Background: Recent advancements in large language models (LLMs) have accelerated their integration into clinical domains, including laboratory medicine. The performance of LLMs in answering board-level laboratory medicine questions has not been comprehensively evaluated. Given the importance of diagnostic accuracy in this field, rigorous and objective evaluations of LLM capabilities are essential. Methods: We assessed 12 LLMs from OpenAI, Anthropic, and Google using 320 Korean Residency Examination questions (2021–2024) spanning six laboratory medicine subspecialties. Standardized prompts were provided via their application programming interfaces under deterministic settings (temperature = 0). Questions were administered thrice to assess response reproducibility. Outputs were compared with validated answers and analyzed for accuracy, reasoning quality, and error typology. Results: Google’s Gemini 2.0 Pro achieved the highest accuracy (80.0%), followed by OpenAI’s GPT-4.5 (77.2%) and Anthropic’s Claude 3.7 Sonnet (74.1%). Accuracy decreased as the difficulty of questions increased (78.0% for easy vs. 45.1% for challenging). Subspecialty performance varied. Al models underperformed on questions on transfusion medicine (mean accuracy: 38.8%), primarily because of limitations in domain-specific and regional knowledge representations. Incorrect answers primarily resulted from reasoning errors. Reproducibility exceeded 95% for most models; however, some residual non-determinism appeared even with greedy decoding (temperature = 0). Conclusions LLMs demonstrated substantial potential for integration into laboratory medicine, particularly in clinical chemistry and immunology. Performance inconsistencies (particularly for high-difficulty questions) and knowledge gaps (notably for transfusion medicine) highlight the necessity for further development—potentially including domain-specific fine-tuning and retrieval-augmented generation integration—and robust expert oversight before clinical application.

Background: Urinalysis, an essential diagnostic tool, faces challenges in terms of standardization and accuracy. The use of artificial intelligence (AI) with mobile technology can potentially solve these challenges. Therefore, we investigated the effectiveness and accuracy of an AI-based program in automatically interpreting urine test strips using mobile phone cameras, an approach that may revolutionize point-of-care testing. Methods: We developed novel urine test strips and an AI algorithm for image capture.Sample images from the Chungnam National University Sejong Hospital were collected to train a k-nearest neighbor classification algorithm to read the strips. A mobile application was developed for image capturing and processing. We assessed the accuracy, sensitivity, specificity, and ROC area under the curve for 10 parameters. Results: In total, 2,612 urine test strip images were collected. The AI algorithm demonstrated 98.7% accuracy in detecting urinary nitrite and 97.3% accuracy in detecting urinary glucose. The sensitivity and specificity were high for most parameters. However, this system could not reliably determine the specific gravity. The optimal time for capturing the test strip results was 75 secs after dipping. Conclusions The AI-based program accurately interpreted urine test strips using smartphone cameras, offering an accessible and efficient method for urinalysis. This system can be used for immediate analysis and remote testing. Further research is warranted to refine test parameters such as specific gravity to enhance accuracy and reliability.

Background: Ensuring reference material (RM) commutability is crucial for evaluating measurement traceability in order to standardize laboratory tests. However, commutability assessment is not routinely performed. We assessed whether RMs prepared following CLSI C37-A guidelines could be used without assessing commutability by evaluating their commutability for four lipid measurements using the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) and CLSI EP14 protocols. Methods: We analyzed total cholesterol (TC), triglycerides (TGs), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C) in frozen sera from 20 individuals and 11 RMs, prepared by the Korea Disease Control and Prevention AgencyLaboratory Standardization Project (per CLSI C37-A), using six routine measurement procedures (MPs). Regression equations and 95% prediction intervals derived from single-donor sera were analyzed following CLSI EP14. The IFCC protocol was used to assess differences in inter-MP biases between RM and clinical samples. The effect of the TG concentration on commutability was evaluated by analyzing biases between MP results and reference procedure-assigned values. Results: RMs were commutable for most MP pairs for TC and TG. Commutability for HDL-C and LDL-C varied across RMs, with RM10 and RM11 showing higher TG levels (2.38 and 2.95 mmol/L, respectively) and lower commutability. Increased bias percentages from assigned values were observed for RMs with higher TG levels. Conclusions RMs prepared per CLSI C37-A were commutable with most MP pairs for TC and TG. Elevated TG levels affected HDL-C and LDL-C commutability, highlighting the need to consider TG concentrations during RM preparation and assess commutability to standardize laboratory tests.

