AI Group Research Articles

Effect on comprehension of an AI patient-friendly hospital discharge letter: a quasi-RCT

Abstract Background During transition from hospital to home care, discharge letters serve as a bridge of information between hospital physicians, general practitioners, and patients. However, the complexity of these letters often hinders patient comprehension, leading to medication errors and adverse outcomes. We assessed the effect on patients’ comprehension and satisfaction of using AI to transform discharge letters into patient-friendly versions. Methods An online quasi-randomized controlled parallel-group trial was conducted in Italy. A synthetic discharge letter for a chronic obstructive pulmonary disease patient was drafted. GPT-4 rephrased the letter into an easier-to-understand version. Participants were assigned to receive either the original or AI-generated letter and completed comprehension assessments and satisfaction survey. Results Out of 460, 166 participants received the original letter, and 175 the AI version. The AI group showed significantly higher comprehension scores (median 4;IQR 2-4 vs. 2; IQR1-2, p < 0.001) and outperformed the control group in understanding diagnosis, medications, and antibiotic instructions. Perceived satisfaction was higher in the AI group across all measures (p < 0.001). Health literacy was a good predictor of comprehension in both arms (p = 0.013; p = 0.015). Subgroup analysis revealed the AI letter nearly bridged the comprehension gap between non-healthcare individuals and healthcare professionals receiving the original letter. Conclusions AI-optimized discharge letters improved patient understanding, and satisfaction compared to traditional letters. The study underscores the influence of sociodemographic factors like healthcare experience and health literacy on comprehension, advocating for adaptive communication strategies. These findings contribute to discussions about integrating AI in healthcare to develop patient-centered communication approaches. Key messages • AI-generated patient-friendly discharge letters significantly improved comprehension and satisfaction compared to traditional letters. • Integrating AI in healthcare communication can help bridge comprehension gaps and promote patient-centered care.

Elevated circulating group-2 innate lymphoid cells expressing activation markers and correlated tryptase AB1 levels in active ascariasis

IntroductionAscaris lumbricoides infection is one of the most common soil-transmitted helminthiasis and IgE response to this helminth may increase the risk of asthma, bronchial hyperreactivity and atopy. There is not enough evidence showing the role of group-2 innate lymphoid cells (ILC2) in the pathogenesis of helminth infections in humans. Here, we aimed to investigate and characterize the influence of Ascaris lumbricoides infection on circulating ILCs in endemically exposed subjects.MethodsNon-infected (NI; n=16) and Ascaris-infected (AI; n=16) subjects from an endemic area were included. Two consecutive stool samples from each subject were examined by Kato-Katz to define parasite infection. Antibodies to the ABA-1 antigen of Ascaris and Ascaris extract were measured by ELISA. ILC subsets and their activation markers (CD25, CD69, thymic stromal lymphopoietin receptor (TSLPR) were evaluated in its PBMC by flow cytometry. Proximity extension assay (PEA) was performed to explore plasma proteins associated to infection.ResultsNo significant differences in the relative or absolute frequencies of total ILCs, ILC1, ILC2 and ILC3 cells were observed regarding the infection status. However, within AI group, IgE-sensitized subjects to ABA-1 had higher frequencies and counts of ILC2 (p<0.05). Frequencies of CD25+, CD69+ and TSLPR+ ILC2 were higher in AI compared to the NI (p<0.01). Additionally, egg burden was positively correlated with CD69+ ILC2 frequencies (r=0.67; p=0.005). Tryptase alpha/beta 1 (TPSAB1), GP6 and several plasma proteins associated with cell growth and granulocyte chemotaxis were highly expressed in the AI group (p<0.05). Interestingly, TPSAB1 levels were positively correlated with ILC2 expressing activation markers frequencies, egg burden and IgE levels against Ascaris.DiscussionAscaris infection is associated with increased expression of ILC2 activation markers and TPSAB1, which may contribute to the type-2 response.

Gut microbiota alterations in adolescent idiopathic scoliosis: a comparison study with healthy control and congenital scoliosis.

