Artificial intelligence (AI), particularly deep learning, has demonstrated remarkable performance in medical imaging across a variety of modalities, including X-ray, computed tomography (CT), magnetic resonance imaging (MRI), ultrasound, positron emission tomography (PET), and pathological imaging. However, most existing state-of-the-art AI techniques are task-specific and focus on a limited range of imaging modalities. Compared to these task-specific models, emerging foundation models represent a significant milestone in AI development. These models can learn generalized representations of medical images and apply them to downstream tasks through zero-shot or few-shot fine-tuning. Foundation models have the potential to address the comprehensive and multifactorial challenges encountered in clinical practice. This article reviews the clinical applications of both task-specific and foundation models, highlighting their differences, complementarities, and clinical relevance. We also examine their future research directions and potential challenges. Unlike the replacement relationship seen between deep learning and traditional machine learning, task-specific and foundation models are complementary, despite inherent differences. While foundation models primarily focus on segmentation and classification, task-specific models are integrated into nearly all medical image analyses. However, with further advancements, foundation models could be applied to other clinical scenarios. In conclusion, all indications suggest that task-specific and foundation models, especially the latter, have the potential to drive breakthroughs in medical imaging, from image processing to clinical workflows.
Humans ; Artificial Intelligence ; Deep Learning ; Diagnostic Imaging/methods* ; Magnetic Resonance Imaging ; Tomography, X-Ray Computed ; Positron-Emission Tomography

Objective:To investigate the distribution of pathogenic bacteria in the ear canal secretions of patients with chronic suppurative otitis media（CSOM）, the changes in the levels of interleukin-8（IL-8） and Toll like receptor 4（TLR4） in the ear canal secretions, and their clinical significance. Methods:This study selected 128 CSOM patients who visited our hospital from January 2022 to February 2024 as the study subjects and recorded them as the CSOM group. Additionally, 135 volunteers who underwent physical examinations at our hospital during the same period were regarded as the control group. Video otoscopy was used to collect and cultivate ear canal secretions, and a fully automated microbial identification instrument was used to identify the bacterial species. ELISA was applied to detect levels of IL-8, TLR4. Multivariate logistic regression was employed to examine the factors that affect the occurrence of CSOM. Pearson correlation was applied to analyze the correlation between IL-8, TLR4 levels and various influencing factors. ROC curve was applied to analyze the diagnostic value of IL-8 and TLR4 levels for the occurrence of CSOM. Z-test was applied to compare the differences in AUC. Results:Among 128 patients, the detection rate was 89.06%, and a total of 181 strains of pathogenic bacteria were cultured, among them, Gram positive bacteria accounted for the highest proportion of 54.14%, followed by Gram negative bacteria, accounting for 34.25%, and finally fungi, accounting for 11.60%. The common bacteria were Staphylococcus aureus （20.44%）, Pseudomonas aeruginosa （13.26%）, and Staphylococcus epidermidis （8.29%）. The resistance of Gram-positive bacteria to penicillin, clindamycin, erythromycin, and amoxicillin is high. Gram-negative bacteria are highly resistant to penicillin, ampicillin and erythromycin. Fungi are resistant to ketoconazole and fluconazole. The levels of IL-8 and TLR4 in CSOM group were higher than those in the control group, and gradually increased with the increase of hearing impairment. （P<0.05）. Elevated levels of IL-8, TLR4 were independent risk factors for the occurrence of CSOM（P<0.05）. The AUC of CSOM diagnosed by IL-8 and TLR4 alone was 0.790 and 0.777, respectively, while the AUC of combined diagnosis was 0.898, which was better than their respective individual diagnoses（both P<0.05）. Conclusion:The distribution of pathogenic bacteria in the ear canal secretions of CSOM patients is mainly Gram positive, with common ones being Staphylococcus aureus and Pseudomonas aeruginosa. The levels of IL-8 and TLR4 in CSOM patients are higher than those in the control group. The higher the levels, the higher the degree of hearing loss, which can be used for clinical diagnosis.
Humans ; Toll-Like Receptor 4/metabolism* ; Interleukin-8/metabolism* ; Otitis Media, Suppurative/metabolism* ; Ear Canal/metabolism* ; Chronic Disease ; Male ; Female ; Adult ; Middle Aged ; Clinical Relevance

