outcomes following AS and surgery. Patients completed a questionnaire to assess their prior knowledge of the disease, considerations in the decision-making process, and reasons for choosing the treatment. We also surveyed 19 physician investigators about their disease management preferences. Variables affecting the patients' choice of AS, including patients' characteristics and their decision-making process, were analyzed in a multivariable analysis. Results: The response rate of the patient survey was 72.8% (857/1177). Among the patients who responded to the survey, 554 patients (128 male; mean age 49.4 ± 11.6 years; response rate 73.4%) with low-risk PTMC chose AS (AS group), whereas 303 patients (55 male; 46.6 ± 10.7 years; 71.8%) chose immediate surgery (iOP group). In the AS group, 424 patients (76.5%) used a decision aid, and 144 (47.5%) used it in the iOP group. The choice of AS was associated with the following variables: patient age >50 years (odds ratio 1.713 [confidence interval, CI 1.090-2.690], p = 0.020), primary tumor size ≤5 mm (odds ratio 1.960 [CI 1.137-3.379], p = 0.015), and consulting an endocrinologist (odds ratio 114.960 [CI 48.756-271.057], p < 0.001), and use of a decision aid (odds ratio 2.469 [CI 1.320-4.616], p = 0.005). The proportion of patients who were aware of AS before their initial consultation for treatment decision was higher in the AS group than in the iOP group (64.6% vs. 56.8%). Family members were reported to have influenced the treatment decisions more in the iOP group (p = 0.025), whereas the AS group was more influenced by information from the media (p = 0.017). Physicians' attitudes regarding AS of low-risk PTMC tended to be more favorable among endocrinologists than surgeons and all became more favorable as the study progressed. Conclusions: Emerging evidence suggests that physicians' attitudes and communication tools influence the treatment decision of low-risk PTMC patients. Support is needed for patient-centered decision making. (Clinical trial No: NCT02938702).

Seaweed and Iodine Intakes and SLC5A5 rs77277498 in Relation to Thyroid Cancer.

BACKGRUOUND: This study aims to elucidate the associations among dietary seaweed (gim and miyeok/dashima) and iodine intakes, the rs77277498 polymorphism of the SLC5A5 gene codifying the sodium/iodine symporter, and thyroid cancer risk in a Korean population. METHODS: We conducted a case-control study of 117 thyroid cancer cases and 173 controls who participated in the Cancer Screenee Cohort between 2002 and 2014 at the National Cancer Center, Korea. The amount of seaweed and iodine consumption (g/day) was estimated using the residual energy adjustment method. We calculated odds ratios (ORs) and their 95% confidence intervals (CIs) using a multivariable logistic regression model for the separate and combined effect of dietary iodine-based intake and SLC5A5 polymorphism (rs77277498, C>G) on thyroid cancer. RESULTS: Dietary gim and iodine intakes were inversely associated with thyroid cancer, with ORs of 0.50 (95% CI, 0.30 to 0.83) and 0.57 (95% CI, 0.35 to 0.95), respectively, whereas the associations for dietary miyeok/dashima and total seaweed intakes were not significant. However, compared with individuals carrying the C/C genotype of the rs77277498 polymorphism with a low intake of all dietary factors, those carrying the G allele with a high intake had a lower risk of thyroid cancer, with ORs of 0.25 (95% CI, 0.10 to 0.56), 0.31 (95% CI, 0.12 to 0.77), 0.26 (95% CI, 0.10 to 0.62), and 0.30 (95% CI, 0.12 to 0.73) for the consumption of gim, miyeok/dashima, total seaweed, and iodine, respectively. CONCLUSION: In summary, our results supported the evidence of the protective effects of dietary gim and iodine intake against thyroid cancer risk, and this association can be strengthened by SLC5A5 rs77277498 genotypes.

Can computed tomography scanning in adults lead to an increased risk of thyroid cancer? A nationwide nested case-control study.