This study aims to compare the composition of GM isolated from individuals with AIS or congenital scoliosis (CS) and age-matched control (Ctr). A total of 48 patients with AIS, 24 patients with CS, and 31 healthy individuals were recruited as the discovery cohort, and 9 pairs of siblings where one was affected by AIS were recruited as the validation cohort. The GM profile was determined with 16S rRNA sequencing, and the alpha-diversity and beta-diversity metrics were performed with Mothur. Linear discriminant analysis (LDA) analysis was performed to identify the enriched species. The α diversity (Chao1 index) was significantly lower in AIS patients with low BMI (< 18.5) than those with normal BMI. The PcoA analysis showed a trend of clustering of GM in AIS compared to that in Ctr and CS groups (r2 = 0.0553, p = 0.001). METASTAT analysis showed Cellulomonadaceae was significantly enriched in AIS groups compared to CS and Ctr. LDA analysis showed 9 enriched species in AIS patients. Compared to Ctr, two species including Hungatella genus and Bacteroides fragilis were significantly enriched, while the Firmicutes versus Bacteroidetes (F/B) ratio and the Ruminococcus genus were significantly decreased in AIS but not CS groups. The significantly reduced F/B ratio and Ruminococcus genus in AIS were replicated in the validation cohort. Our study elucidated an association between low BMI and GM diversity in AIS patients. The reduced F/B ratio and Ruminococcus genus in AIS patients were identified and validated in 9 pairs of AIS patients and their unaffected siblings. Our pilot results may help understand the anthropometric discrepancy in these patients and support a possible role of GM in the pathogenesis of AIS.

Trust Dynamics in Financial Decision Making: Behavioral Responses to AI and Human Expert Advice Following Structural Breaks.

This study explores the trust dynamics in financial forecasting by comparing how individuals perceive the credibility of AI and human experts during significant structural market changes. We specifically examine the impact of two types of structural breaks on trust: Additive Outliers, which represent a single yet significant anomaly, and Level Shifts, which indicate a sustained change in data patterns. Grounded in theoretical frameworks such as attribution theory, algorithm aversion, and the Technology Acceptance Model (TAM), this research investigates psychological responses to AI and human advice under uncertainty. This experiment involved 157 participants, recruited via Amazon Mechanical Turk (MTurk), who were asked to forecast stock prices under different structural break scenarios. Participants were randomly assigned to either the AI or human expert treatment group, and the experiment was conducted online. Through this controlled experiment, we find that, while initial trust levels in AI and human experts are comparable, the credibility of advice is more severely compromised following a structural break in the Level Shift condition, compared to the Additive Outlier condition. Moreover, the decline in trust is more pronounced for human experts than for AI. These findings highlight the psychological factors influencing decision making under uncertainty and offer insights into the behavioral responses to AI and human expert systems during structural market changes.

Metabolic profile and glycemic response in fully-grown sows born using assisted reproductive technologies

The aim of the present work was to gain insight into the metabolism of pigs derived from assisted reproductive technologies during their adulthood. Approximately 4h after feeding, a blood sample was taken from 3.5 year old sows born by artificial insemination (AI group, n = 7) and transfer of in vitro produced embryos (IVP group, n = 11) to determine the physiological concentrations of the main biomarkers of carbohydrates (glucose and lactate), proteins (albumin, creatinine and urea) and lipids (cholesterol and triglycerides). Four weeks later, an oral glucose tolerance test (OGTT; 1.75g glucose/kg body weight) was performed after an overnight fast and 1h of water withdrawal. Blood samples were obtained prior (T = 0 min; fasting conditions) and 15, 30, 45, 60, 90, 120, 150, 180, 210 and 240 min after glucose intake. At each time point, glycemia was measured immediately using glucometer test strips, and serum was collected to determine the above metabolites along with insulin and glucagon. After OGTT, the area under the curve (AUC) between sampling times and homeostasis model assessment of insulin resistance (HOMA) indices were calculated. Under physiological conditions, the concentration of metabolites studied was similar between AI and IVP sows. In both groups, fasting decreased cholesterol and increased triglycerides and urea (P < 0.001). However, creatinine and lactate were similar in both groups under physiological and fasting conditions. The expected increase in albuminemia and decrease in glycaemia after fasting was only observed in IVP sows. OGTT revealed a different glucose curve pattern (monophasic in AI and biphasic in IVP group), a lower mean concentration of cholesterol, glucose, lactate, triglycerides in IVP compared to AI pigs (P < 0.01), and a higher mean concentration of albumin, creatinine and insulin in IVP compared to AI group (P < 0.05). On the contrary, no differences were found between groups for mean serum glucagon and urea levels, nor for glucose homeostasis indices HOMA-IR and HOMA-%B. The AUC differed between groups at several time points with larger AUC for creatinine, and smaller AUC for glucose, glucagon, and triglycerides, in IVP pigs than in AI pigs at 180–210 min (P < 0.05). In conclusion, under physiological conditions the metabolic profile of fully-grown AI and IVP sows is similar and within normal ranges. Glucose challenge revealed differences in metabolic and insulin responses between groups but with normal glucose tolerance in both cases.

