南方医科大学学报杂志最新文献

Objectives: To improve the accuracy of machine learning models for preoperative prediction of high-intensity focused ultrasound (HIFU) ablation efficacy for uterine fibroids by correcting class imbalance in small sample datasets using undersampling methods.

Methods: Clinical and imaging data were collected from 140 patients with uterine fibroids undergoing HIFU treatment at Foshan Women and Children Hospital, including 104 with high ablation rates and 36 with low ablation rates. Radiomic features were extracted from MRI T2-weighted images (T2WI) of the patients, and machine learning models were constructed to predict HIFU treatment outcomes. Four machine learning algorithms, including k-Nearest Neighbors (KNN), Random Forest (RF), Support Vector Machine (SVM), and Multilayer Perceptron (MLP), were coupled with 7 undersampling methods, namely Random Undersampling (RUS), Repeated Edited Nearest Neighbors (RENN), All k-Nearest Neighbors (AllKNN), Neighborhood Cleaning Rule-3 (NM), Condensed Nearest Neighbor (CNN), Neighborhood Cleaning Rule (NCR), and Instance Hardness Threshold (IHT), for handling class imbalance in the datasets. The 28 prediction models were evaluated using 5-fold cross-validation for areas under the receiver operating characteristic curve (AUC), accuracy, recall, and specificity.

Results: The best combinations of undersampling methods and machine learning models CNN-RF, NM-SVM, CNN-KNN, and NM-MLP had AUCs of 0.772 (95% CI: 0.566-0.942), 0.797 (95% CI: 0.600-0.950), 0.822 (95% CI: 0.635-0.964), and 0.822 (95% CI: 0.632-0.960), respectively. The AUCs of the machine learning models significantly increased after coupling with undersampling methods, with the MLP model showing the most pronounced improvement. The recall rates of the 4 combined models also improved significantly (by 0.389 for CNN-RF, 0.836 for NM-SVM, 0.532 for CNN-KNN, and 0.372 for NM-MLP).

Conclusions: The use of undersampling methods can effectively correct class imbalance in small sample datasets to improve the accuracy of machine learning models for predicting the efficacy of HIFU ablation for uterine fibroids.

Objectives: To establish a pelvic active bone marrow (ABM) segmentation method based on diffusion cycle-consistent generative adversarial networks for improving individualized precision of conventional anatomical atlas-based methods.

Methods: We collected pelvic PET-CT data from 253 patients and constructed a 3-stage cascaded cross-modal learning framework for precise individualized ABM identification from CT images. The framework used cycle-consistent generative adversarial networks for bidirectional CT-PET mapping, conditional diffusion modules with 1000-step Markov chains for progressive denoising, and multi-scale progressive feature pyramid fusion networks for segmentation. The peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), normalized mean square error (NMSE), Dice similarity coefficient (DSC), and average symmetric surface distance (ASSD) were used for evaluation of the model performance for ABM segmentation.

Results: The proposed method outperformed the existing methods with a PSNR of 26.42±0.63 dB, an SSIM of 0.894±0.011, and an NMSE of 0.0235±0.0026. For ABM segmentation, the average Dice coefficient of the model reached 0.777±0.023 with an ASSD of 3.52±0.41 mm.

Conclusions: Compared with the conventional methods, the propose method significantly improves individualized segmentation accuracy of the ABM and is thus suitable use in individualized bone marrow protection radiotherapy for rectal cancer.

Objectives: Sepsis patients exhibit diverse immune states, making it crucial to identify subtypes with distinct inflammatory profiles through Th1/Th2 cytokine data for personalized treatment and improved prognosis.

Methods: We retrieved data from sepsis patients who underwent Th1/Th2 cytokine testing in Nanfang Hospital, Southern Medical University from June 1, 2020, to February 1, 2022. An unsupervised K-means clustering method classified participants based on Th1/Th2 cytokine levels, with the primary outcome being the 7-day mortality rate post-ICU admission. Cox proportional hazards and Restricted Mean Survival Time (RMST) analyses were utilized to explore survival outcomes.