ObjectiveTo investigate the individual factors of postural adjustment reaction time and movement time during adaptive equilibrium. MethodsFrom March to December, 2021, 126 healthy adults aged 18 to 80 years were recruited at the First Medical Center of the General Hospital of the Chinese PLA. The DE-A somatosensory balance detection system was used to detect their postural adjustment reaction time (RT) and movement time (MT) as the platform tilting in multiple directions during standing (static) or walking (dynamic). The ages, genders, body mass index (BMI) and physical activity level of them were investigated. ResultsThe age was the only factor independently associated with dynamic RT and MT in all the directions (β > 0.632, P < 0.05). For static MT, as the platform tilting forward, physical activity level (β = -0.143, P < 0.05), BMI (β = 0.154, P < 0.05) and age (β = 0.663, P < 0.05) were the independently associated factors; as the platform tilting leftward, gender (β = -0.173, P < 0.05) and age (β = 0.647, P < 0.05) were the independently associated factors; and age was the only independently associated factor for other directions (β > 0.571, P < 0.05). For the static RT, age was the only independently associated factor for all the directions (β > 0.615, P < 0.05). ConclusionAge is the most important independently factor related to postural adjustment during adaptive equilibrium, and aging may delay the postural adjustment after instability.

Objective:To examine influencing factors for social support satisfaction among disabled elderly people at home in rural areas and to provide evidence for policies on further improving the level of social support and the quality of care for disabled elderly people.Methods:We conducted a self-administered questionnaire survey with 211 disabled elderly people, using a random sampling method in 6 villages and towns in the Southern, Northern and Guanzhong areas of Shaanxi province from June to October 2020.An ordinal logistic regression model was adopted to analyze influencing factors for social support satisfaction among the participants.Results:The survey results indicated that, of the 211 disabled elderly people, 49(23.22%)were unsatisfied with social support, 110(52.13%)were generally satisfied, and 52(24.65%)were satisfied.Ordinal logistic regression analysis showed that the degree of disability, the number of chronic diseases, with or without hospitalization in 2019, the most important daily caregiver, the time needed to the nearest medical institution, and relationships with family members were important factors influencing social support satisfaction for disabled elderly people(all P<0.05).For the elderly with disability, female, a low degree of disability, no hospitalization in 2019, the spouse as the most important caregiver, and a short time needed to the nearest medical institution tended to be associated with a high level of social support satisfaction( OR=2.07, 71.52, 30.75 and 19.22, respectively).Poor relationships with family members were associated with a low level of social support satisfaction( OR=0.39). Conclusions:The satisfaction of disabled elderly people with social support is at a moderate level.For elderly people with disability, the degree of disability, the number of chronic diseases, recent hospitalization, the most important daily caregivers, the time needed to the nearest medical institution, and relationships with family members are important influencing factors for social support satisfaction.

Objective: To explore the possibility that the neck extension in chest CT scanning would make thyroid move upward and reduce thyroid radiation exposure. Methods: The images of 1 994 patients who underwent chest CT examination in Beijing Chaoyang Hospital from March to June in 2017 were compared and analyzed retrospectively. According to different cervical positions, patients were divided into extension group and routine group with 997 cases in each group. In the extension group, the head and neck bended back to the maxillary top position while in the regular group with normal position. The length of thyroid glands exposed to the scanning range and the number of cases of all glands moved out of the scanning range were observed with the first rib head as the baseline counting mark. The differences in exposed length of thyroid gland in primary radiation area between the two groups were compared. Results: The length of exposed thyroid gland by primary radiation in the extension group (4.69±5.68) mm was significantly shorter (U=91 073.5, P<0.05) than that in the routine group (17.16±6.68) mm. In addition, the number of thyroid glands out of scanning range completely in the extension group (n=519, with ratio of 52.1%) was significantly more (U=594.8, P<0.05) than those in the routine group (n=32, with ratio of 3.2%). Conclusions During chest CT scan, the length of thyroid gland in the scanning field can be shortened with neck backbend position and the radioation protection for thyroid gland can be effectively improved.