Gemella sanguinis is a gram-positive, facultatively anaerobic coccus bacterium that has rarely been reported as a cause of infective endocarditis. A 41-year-old male patient with mitral valve prolapse visited the outpatient clinic presenting with fever. Transthoracic echocardiography and transesophageal echocardiography revealed myxomatous change and vegetation of the mitral valve. We isolated G. sanguinis from the patient’s blood, cultured it in both aerobic and anaerobic blood culture bottles, and identified it using matrix-assisted laser desorption/ionization-time of flight mass spectrometry (MALDI-TOF MS; bioMérieux, France) and 16s rRNA sequencing. The isolated G. sanguinis was highly susceptible to penicillin and vancomycin and intermediately susceptible to erythromycin, clindamycin, and levofloxacin. Following the American Heart Association recommendations, this highly penicillin-sensitive isolate was eradicated with ceftriaxone and gentamicin, and the patient recovered and was discharged. To the best of our knowledge, this is the first reported case in Korea where G. sanguinis, the causative agent of endocarditis, was identified using MALDI-TOF MS and 16s rRNA sequencing and was treated with only antibiotics and without surgical valve replacement.

PURPOSE: Refractory Mycoplasma pneumonia (RMP) has been increasing not only in Korea but worldwide. We investigated the incidence of M. pneumonia resistant to macrolides and risk factors for RMP. METHODS: From October 2015 to May 2016, 62 pediatric patients who were admitted due to pneumonia diagnosed on the basis of chest x-ray with respiratory symptoms and positive for M. pneumoniae in polymerase chain reaction with no evidence of other bacterial or viral infections were included. Sequence analysis of the 23S rRNA gene in M. pneumoniae was performed to identify macrolide resistance. Patients with congenital anomalies, history of pulmonary disease, and unclear information on antibiotic use were excluded. RESULTS: Mutations in the 23S rRNA gene were detected in 50 of 62 patients (80.6%). Risk factors were analyzed in only 45 patients. The RMP group consisted of 26 patients (57.8%) who had fever lasting more than 5 days and deteriorating chest x-ray findings. The lactate dehydrogenase (LDH) and C-reactive protein (CRP) levels were significantly higher in the RMP group than in the non-RMP group (LDH: 300±79 U/L vs. 469±206 U/L, CRP: 4.9±4.3 mg/dL vs. 2.5±1.7 mg/dL; P = 0.04 vs. P = 0.026). In univariate analysis, the RMP group was significantly associated with 23S rRNA gene mutation, lobar pneumonia, and pleural effusion (odds ration [OR]: 10.8, 4.1, 5.3; P = 0.004, P = 0.036, P = 0.046). The presence of macrolide resistance was found to be only a significant risk factor in logistic regression (OR; 8.827; 95% confidence interval, 1.376–56.622; P = 0.022). CONCLUSION: Macrolide resistance was a significant risk factor in patients with RMP and identification of macrolide resistance might be helpful in predicting RMP and establishing target therapy for RMP.
C-Reactive Protein ; Child ; Fever ; Genes, rRNA ; Humans ; Incidence ; Korea ; L-Lactate Dehydrogenase ; Logistic Models ; Lung Diseases ; Macrolides ; Mycoplasma ; Pleural Effusion ; Pneumonia ; Pneumonia, Mycoplasma ; Polymerase Chain Reaction ; Risk Factors ; Sequence Analysis ; Thorax

BACKGROUND: Fecal occult blood tests have been widely used as a means of gastrointestinal bleeding and colorectal cancer screening. HM-JACKarc (Kyowa Medex Co. Ltd, Japan) is a recently introduced automated fecal occult blood test analyser, which uses latex agglutination method. We evaluated the analytical performance of HM-JACKarc. METHODS: The linearity and precision for HM-JACKarc were evaluated according to the corresponding Clinical and Laboratory Standard Institute guidelines. The comparison study between HM-JACKarc and OC-SENSOR DIANA (Eiken Chemical Co. Ltd., Japan) was done with stool specimens. RESULTS: The linearity was good (R²=0.999) and the coefficients of variation of within-day precision and between-day precision were 5.2% and 4.9%, respectively, in low concentration and 2.7% each in high concentration. The concordance rate between HM-JACKarc and OCSENSOR DIANA was 99.0% (198 out of 200). CONCLUSIONS: HM-JACKarc showed excellent performance in linearity, precision, and comparison studies. Therefore, it appears to be a useful automated fecal occult blood test analyser.
Agglutination ; Colorectal Neoplasms ; Hemorrhage ; Latex ; Mass Screening ; Methods ; Occult Blood*