OBJECTIVE: To evaluate the association between computed tomography (CT) scanning and newly diagnosed thyroid cancer cases in relation to the confounding effect of the healthcare utilization rate. METHODS: This nested case-control study used the Korean National Health Insurance Service-National Sample Cohort 2002-2015: 3557 adult thyroid cancer cases were matched to 17,785 controls by age, sex, and diagnosis date. Odds ratios (ORs) were estimated for thyroid cancer associated with cumulative exposure to CT scanning > 3 years before cancer diagnosis. Changes in estimated ORs with and without adjustment for outpatient visit frequency were investigated. RESULTS: ORs for newly diagnosed thyroid cancer increased according to the higher number of total CT scans and thyroid-exposing CT scans (CT scans of the head, neck, or chest compartment; OR and 95% confidence interval [CI], 1.09 [1.03-1.16] and 1.28 [1.05-1.57], respectively). ORs for thyroid cancer increased according to higher outpatient visit frequency. The association between thyroid cancer incidence and CT scans became insignificant when outpatient visit frequency was adjusted in the models (OR [95% CI], 1.03 [0.97-1.10]: total CT scans, 1.14 [0.93-1.41]: thyroid-exposing CT scans). Subgroup analyses stratified by age, sex, and history of other malignancies did not reveal independent associations between CT scanning and thyroid cancer. CONCLUSIONS: The high incidence of thyroid cancer in adults exposed to ionizing radiation during CT scanning can be largely explained by the confounding effect of the healthcare utilization rate. These effects should be considered to avoid overestimation of the CT scanning-associated risk of thyroid cancer. KEY POINTS: • Studies indicate that diagnostic imaging using low-ionizing radiation may increase risks for thyroid cancer in adults. • Our findings suggest that the risk for radiation-induced thyroid cancer following CT scanning in adults may have been overestimated in observational studies due to medical surveillance-related biases.

Metabolic Effects of Gastrectomy and Duodenal Bypass in Early Gastric Cancer Patients with T2DM: A Prospective Single-Center Cohort Study.

We evaluated the metabolic effects of gastrectomies and endoscopic submucosal dissections (ESDs) in early gastric cancer (EGC) patients with type 2 diabetes mellitus (T2DM). Forty-one EGC patients with T2DM undergoing gastrectomy or ESD were prospectively evaluated. Metabolic parameters in the patients who underwent gastrectomy with and without a duodenal bypass (groups 1 and 2, n = 24 and n = 5, respectively) were compared with those in patients who underwent ESD (control, n = 12). After 1 year, the proportions of improved/equivocal/worsened glycemic control were 62.5%/29.2%/8.3% in group 1, 40.0%/60.0%/0.0% in group 2, and 16.7%/50.0%/33.3% in the controls, respectively (p = 0.046). The multivariable ordered logistic regression analysis results showed that both groups had better 1-year glycemic control. Groups 1 and 2 showed a significant reduction in postprandial glucose (-97.9 and -67.8 mg/dL), body mass index (-2.1 and -2.3 kg/m2), and glycosylated hemoglobin (group 1 only, -0.5% point) (all p < 0.05). Furthermore, improvements in group 1 were more prominent when preoperative leptin levels were high (p for interaction < 0.05). Metabolic improvements in both groups were also observed for insulin resistance, leptin, plasminogen activator inhibitor-1, and resistin. Gastrectomy improved glycemic control and various metabolic parameters in EGC patients with T2DM. Patients with high leptin levels may experience greater metabolic benefits from gastrectomy with duodenal bypass.

Global Trends of Regional Health Information Systems and Suggested Strategic Utilization of their Medical Information.

OBJECTIVE: Along with the exponentially-growing data produced and accumulated every day through mobile platforms, social networking services, the Internet, and other media, information is becoming increasingly important as a strategic resource. This report presents specific and clear directions and suggests empirical project plans regarding innovations in regional health information systems to promote the utilization of medical information. METHODS: We reviewed and examined documents about global trends and examples of regional health information systems. The problems and solutions of health information utilization and regional health information systems in Korea were analyzed. RESULTS: This study presented examples of the establishment of health information systems, problems in the use of local healthcare information, and an empirical project for improvement. CONCLUSIONS: The results of this study imply the need for long-term and systematic approaches for the use of medical information and the establishment of a local healthcare information system, along with implementation plans. As a first step, it is imperative to clarify the goal of building a medical information system, the information that must be provided to build the system, and the data that should be collected to provide such information, while moving away from the mentality of focusing on technology-oriented medical information services. In addition, it is necessary to consider information governance, data-based service development, and the medical innovation framework, which are ways to efficiently manage, utilize, and systemize the data to be collected.

Application of Machine Learning to Identify Clinically Meaningful Risk Group for Osteoporosis in Individuals Under the Recommended Age for Dual-Energy X-Ray Absorptiometry.