Objective: To study the effects of patient-controlled intravenous analgesia(PCIA) with Dexmedetomidine, Ropivacaine by local incision infiltration and combined analgesia on inflammatory factors and rapid rehabilitation in infants aged 3-6 months after radical resection of megacolon. Methods: From June 2016 to March 2018, 90 infants aged 3-6 months underwent radical resection of megacolon in Henan Provincial Children′s Hospital as the subjects.According to the postoperative analgesia mode applied to the children, they were divided into Dexmedetomidine group, Ropivacaine group and combined group (Dexmedetomidine PCIA combined with local infiltration of Ropivacaine), 30 cases in each group.The serum levels of interleukin(IL)-6 and IL-10 were measured and recorded 2 hours before operation and 24 hours after operation, and the analgesic scores of 4, 8, 12 and 24 hours after operation were recorded.The rapid recovery index of each group, standard and the number of complications were compared. Results: Compared with 2 hours before operation, the serum levels of IL-6 in the Dexmedetomidine group, Ropivacaine group and combined group [(24.61±1.44) ng/L vs.(13.84±0.65) ng/L, (26.39±1.29) ng/L vs.(13.85±0.67) ng/L, (20.58±2.06) ng/L vs.(13.87±0.63) ng/L], IL-10[(27.63±1.52) ng/L vs.(15.79±1.48) ng/L, (28.29±2.34) ng/L vs.(15.41±1.37) ng/L, (23.21±2.71) ng/L vs.(15.47±1.52) ng/L] were significantly higher (all P<0.05), while the levels of IL-6, IL-10 in combined group were lower than those in Dexmedetomidine group and Ropivacaine group (t_a=9.172, 9.835; t_b=10.221, 11.034, all P<0.05). At 4 h, 8 h, 12 h, 24 h after operation.The analgesic scores in combined group were(1.89±0.23) scores, (1.87±0.15) scores, (1.95±0.17) scores, (2.08±0.18) scores, those of dexmedetomidine group were(2.06±0.24) scores, (2.08±0.18) scores, (2.76±0.29) scores, (2.55±0.31) scores, those of Ropivacaine group were(2.10±0.26) scores, (2.15±0.26) scores, (2.32±0.19) scores, (3.00±0.28) scores, and the analgesic scores in combined group were significantly lower than those in Dexmedetomidine group and Ropivacaine group(t_a=8.526, 9.364, 10.287, 9.521; t_b=8.729, 9.514, 11.037, 9.184, all P<0.05). In comparison with the first exhaust time, anal extubation time, hospitalization days, hospitalization expenses and antibiotic use time in the three groups, it was found that combined group was significantly better than Dexmedetomidine group and Ropivacaine group(t_a=8.315, 8.294, 7.883, 9.261, 10.487; t_b=8.582, 8.329, 7.916, 9.348, 10.862, all P<0.05). The incidence of complications in the combined group, Dexmedetomidine group, and Ropivacaine group were not significantly different in 3 groups(χ²=0.577, P=0.749). Conclusions The continuous anesthesia of Dexmedetomidine combined with local infiltration of Ropivacaine incision can alleviate the postoperative inflammatory reaction and promote the rapid recovery of children.

Objective To investigate the value of Chinese brain template Chinese2020 in detecting gray matter (GM) alterations in patients with Parkinson disease (PD) in Chinese population using voxel-based morphometry (VBM). Methods Totally 15 PD patients (PD group)and 15 gender- and age-matched healthy volunteers (control group) were enrolled. Chinese2020 based on Chinese and MNI152 based on Caucasian were used for spatial normalization, respectively. The deformations induced by spatial normalization and gray matter (GM) proportion in the identified brain areas were compared between Chinese2020 and MNI152. The differences of GM volume between PD group and control group were examined using VBM. Results Using MNI152, GM atrophies in PD patients were detected in bilateral temporal lobe extending into ipsilateral insula/parahippocampus/hippocampus, the left superior occipital cortex/cuneus/precuneus, as well as the right putamen. Besides the above mentioned regions, the right middle frontal cortex atrophy was additionally identified using Chinese2020. Particularly, fewer deformations were found when spatially normalizing Chinese population's brain to Chinese2020 than to MNI152. Moreover, GM percentages of detected regions using Chinese2020 were significantly higher than those using MNI152 (t=2.502, P=0.037). Conclusion CBT should be used in neuroimaging studies of Chinese population, which may achieve fewer deformations during spatial normalization, more GM proportions in the identified clusters and improve the accuracy of detection.