Results: A total of 321 sepsis patients were included. IL-6 (HR 1.69, 95%CI: 1.22, 2.34) and IL-10 (HR 1.81, 95% CI: 1.37, 2.40) emerged as independent predictors of 7-day mortality. Unsupervised K-means clustering revealed 3 inflammatory/immune subgroups: Cluster 1 (n=166, low inflammatory response), Cluster 2 (n=99, moderate inflammatory response with immune suppression), and Cluster 3 (n=56, strong inflammatory and immune suppression). Compared to Cluster 1, Clusters 2 and 3 had higher 7-day mortality risks (14.4% vs 23.2%, HR=4.30, 95% CI: 1.51-12.26; 14.4% vs 35.7%, HR=7.32, 95% CI: 2.57-20.79).

Conclusions: Septic patients in a protective immune response state (Cluster 1) exhibit better short-term prognoses, suggesting the importance of understanding inflammatory/immune states for precise treatment and improved outcomes.

Objectives: To investigate the regulatory role of lncRNA SNHG12 in docetaxel (DTX) resistance of prostate cancer (PCa) cells.

Methods: Tumor-bearing male BALB/c nude mouse models were stablished by dorsal subcutaneous injection of PC-3 cells or DTX-resistant PC-3 (PC-3R) cells, either with or without transfection with sh-SNHG12 prior to the injection (n=5). The expressions of the key genes and proteins in the tumor tissues were detected using RT-qPCR, Western blotting, immunofluorescence staining or immunohistochemistry. The proliferation and migration of the treated cells were evaluated with CCK-8, clone formation and Transwell migration assays. RIP-qPCR technique was used to determine the binding between the RNAs and proteins.

Results: SNHG12 expression was significantly up-regulated in PC-3R cells. SNHG12 knockdown effectively inhibited proliferation and migration of PC-3R cells in vitro and suppressed tumor growth in nude mice. While 10 nmol/L DTX treatment alone did not significantly affect proliferation or migration of PC-3R cells, its combination with SNHG12 knockdown strongly inhibited cell proliferation and migration both in vitro and in the tumor-bearing mice. The expression of ELAVL1 was obviously up-regulated in PC-3R cells, and increased activation level of PI3K/AKT signaling pathway was detected in both PC-3R cells and the xenografts. The effect of SNHG12 knockdown was significantly weakened by treatment with the PI3K activator 740 Y-P. SNHG12 was found to bind to ELAVL1 in PC-3R cells, and mechanistic studies showed that their binding activated the PI3K/AKT signaling pathway to result in DTX resistance in PCa.

Conclusions: SNHG12 knockdown inhibits DTX resistance of PCa cells by reducing SNHG12 binding to ELAVL1 to inhibit the activation the PI3K/AKT signaling pathway.

Large language models (LLMs) are emerging artificial intelligence technologies with strong text and image processing capabilities, offering critical support for the intelligent transformation of healthcare and improving clinical efficiency and quality. This review summarizes the current applications, technical features, and future directions of LLMs in cancer diagnosis, focusing on two key scenarios: automated analysis of textual reports (e.g., imaging, pathology, and case summaries) and multimodal diagnosis combining text and medical images. Findings show that LLMs now perform at a level comparable to general resident physicians in cancer diagnosis but are still incapable of making specialized and precise judgments. They also exhibit application-specific traits, such as parameter-efficient models adapted for grassroots-level scenario and divergent versatility in multilingual report analysis. Future efforts should prioritize developing specialized, practical medical LLMs through optimized fine-tuning strategies, construction of high-quality Chinese medical datasets, and integration with vision-language models to promote the clinical application of these models and increase the accessibility of healthcare resources.

Objectives: To explore pathological and immune cell infiltration characteristics of pigmented pretibial patches in diabetic patients.

Methods: Forty-two diabetic patients undergoing thigh amputation at Tianjin Medical University Chu Hsien-I Memorial Hospital were enrolled. Before the operation, the pretibial skin of the patients were examined and sampled for HE and Masson staining. The thickness of the epidermis and the density of blood vessels in the dermis were compared between patients with and without pigmented pretibial patches. The expressions of VEGFA and VEGFR2 in the skin tissues were detected using Western blotting, and CD4⁺ and CD8⁺ cells and the CD4/CD8 ratio were analyzed with immunohistochemical staining.

