Retrieving keyframes most relevant to text from small intestine videos with given labels can efficiently and accurately locate pathological regions. However, training directly on raw video data is extremely slow, while learning visual representations from image-text datasets leads to computational inconsistency. To tackle this challenge, a small bowel video keyframe retrieval based on multi-modal contrastive learning (KRCL) is proposed. This framework fully utilizes textual information from video category labels to learn video features closely related to text, while modeling temporal information within a pretrained image-text model. It transfers knowledge learned from image-text multimodal models to the video domain, enabling interaction among medical videos, images, and text data. Experimental results on the hyper-spectral and Kvasir dataset for gastrointestinal disease detection (Hyper-Kvasir) and the Microsoft Research video-to-text (MSR-VTT) retrieval dataset demonstrate the effectiveness and robustness of KRCL, with the proposed method achieving state-of-the-art performance across nearly all evaluation metrics.
Humans ; Video Recording ; Intestine, Small/diagnostic imaging* ; Machine Learning ; Image Processing, Computer-Assisted/methods* ; Algorithms

Objective:To investigate the value of midtreatment 18F-FDG PET/CT metabolic parameters for predicting the pathological response in patients with locally advanced gastric cancer (LAGC) after neoadjuvant immunochemotherapy (NICT). Methods:Twenty-five LAGC patients (19 males, 6 females, age: (64.8±8.6) years) who underwent 18F-FDG PET/CT after NICT in Henan Cancer Hospital from August 2019 to June 2024 were retrospectively analyzed. The lesion′s ROI was delineated, then the SUV max and metabolic tumor volume (MTV) were measured, and the SUV max was divided by SUV mean of the descending aorta to obtain the tumor-to-background ratio (TBR). Patients underwent surgery after PET/CT imaging. Based on the tumor regression grade (TRG) system by the American Joint Committee on Cancer (AJCC) criteria on surgical specimen, patients were divided into responders (TRG0+ 1) and non-responders (TRG2+ 3). Independent-sample t test, Mann-Whitney U test, one-way analysis of variance, and Kruskal-Wallis rank-sum test were used to compare the differences of data. The predictive efficacy of PET/CT metabolic parameters was assessed by the ROC curve analysis. Results:Postsurgical pathology showed that 9 patients were responders and 16 were non-responders. The SUV max (3.10±1.95) and TBR (2.44±1.54) of primary lesions in responders were lower than those in non-responders (7.40±4.68, 5.85±3.74; t values: -2.61, -2.59, both P<0.05), while the MTV of primary tumors and short diameter and metabolic parameters of positive lymph nodes were not significantly different between those 2 groups ( t=-1.50, Z values: -1.09 to -0.75, all P>0.05). No significant relation was found between PET/CT parameters and pathological differentiation or Lauren classification, or other pathological features ( t values: -1.55 to 1.38, Z values: -1.84 to 0, F values: 0.12-2.43, H values: 0.13-0.98, all P>0.05). ROC curve showed that the cut-off value of SUV max for predicting postoperative TRG was 5.40, and the AUC reached 0.77 (95% CI: 0.56-0.91), with the sensitivity and specificity of 9/16, 9/9, respectively. With TBR=3.54 as the cut-off value, its AUC reached 0.77 (95% CI: 0.56-0.91), and the sensitivity and specificity were 11/16, 8/9, respectively. The sensitivity and specificity of PET/CT for predicting lymph node positivity of patients were 8/12 and 13/13, respectively. Conclusion:Interim 18F-FDG PET/CT metabolic parameters can accurately predict the pathological response of LAGC patients after NICT.

Objective To investigate the value of cranial CT cisternal grading combined with D-dimer(D-D)and Glasgow Coma Scale(GCS)score in predicting the short-term postoperative prog-nosis of patients with severe traumatic brain injury.Methods A total of 165 patients with severe trau-matic brain injury who were treated in the hospital from January 2019 to May 2024 were selected as study subjects,all underwent craniotomy surgery.Postoperative follow-up was conducted for 3 months to analyze the differences in clinical data and preoperative indicators such as cranial CT cisternal grad-ing,D-D levels,and GCS scores between patients with poor and good prognosis.The value of cranial CT cisternal grading,D-D levels,and GCS scores in predicting short-term postoperative poor prognosis in patients with severe traumatic brain injury was also analyzed.Results Compared with patients with good prognosis,patients with poor prognosis had higher proportion of age,cranial CT cisternal grading of Ⅰ to Ⅱ,D-D levels,and GCS scores＜6(P＜0.05).There were no statistically significant differences in C-reactive protein,prothrombin time,activated partial thromboplastin time,international normalized ratio,total cholesterol,triglycerides,high-density lipoprotein cholesterol,and low-density lipoprotein cholesterol levels between patients with poor and good prognosis(P＞0.05).Cranial CT cisternal grading,D-D levels,and GCS scores were influencing factors for short-term postoperative poor prognosis in patients with severe traumatic brain injury(P＜0.05).The area under the curve for poor prognosis by three indicators in combination was 0.941(95％CI,0.906 to 0.975),which was higher than the area under the curve for the individual predictions of cranial CT cisternal grad-ing,D-D levels,and GCS scores(P＜0.05).Conclusion The influencing factors for short-term postoperative prognosis in patients with severe traumatic brain injury include cranial CT cisternal grading,D-D levels,and GCS scores.The model based on these three indicators has certain appli-cation value in predicting patient prognosis.

Objective To compare the diagnostic performance of two syphilis reverse testing protocols recommended by the U.S.Centers for Disease Control and Prevention(USCDC)and the European Centre for Disease Prevention and Control(ECDC)within maternal and child populations,and to explore the factors contributing to false-positive results in chemiluminescent immunoassays(CLIA).Methods A retrospective analysis was performed on serological test results from 109,003 patients collected between 2021 and 2023.All participants were initially screened for specific syphilis antibodies using CLIA,and those with positive results underwent confirmatory testing with both the toluidine red unheated serum test(TRUST)and the Treponema pallidum particle agglutination assay(TPPA).The kappa statistic was employed to assess the diagnostic agreement between the USCDC and ECDC protocols,while also analyzing the distribution patterns of CLIA false-positive results across varying S/CO value ranges and demographic groups.Results Both protocols achieved a syphilis positivity rate of 0.34%.The hospital's high-prevalence syphilis screening program identified no cases with CLIA-positive or TRUST-positive but TPPA-negative results,suggesting high specificity in this population.The two protocols demonstrated perfect agreement with a Kappa value of 1.0,indicating no significant difference in diagnostic performance between maternal and child populations.CLIA exhibited a true positive rate of 71.04%,which was positively correlated with S/CO values:21.88%when 1.010.The highest false positive rate for CLIA(38.75%)was observed among young adults,while the largest proportion of false positives-reaching 40.00%—was associated with pregnancy and abortion status.Conclusions The two reverse detection methods for syphilis exhibit comparable diagnostic efficacy in women and children from populations with low syphilis prevalence.CLIA demonstrates high sensitivity;however,when the S/CO ratio is less than 10,particularly in adult women of childbearing age,clinicians should remain vigilant for potential false-positive results to prevent misdiagnosis and unnecessary medical interventions.