Dual-energy X-ray absorptiometry (DXA) is the gold standard for diagnosing osteoporosis; it is generally recommended in men ≥ 70 and women ≥ 65 years old. Therefore, assessment of clinical risk factors for osteoporosis is very important in individuals under the recommended age for DXA. Here, we examine the diagnostic performance of machine learning-based prediction models for osteoporosis in individuals under the recommended age for DXA examination. Data of 2210 men aged 50-69 and 1099 women aged 50-64 obtained from the Korea National Health and Nutrition Examination Survey IV-V were analyzed. Extreme gradient boosting (XGBoost) was used to find relevant clinical features and applied to three machine learning models: XGBoost, logistic regression, and a multilayer perceptron. For the prediction of osteoporosis, the XGBoost model using the top 20 features extracted from XGBoost showed the most reliable performance with area under the receiver operating characteristic curve (AUROC) of 0.73 and 0.79 in men and women, respectively. We compared the diagnostic accuracy of the Shapley additive explanation values based on a risk-score model obtained from XGBoost and conventional osteoporosis risk assessment tools for prediction of osteoporosis using optimal cut-off values for each model. We observed that a cut-off risk score of ≥ 28 in men and ≥ 47 in women was optimal to classify a positive screening for osteoporosis (an AUROC of 0.86 in men and 0.91 in women). The XGBoost-based osteoporosis-prediction model outperformed conventional risk assessment tools. Therefore, machine learning-based prediction models are a more suitable option than conventional risk assessment methods for screening osteoporosis in individuals under the recommended age for DXA examination.

Evaluation of modifiable factors and polygenic risk score in thyroid cancer.

The cumulative effect of single-nucleotide polymorphisms (SNPs) on thyroid cancer has been adequately defined in individuals of European ancestry; however, similar evidence in the Korean population is limited. This study aimed to investigate the influence of modifiable factors and the polygenic risk score (PRS) and their interactive and combined effects on thyroid cancer. Using data from the cancer screenee cohort, this study included 759 thyroid cancer cases and 759 age- and sex-matched controls. We examined the effects of tobacco smoking, alcohol consumption, and regular exercise habits, BMI, and the PRS of six SNPs on thyroid cancer. Odds ratios (ORs) and 95% confidence intervals (CIs) for the associations were obtained using a conditional logistic regression model. The results indicated that family history, obesity, and the unweighted and weighted PRS were independently associated with susceptibility to thyroid cancer, with ORs (95% CIs) of 2.96 (1.63-5.36), 1.72 (1.20-2.48), 1.46 (1.10-1.93), and 1.56 (1.19-2.03), respectively, whereas the effect of smoking, drinking, and regular exercise was not significant. The contribution of the PRS remained after stratifying participants with healthy behaviors, such as nonsmokers/nondrinkers, and regular exercise. Although the PRS did not significantly contribute to the risk for thyroid cancer when participants were stratified according to BMI, BMI and the PRS had a cumulative effect on thyroid cancer risk. The combined effect of genetic polymorphisms on predisposition to thyroid cancer may differ based on tobacco smoking, alcohol consumption, regular exercise behaviors and cumulative BMI. Larger population-based studies are needed to validate these findings.

Impact of image compression on deep learning-based mammogram classification.

Image compression is used in several clinical organizations to help address the overhead associated with medical imaging. These methods reduce file size by using a compact representation of the original image. This study aimed to analyze the impact of image compression on the performance of deep learning-based models in classifying mammograms as "malignant"-cases that lead to a cancer diagnosis and treatment-or "normal" and "benign," non-malignant cases that do not require immediate medical intervention. In this retrospective study, 9111 unique mammograms-5672 normal, 1686 benign, and 1754 malignant cases were collected from the National Cancer Center in the Republic of Korea. Image compression was applied to mammograms with compression ratios (CRs) ranging from 15 to 11 K. Convolutional neural networks (CNNs) with three convolutional layers and three fully-connected layers were trained using these images to classify a mammogram as malignant or not malignant across a range of CRs using five-fold cross-validation. Models trained on images with maximum CRs of 5 K had an average area under the receiver operating characteristic curve (AUROC) of 0.87 and area under the precision-recall curve (AUPRC) of 0.75 across the five folds and compression ratios. For images compressed with CRs of 10 K and 11 K, model performance decreased (average 0.79 in AUROC and 0.49 in AUPRC). Upon generating saliency maps that visualize the areas each model views as significant for prediction, models trained on less compressed (CR < = 5 K) images had maps encapsulating a radiologist's label, while models trained on images with higher amounts of compression had maps that missed the ground truth completely. In addition, base ResNet18 models pre-trained on ImageNet and trained using compressed mammograms did not show performance improvements over our CNN model, with AUROC and AUPRC values ranging from 0.77 to 0.87 and 0.52 to 0.71 respectively when trained and tested on images with maximum CRs of 5 K. This paper finds that while training models on images with increased the robustness of the models when tested on compressed data, moderate image compression did not substantially impact the classification performance of DL-based models.

Incidence and disease course of new-onset diabetes mellitus in breast and colorectal cancer patients undergoing chemotherapy: A prospective multicenter cohort study.