Efficacy of a single 5mg zoledronic acid infusion in preventing bone loss and fracture in postmenopausal women with breast cancer.

Chemotherapy-induced bone loss (CTIBL) is common among breast cancer patients, requiring comprehensive assessment and intervention. Zoledronic acid, a strong inhibitor of bone resorption, is effective in CTIBL management, though information on dosing and intervals, particularly the efficacy of the 5mg annual dose for osteoporosis in breast cancer patients, is limited. In this 12-month prospective observational study, 85 breast cancer patients were divided into three groups: 17 received no treatment, 17 received tamoxifen, and 51 received anastrozole or letrozole (AI). Post-surgery, patients were administered a single 5mg dose of zoledronic acid and monitored over 12months for changes in bone mineral density (BMD), fracture rates, and biochemical markers. Initially, the AI group was the oldest, averaging 59.1 ± 8.7years. At baseline, no significant differences in variables, except age, were observed. After 12months, BMD increased in all groups following a single zoledronic acid dose, with the smallest increase in the AI group at the lumbar spine: no treatment (2.4% ± 6.1%), tamoxifen (2.6% ± 3.4%), AI (0.6% ± 14.5%) (p = 0.778). CTx and P1NP levels were consistently suppressed up to 12months post-treatment, with smaller reductions in the AI group. There were no significant differences in fracture or bone metastasis rates among groups. A single infusion of 5mg zoledronic acid was effective in increasing bone density in breast cancer patients. However, AI-treated patients showed less improvement in vertebral bone mineral density and biochemical markers. Further long-term studies with larger cohorts are needed.

Evaluation of Middle Cerebral Artery Culprit Plaque Inflammation in Ischemic Stroke Using CAIPIRINHA-Dixon-TWIST Dynamic Contrast-Enhanced Magnetic Resonance Imaging.

Middle cerebral artery (MCA) plaques are a leading cause of ischemic stroke (IS). Plaque inflammation is crucial for plaque stability and urgently needs quantitative detection. To explore the utility of Controlled Aliasing in Parallel Imaging Results in Higher Acceleration (CAIPIRINHA)-Dixon-Time-resolved angiography With Interleaved Stochastic Trajectories (TWIST) (CDT) dynamic contrast-enhanced MRI (DCE-MRI) for evaluating MCA culprit plaque inflammation changes over stroke time and with diabetes mellitus (DM). Prospective. Ninety-four patients (51.6 ± 12.23 years, 32 females, 23 DM) with acute IS (AIS; N = 43) and non-acute IS (non-AIS; 14 days < stroke time ≤ 3 months; N = 51). 3-T, CDT DCE-MRI and three-dimensional (3D) Sampling Perfection with Application optimized Contrast using different flip angle Evolution (3D-SPACE) T1-weighted imaging (T1WI). Stroke time (from initial IS symptoms to MRI) and DM were registered. For 94 MCA culprit plaques, Ktrans from CDT DCE-MRI and enhancement ratio (ER) from 3D-SPACE T1WI were compared between groups with and without AIS and DM. Shapiro-Wilk test, Bland-Altman analysis, Passing and Bablok test, independent t-test, Mann-Whitney U test, Chi-squared test, Fisher's exact test, receiver operating characteristics (ROC) with the area under the curve (AUC), DeLong's test, and Spearman rank correlation test with the P-value significance level of 0.05. Ktrans and ER of MCA culprit plaques were significantly higher in AIS than non-AIS patients (Ktrans = 0.098 s-1 vs. 0.037 s-1; ER = 0.86 vs. 0.55). Ktrans showed better AUC for distinguishing AIS from non-AIS patients (0.87 vs. 0.75) and stronger negative correlation with stroke time than ER (r = -0.60 vs. -0.34). DM patients had significantly higher Ktrans and ER than non-DM patients in IS and AIS groups. Imaging by CDT DCE-MRI may allow to quantitatively evaluate MCA culprit plaques over stroke time and DM. 2 TECHNICAL EFFICACY: Stage 2.

Peri-procedural anesthesia and patient pain experience in pulmonary vein isolation by means of very high-power short-duration radiofrequency ablation.