Results: Compared with the patients without pigmented pretibial patches, the patients with pigmented pretibial patches showed obvious thickening of the epidermal spinous layer, irregular downward extension of the epidermal projections, hyperkeratosis, melanin deposition in the basal layer, increased capillaries in the dermis, and localized, well-defined inflammatory cell infiltration around the blood vessels. In pigmented pretibial patches group, Masson staining revealed irregular arrangement, thickening and hyaline degeneration of collagen fibers, significantly increased epidermal thickness and blood vessel density in the dermis, increased CD4⁺ cells and the CD4/CD8 ratio, and reduced CD8⁺ cells.

Conclusions: The pigmented pretibial patches in diabetic patients show obvious pathological changes possibly due to vascular and immune abnormalities.

Objectives: To explore the role of miR-221-3p in mediating the positive effects of aerobic exercise on macrophage polarization in the adipose tissues and insulin resistance (IR).

Methods: Sixteen normal C57BL/6J mice and 16 mice with IR induced by high-fat diet (HFD) feeding for 12 weeks were both randomized into sedentary group and exercise group with aerobic exercise training on a treadmill (5 times per week for 8 consecutive weeks). All the mice were examined for changes in body weight, body composition, fasting blood glucose, blood lipid levels, insulin levels, miR-221-3p expression level, mRNA levels of Socs1, Tnf-α and Arg-1, and protein levels of SOCS1, JAK1, p-STAT1, and p-STAT3 in the adipose tissues, and the targeting relationship between miR-221-3p and SOCS1 was validated using dual-luciferase reporter gene assay. In RAW264.7 macrophages, the effects of transfection with miR-221-3p mimic or inhibitor on macrophage polarization were observed.

Results: In mice with normal feeding, aerobic exercise significantly decreased body weight, fat mass, fat percent, fasting blood glucose, serum insulin level, HOMA-IR, and TC and TG levels, and reduced miR-221-3p levels in both the plasma and the adipose tissues. The sedentary IR mice showed significantly increased miR-221-3p levels in both the plasma and adipose tissue, increased protein levels of iNOS, JAK1, and p-STAT1/STAT1, and decreased protein levels of Arg-1, SOCS1 and p-STAT3/STAT3, which were significantly reversed after aerobic exercise intervention. Dual-luciferase reporter gene assays validated the targeting relationship between miR-221-3p and SOCS1. In RAW264.7 macrophages, miR-221-3p overexpression significantly reduced Socs1 and Arg-1 mRNA expression, whereas miR-221-3p inhibition obviously promoted M2 polarization of the macrophages.

Conclusions: Aerobic exercise improves HFD-induced IR in mice possibly by inhibiting miR-221-3p to activate the SOCS1 and JAK/STAT signaling pathway, thereby promoting macrophage M2 polarization and alleviating chronic inflammation in the adipose tissue.

Objectives: To evaluate the effect of exosomes derived from folic acid (FA)-treated infrapatellar fat pad mesenchymal stem cells (IPFP-MSCs) on M1 and M2 polarization of macrophages in vitro.

Methods: Infrapatellar fat pad tissues were obtained from surgical patients without knee osteoarthritis to isolate IPFP-MSCs. The exosomes were extracted from the cell cultures with or without FA treatment and identified by transmission electron microscopy, TEM, NTA and Western blotting. RAW264.7 cells were induced with lipopolysaccharide (LPS) and incubated with exosomes from FA-treated or untreated IPFP-MSCs for 12 h, and Exos uptake was observed using confocal microscopy. The changes in expression levels of IL-1β, IL-6, TNF-α, iNOS, ARG1, MRC1, and CD206 in the macrophages were detected using qRT-PCR, ELISA, flow cytometry and immunofluorescence staining.

Results: The exosomes derived from IPFP-MSCs showed a typical cup shape, were positive for CD9 and CD81, and could be uptaken by macrophages. In LPS-induced macrophages, incubation with exosomes from FA-treated IPFP-MSCs significantly decreased the expressions of IL-1β, IL-6, TNF‑α, and NOS2, and increased the expressions of ARG1 and MRC1. Treatment of the macrophages with exosomes from FA-treated IPFP-MSCs significantly lowered CD86-positive cell percentage, increased CD206-positive cells and the CD206/CD86 ratio, lowered cellular expression of iNOS, and enhanced the expression of CD206.