Objective By investigating the obesity of the adult population in Hunan,we explored the relationship between body mass index (BMI) [including waist circumference (WC),waist-to-hip ratio (WHR),waist-to-body ratio (WHtR) and H-type hypertension] in predicting H-type hypertension.Methods A multi-stage random sampling method was used to conduct a questionnaire survey,physical examination and laboratory testing on 4 012 adults ＞ 30 years old in 6 districts of Hunan Province.The receiver operator characteristic (ROC) curve was used to evaluate the predictive effects of BMI,WC,WHR and WHtR on H-type hypertension.Results ROC curve analysis indicated that the area under curve (AUC) of the evaluation of H-type hypertension of WC,WHtR,WHR and BMI in men were:0.637,0.640,0.601,0.639,of which the largest were combination of WC + BMI and WHtR + BMI,both being 0.647.In women,the AUC of individual indexes WC,WHtR,WHR and BMI for evaluating H-type hypertension were 0.660,0.670,0.660 and 0.604 respectively.The combination of WHtR and WHR had the largest AUC,which was 0.675.The WC for evaluating H-type hypertension in youth people had the largest AUC (0.732);the WC + BMI,WHR + BMI for evaluating H-type hypertension in middle-aged people had the largest AUC,all being 0.687;the WC predicted the largest AUC of H-type hypertension in elderly people (0.590).Conclusions WC,WHtR,WHR and BMI are important predictors of H-type hypertension.The optimal indicators for different genders and ages are different,and the combined indicators are better than the individual indicators.

Objective: To explore the acute toxicity of Diquat in mice and to calculate the median lethal dose (LD₅₀) of Diquat to rats and observe the pathological changes of tissues and organs in rats with different concentrations of Diquat. Methods: Diquat solution of 50 mg/kg was prepared freshly with 1 000 mg of Diquat and dilute the solution with water to a total of 20 ml. A total of 99 healthy adult male Wistar rats were randomly divided into part one, part two and control groups. In the first part, 36 rats were randomly divided into 4 groups: 100 mg/kg group, 200 mg/kg group, 300 mg/kg group and 400 mg/kg group, which were treated with 100 mg/kg, 200 mg/kg, 300 mg/kg and 400 mg/kg of Diquat solution by gavage, respectively. The death and symptoms of poisoning after intragastric administration were recorded, and the maximum tolerated dose and absolute lethal dose were measured. In the second part, 54 rats were randomly divided into 6 groups: 200 mg/kg group, 220 mg/kg group, 240 mg/kg group, 260 mg/kg、280 mg/kg group and 300 mg/kg group, whichwere treated with 200 mg/kg, 220 mg/kg, 240 mg/kg, 260 mg/kg, 280 mg/kg and 300 mg/kg of Diquat solution by gavage, respectively. The survival of rats in different concentration of Diquat was observed and the LD₅₀ was calculated by Excel processing the formula of Koch's method. The control group were given equal volume water under the same experimental conditions. And moreover, the lungs, kidneys, hearts, livers, and brain tissues were collected and fixed by formaldehyde, embedded by paraffin, and sectioned for histopathological light microscopy. Results: The maximum tolerated dose was 240 mg/kg and the absolute lethal dose was 300 mg/kg. The LD₅₀ of Diquat for Rats was 280.58 mg/kg. The high-dose group had significantly more organ damage than the low-dose group after diquat poisoning. Conclusion The determination of the half-lethal dose of diquat, at the same time observed multiple organs damaged in rats after the diquat quickly poisoned. Kidneys, lungs and heart might be the main organ which was heavily damaged. With the extension of observation time, the organ damage of rats exposed to small doses gradually stabilized.

Objective To investigate the effect of transcranial direct current stimulation (tDCS) on aphasia recovery after stroke. Meth-ods From April, 2012 to January, 2013, 20 aphasic patients after stroke were enrolled in an A-B experiment design. During phase A, ten times of sham tDCS and language training (five days a week) were implemented, then ten times language training combined with tDCS (five days a week) were implemented in phase B. The treatment lasted for four weeks. Picture naming was measured for all patients before and af-ter treatment both in phase A and phase B. Results The D-value scores of picture naming before and after treatment were significantly more in phase B than in phase A in both treatment items and non-treatment items (t>3.030, P<0.05). Conclusion tDCS could raise the accuracy of picture naming in patients with aphasia after stroke.