Very high-power short-duration (vHPSD) temperature-controlled radiofrequency ablation (vHPSD) may reduce ablation times and improve patient tolerability, permitting pulmonary vein (PV) isolation under mild conscious sedation. We evaluated of the anesthetic drugs use and patients' pain experience during vHPSD PV isolation. Fifty-eight patients, with paroxysmal and persistent atrial fibrillation (AF), treated with QDot Micro catheter and vHPSD (90 w for 4s) (vHPSD group), were compared with the last 33 patients treated with a surround flow contact force-sensing catheter guided by the ablation index (450 anteriorly at 50 W, 330 posteriorly at 40 W) (AI group). Anesthetic drugs use was compared as well as pain experience, measured using a 0-10 scale. All PVs were acutely isolated. Procedural time (78 ± 10min vs 84 ± 12min, p = 0.012), fluoroscopy time (369 ± 139s vs 441 ± 172s, p = 0.03), and RF time in the vHPSD group (8.3 ± 2.1min) were shorter than in the AI group (25 ± 11min, p < 0.001). Only 4 patients experienced an access site-related vascular complication (groin hematoma). Midazolam was required in 36 (62%) vHPSD group patients vs 31 (94%) AI group patients (p < 0.001). Fentanyl was required in 4 (7%) vHPSD group patients vs 25 (76%) AI group patients (p < 0.001). No patients required general anesthesia. Twenty-two (38%) vHPSD group patients underwent PV isolation without any anesthetic drug. Pain experience was significantly lower in vHPSD group (4.9 ± 2 vs 6.6 ± 1.8, p < 0.001). vHPSD radiofrequency ablation for PVI can be performed under conscious sedation using only benzodiazepine in most of patients without compromising patient pain experience.

Surgical Repair vs Splenectomy in Patients With Severe Traumatic Spleen Injuries

The spleen is often removed in laparotomy after traumatic abdominal injury, with little effort made to preserve the spleen. To explore the association of surgical management (splenic repair vs splenectomy) with outcomes in patients with traumatic splenic injuries undergoing laparotomy and to determine whether splenic repair is associated with lower mortality compared with splenectomy. This is a trauma registry-based cohort study using the American College of Surgeons Trauma Quality Improvement Program database from January 2013 to December 2019. Participants included adult patients with severe splenic injuries (Abbreviated Injury Scale [AIS] grades 3-5) undergoing laparotomy after traumatic injury within 6 hours of admission. Data analysis was performed from April to August 2023. Splenic repair vs splenectomy in patients with severe traumatic splenic injury. The primary outcome was in-hospital mortality. Outcomes were compared using different statistical approaches, including 1:1 exact matching with consecutive conditional logistic regression analysis as the primary analysis and multivariable logistic regression, propensity score matching, and inverse-probability weighting as sensitivity analyses. A total of 11 247 patients (median [IQR] age, 35 [24-52] years; 8179 men [72.7%]) with a severe traumatic splenic injury undergoing laparotomy were identified. Of these, 10 820 patients (96.2%) underwent splenectomy, and 427 (3.8%) underwent splenic repair. Among patients who underwent an initial splenic salvage procedure, 23 (5.3%) required a splenectomy during the subsequent hospital stay; 400 patients with splenic preservation were matched with 400 patients who underwent splenectomy (matched for age, sex, hypotension, trauma mechanism, AIS spleen grade, and AIS groups [0-2, 3, and 4-5] for head, face, neck, thorax, spine, and lower and upper extremity). Mortality was significantly lower in the splenic repair group vs the splenectomy group (26 patients [6.5%] vs 51 patients [12.8%]). The association of splenic repair with lower mortality was subsequently verified by conditional regression analysis (adjusted odds ratio, 0.4; 95% CI, 0.2-0.9; P = .03). Multivariable logistic regression, propensity score matching, and inverse-probability weighting confirmed this association. In this retrospective cohort study, splenic repair was independently associated with lower mortality compared with splenectomy during laparotomy after traumatic splenic injury. These findings suggest that efforts to preserve the spleen might be indicated in selected cases of severe splenic injuries.

Can ChatGPT Fool the Match? Artificial Intelligence Personal Statements for Plastic Surgery Residency Applications: A Comparative Study