Conclusions: Exosomes from FA-treated IPFP-MSCs promotes M2 polarization of macrophages more effectively than exosomes from unmodified IPFP-MSCs, suggesting a new exosome modification strategy for targeted treatment of knee osteoarthritis.

Objectives: To explore the therapeutic mechanism of Jiangzhi Quban Recipe (JZQBR) for type 2 diabetes mellitus (T2DM) complicated with hyperlipidemia and validate its clinical efficacy and safety.

Methods: The active components and disease targets of JZQBR were screened using TCMSP and GeneCards databases, followed by protein-protein interaction analysis and GO and KEGG enrichment analyses. In the animal experiments, ApoE^-/- mice were randomized into blank control, model, simvastatin treatment, and low- and high-dose JZQBR groups. In the latter 4 groups, the mice were fed a high-fat diet for 24 weeks with corresponding treatments from Weeks 9 to 24. The changes in body weight, blood glucose, lipids, liver pathology, and inflammatory cytokine expressions of the mice were examined. In the clinical study, 72 T2DM patients with hyperlipidemia were randomized equally into control group for treatment with metformin plus empagliflozin and JZQBR group with additional JZQBR for 12 consecutive weeks.

Results: Network pharmacology identified 65 potential targets, with quercetin, kaempferol, and luteolin as the core components and IL-6, IL-1β, and TNF‑α as the key targets. The targets were enriched mainly in the pathways involving inflammatory responses and diabetic complications. In the ApoE^-/- mouse models, JZQBR treatment dose-dependently improved body weight, blood glucose, and blood lipid profiles, and high-dose JZQBR produced a stronger effect than simvastatin for improving hepatic steatosis and significantly reduced inflammatory cytokine levels. In the clinical trial, 29 patients in JZQBR group and 31 in the control group completed the trial. The patients in JZQBR group showed significant improvements in body weight, FBG, TG, HbA1c, and liver enzymes with significantly lower fasting blood glucose level than the control group. The total effective rates were comparable between the two groups.

Conclusions: JZQBR improves T2DM complicated with hyperlipidemia possibly by multi-target regulation of the inflammation-metabolism network.

Objectives: To explore the therapeutic mechanism of Qingjie Fuzheng granules (QFG) for alleviating 5-fluorouracil (5-FU)-induced skeletal muscle atrophy.

Methods: Male BALB/c mice bearing subcutaneous colorectal cancer CT26 cell xenografts were randomized into control group, model group, and treatment group. The mice in model and treatment groups were given intraperitoneal 5-FU injections every 3 days and treated with daily gavage of saline and QFG for 21 days, respectively; the mice in the control group and normally fed mice were given only saline gavage. Gripping test and hanging test of the mice were performed before and after the treatment, and on day 21, tumor weight and gastrocnemius muscle weight were measured, and histopathology and cell apoptosis in the gastrocnemius muscle were examined with HE staining, transmission electron microscopy and TUNEL assay. ATP content in the muscle was measured, and protein expressions of AMPK, PGC-1α, Cyt c, AIF, Apaf-1, Smac, Bcl-2, Bax, cleaved caspase-3 and cleaved caspase-9 were determined with immunohistochemistry.

Results: The tumor-bearing mice in the control group showed significantly decreased gastrocnemius muscle weight and grip and suspension test scores. The gastrocnemius muscle showed ultrastructure injuries with lowered ATP content, obvious cell apoptosis, decreased expressions of AMPK, PGC-1 α, and Bcl-2, and increased expressions of Bax, Cyto C, AIF, Apaf-1, Smac, cleaved caspase-3 and cleaved caspase-9. These changes were obviously worsened in 5-FU-treated mice, while QFG treatment significantly increased gastrocnemius muscle weight and strength, ameliorated its ultrastructural injuries, reduced cell apoptosis, and reversed the abnormal protein expressions.

Conclusions: QFG alleviates 5-FU-induced skeletal muscle fatigue in tumor-bearing mice by activating the AMPK/PGC-1α pathway and inhibiting mitochondria-dependent apoptosis in the gastrocnemius muscle.