BACKGROUND: Dysfunctional natural anticoagulant systems enhance intravascular fibrin for mation in disseminated intravascular coagulation (DIC), and plasma levels of natural anti coagulants can be used in the diagnosis and prognosis of DIC. Herein, the diagnostic value of 4 natural anticoagulants was assessed, and the prognostic value of antithrombin and protein C were validated in a large population. METHODS: Part 1 study included 126 patients with clinically suspected DIC and estimated plasma levels of 4 candidate anticoagulant proteins: antithrombin, protein C, protein S, and protein Z. Part 2 comprised 1,846 patients, in whom plasma antithrombin and protein C levels were compared with other well-known DIC markers according to the underlying dis eases. The 28-day mortality rate was used to assess prognostic outcome. RESULTS: Antithrombin and protein C showed higher areas under the ROC curve than pro tein S and protein Z. In part 2 of the study, antithrombin and protein C levels significantly correlated with DIC score, suggesting that these factors are good indicators of DIC severity. Antithrombin and protein C showed significant prognostic power in Kaplan-Meier analyses. In patients with sepsis/severe infection, antithrombin and protein C showed higher hazard ratios than D-dimer. Platelet count showed the highest hazard ratio in patients with hemato logic malignancy. In patients with liver disease, the hazard ratio for antithrombin levels was significantly high. CONCLUSIONS: Decreased plasma anticoagulant levels reflect florid consumption of the phys iologic defense system against DIC-induced hypercoagulation. Plasma antithrombin and protein C levels are powerful prognostic markers of DIC, especially in patients with sepsis/severe infection.
Adult ; Aged ; Anticoagulants/*blood ; Antithrombins/*blood ; Blood Platelets/cytology ; Blood Proteins/analysis ; Disseminated Intravascular Coagulation/complications/*diagnosis/mortality ; Female ; Fibrin Fibrinogen Degradation Products/analysis ; Humans ; Male ; Middle Aged ; Platelet Count ; Prognosis ; Protein C/*analysis ; Protein S/analysis ; Prothrombin Time ; Regression Analysis ; Sepsis/complications/*diagnosis ; Severity of Illness Index

The cytogenetic analysis of mesenchymal stromal cells (MSCs) is essential for verifying the safety and stability of MSCs. An in situ technique, which uses cells grown on coverslips for karyotyping and minimizes cell manipulation, is the standard protocol for the chromosome analysis of amniotic fluids. Therefore, we applied the in situ karyotyping technique in MSCs and compared the quality of metaphases and karyotyping results with classical G-banding and chromosomal abnormalities with fluorescence in situ hybridization (FISH). Human adipose- and umbilical cord-derived MSC cell lines (American Type Culture Collection PCS-500-011, PCS-500-010) were used for evaluation. The quality of metaphases was assessed by analyzing the chromosome numbers in each metaphase, the overlaps of chromosomes and the mean length of chromosome 1. FISH was performed in the interphase nuclei of MSCs for 6q, 7q and 17q abnormalities and for the enumeration of chromosomes via oligo-FISH in adipose-derived MSCs. The number of chromosomes in each metaphase was more variable in classical G-banding. The overlap of chromosomes and the mean length of chromosome 1 as observed via in situ karyotyping were comparable to those of classical G-banding (P=0.218 and 0.674, respectively). Classical G-banding and in situ karyotyping by two personnel showed normal karyotypes for both cell lines in five passages. No numerical or structural chromosomal abnormalities were found by the interphase-FISH. In situ karyotyping showed equivalent karyotype results, and the quality of the metaphases was not inferior to classical G-banding. Thus, in situ karyotyping with minimized cell manipulation and the use of less cells would be useful for karyotyping MSCs.
Azure Stains ; Chromosome Banding/*methods ; Humans ; In Situ Hybridization, Fluorescence/*methods ; Karyotyping/*methods ; Mesenchymal Stromal Cells/*cytology

BACKGROUND: Hemoglobin A1c (HbA1c) levels are widely used to monitor glycemic control in diabetes mellitus patients, and various methods are used for determining HbA1c levels. The ADAMS A1c HA-8180 (Arkray, Inc., Japan) is a fully automated HbA1c analyzer based on high-performance liquid chromatography (HPLC). METHODS: The analytical performance of the ADAMS A1c HA-8180 analyzer was evaluated on the basis of its precision, linearity, correlation with the Variant II Turbo (Bio-Rad Laboratories, USA), and agreement with the National Glycohemoglobin Standardization Program (NGSP) targets. All evaluations were performed according to Clinical and Laboratory Standard Institute (CLSI) guidelines EP05, EP06, and EP09. RESULTS: Coefficients of variation (CVs) for total precision at low and high levels were 0.99% and 1.16%, respectively. The linearity was excellent with R2 = 0.99 in the range of 4.98-15.10%. Its analytical performance was well correlated with that of Variant II Turbo (r = 0.9987). The 95% confidence interval of bias between the NGSP target and the levels measured using the ADAMS A1c HA-8180 was -0.402-0.225. CONCLUSIONS: The ADAMS A1c HA-8180 showed excellent precision, linearity, correlation with Variant II Turbo, and agreement with the NGSP target. Therefore, its analytical performance is satisfactory for diagnosis and treatment monitoring of diabetes.
Bias (Epidemiology) ; Chromatography, Liquid ; Diabetes Mellitus ; Hemoglobins ; Humans ; Organothiophosphorus Compounds