AIMS: To investigate the incidence of and risk factors for new-onset type 2 diabetes mellitus (DM) developed during chemotherapy that included steroids in cancer patients without DM. METHODS: This multicenter, prospective, and observational cohort study enrolled 299 cancer patients without DM (aged > 18 years), planning 4-8 cycles of adjuvant chemotherapy. The endpoints were the incidence, remission rate, and independent determinants of new-onset DM during chemotherapy. RESULTS: Between April 2015 and March 2018, 270 subjects with colorectal cancer or breast cancer (mean age, 51.0 years) completed the follow up (mean 39 months). Of whom, 17 subjects (6.3%) developed DM within a median time of 90 days (range, 17-359 days). Male sex (hazard ratio [HR], 15.839; 95% confidence interval [CI], 2.004-125.20) and impaired fasting glucose (IFG) at baseline (HR, 8.307; CI, 1.826-37.786) were independent risk factors. Six months after chemotherapy completion, 11/17 subjects (64.7%) experienced DM remission, associated with a significantly higher C-peptide level at baseline (C-peptide levels, 1.3 ng/mL in subjects with remission and 0.9 ng/mL in subjects without remission, age- and sex-adjusted P = 0.007). CONCLUSIONS: DM incidence was 6.3% in patients who received chemotherapy with dexamethasone. Close monitoring for hyperglycemia is recommended, especially for men with IFG. TRIAL REGISTRATION: ClinicalTrials.gov (NCT03062072).

Prediction of Prolonged Length of Hospital Stay After Cancer Surgery Using Machine Learning on Electronic Health Records: Retrospective Cross-sectional Study.

BACKGROUND: Postoperative length of stay is a key indicator in the management of medical resources and an indirect predictor of the incidence of surgical complications and the degree of recovery of the patient after cancer surgery. Recently, machine learning has been used to predict complex medical outcomes, such as prolonged length of hospital stay, using extensive medical information. OBJECTIVE: The objective of this study was to develop a prediction model for prolonged length of stay after cancer surgery using a machine learning approach. METHODS: In our retrospective study, electronic health records (EHRs) from 42,751 patients who underwent primary surgery for 17 types of cancer between January 1, 2000, and December 31, 2017, were sourced from a single cancer center. The EHRs included numerous variables such as surgical factors, cancer factors, underlying diseases, functional laboratory assessments, general assessments, medications, and social factors. To predict prolonged length of stay after cancer surgery, we employed extreme gradient boosting classifier, multilayer perceptron, and logistic regression models. Prolonged postoperative length of stay for cancer was defined as bed-days of the group of patients who accounted for the top 50% of the distribution of bed-days by cancer type. RESULTS: In the prediction of prolonged length of stay after cancer surgery, extreme gradient boosting classifier models demonstrated excellent performance for kidney and bladder cancer surgeries (area under the receiver operating characteristic curve [AUC] >0.85). A moderate performance (AUC 0.70-0.85) was observed for stomach, breast, colon, thyroid, prostate, cervix uteri, corpus uteri, and oral cancers. For stomach, breast, colon, thyroid, and lung cancers, with more than 4000 cases each, the extreme gradient boosting classifier model showed slightly better performance than the logistic regression model, although the logistic regression model also performed adequately. We identified risk variables for the prediction of prolonged postoperative length of stay for each type of cancer, and the importance of the variables differed depending on the cancer type. After we added operative time to the models trained on preoperative factors, the models generally outperformed the corresponding models using only preoperative variables. CONCLUSIONS: A machine learning approach using EHRs may improve the prediction of prolonged length of hospital stay after primary cancer surgery. This algorithm may help to provide a more effective allocation of medical resources in cancer surgery.

Risk Factors of Undiagnosed Diabetes Mellitus among Korean Adults: A National Cross-Sectional Study Using the KNHANES Data.

In this cross-sectional study, we investigated the baseline risk factors of diabetes mellitus (DM) in patients with undiagnosed DM (UDM). We utilized the Korean National Health and Nutrition Examination Survey (KNHANES) 2010-2017 data. Data regarding the participants' demographic characteristics, health status, health determinants, healthcare accessibility, and laboratory tests were gathered to explore the differences between the DM, UDM, and without-DM groups. Among the 64,759 individuals who participated in the KNHANES 2010-2017, 32,611 individuals aged ≥20 years with fasting plasma glucose levels of <100 or ≥126 mg/dL were selected. The odds ratios (ORs) regarding family history of diabetes and the performance of national health and cancer screening tests were lower in the UDM group than in the DM group (adjusted OR: 0.54; 95% confidence interval (CI): 0.43, 0.66; adjusted OR: 0.74; 95% CI: 0.62, 0.89; adjusted OR: 0.71; 95% CI: 0.60, 0.85). The ORs of hypertension and obesity were higher in the UDM group than in the DM group (adjusted OR: 1.32; 95% CI: 1.06, 1.64; adjusted OR: 1.80; 95% CI: 1.37, 2.36, respectively). Patients with UDM were more likely to be exposed to DM-related risk factors than those with and without DM. Public health interventions to prevent UDM development are necessary.