Introduction: Personal statements can be decisive in Canadian residency applications. With the rise in AI technology, ethical concerns regarding authenticity and originality become more pressing. This study explores the capability of ChatGPT in producing personal statements for plastic surgery residency that match the quality of statements written by successful applicants. Methods: ChatGPT was utilized to generate a cohort of personal statements for CaRMS (Canadian Residency Matching Service) to compare with previously successful Plastic Surgery applications. Each AI-generated and human-written statement was randomized and anonymized prior to assessment. Two retired members of the plastic surgery residency selection committee from the University of British Columbia, evaluated these on a 0 to 10 scale and provided a binary response judging whether each statement was AI or human written. Statistical analysis included Welch 2-sample t tests and Cohen's Kappa for agreement. Results: Twenty-two personal statements (11 AI-generated by ChatGPT and 11 human-written) were evaluated. The overall mean scores were 7.48 (SD 0.932) and 7.68 (SD 0.716), respectively, with no significant difference between AI and human groups ( P = .4129). The average accuracy in distinguishing between human and AI letters was 65.9%. The Cohen's Kappa value was 0.374. Conclusions: ChatGPT can generate personal statements for plastic surgery residency applications with quality indistinguishable from human-written counterparts, as evidenced by the lack of significant scoring difference and moderate accuracy in discrimination by experienced surgeons. These findings highlight the evolving role of AI and the need for updated evaluative criteria or guidelines in the residency application process.

The effectiveness of endoscopic diagnostics of colon tumors using artificial intelligence: prospective tandem study

AIM: to estimate the effectiveness of a medical decision support system based on artificial intelligence in the endoscopic diagnosis of benign tumors during tandem study.PATIENTS AND METHODS: from October to December 2023, a single-center comparative tandem study of medical decision support system based on artificial intelligence “ArtInCol” was done. The first stage was a traditional colonoscopy under sedation, the second one — colonoscopy using AI. A pairwise comparison of the main indicators of the effectiveness was made.RESULTS: in the primary endpoint, the polyp detection rate (PDR) in the traditional colonoscopy group was 40.6% vs 56.4% in the AI group, p = 0.0001 (RR = 1.39; 95% CI: 1.04–1.87). The mean number of lesions detected (MPP) was 1.63 (± 1.2) vs 2.47 (± 1.8) in the AI group (mean difference = 0.84; (95% CI: 1.07–0.61).CONCLUSION: the study demonstrated the effectiveness of the original medical decision support system for benign colon tumors detection in real clinical practice. The further stage, a multicenter randomized trial is needed.

Relevance of timing intake of everolimus (EVE) combined with adjuvant endocrine therapy (ET) in patients (pts) with high-risk early breast cancer: Results of a sub-study of the UCBG-UNIRAD trial.

542 Background: We reported the benefit of taking evening tamoxifen compared to morning/afternoon as adjuvant treatment for high-risk breast cancers in the UNIRAD phase III trial (NCT01805271) (1). Here, we examine the relevance of EVE timing intake in this trial. Methods: 1,277 pts with high-risk HR+/ HER2- primary BC were randomly assigned to adjuvant ET with placebo or EVE (1). Patients prospectively reported in a diary the daily timing for ET and EVE intake among four 6-h slots (06:00 to 11:59 (morning), 12:00 to 17:59 (afternoon), 18:00 to 23:59 (evening), or 24:00 to 05:59 (night). The association between EVE, ET timing and disease-free survival (DFS) was a prespecified secondary endpoint of the trial. Results: EVE timing was recorded by 513 of 632 patients (81.2%). Patients declaring EVE intake in the morning (n= 248, 48 %) or in the afternoon (n=52, 10%) were older than those declaring evening (n=207, 40 %) or night (n=6,1%) intakes ( p&lt;0.001). They were more likely to be postmenopausal (p=0.001) and to take AI rather than tamoxifen (p = 0.03). Only 17 patients (3%) changed EVE intake timing throughout the study. EVE intake dialy times were binned into 2 categories for analyses: morning/afternoon group (n=300) and evening/night group (n = 213). The patients who took morning/afternoon EVE chose another timing for ET in 38% of the cases compared to 17.1% in evening/Night EVE group (p&lt;0.01). The timing of EVE intake was not significantly associated with definitive EVE withdrawal ( p=0.62) in the whole population. Time to EVE withdrawal was significantly longer in the patients taking tamoxifen rather than AI (17.2 vs 10.9 months, p = 0.0047). Prolonged duration of EVE treatment was associated with evening/night intakes of both EVE and tamoxifen, compared to other drug timing modalities (p = 0.04). With a median follow-up of 85months, 15 patients relapsed locally (3%), 55 developed metastases (11%), 36 patients died (7%). Overall, EVE intake timing was not associated with DFS (HR=0.84 [95% CI, 0.53 - 1.35], p=0.4). There was a significant interaction between EVE timing intake and ET type ( P interaction= 0.001). Patients on evening/night EVE combined with tamoxifen had a reduced likelihood of relapse compared to those taking morning/afternoon EVE (HR = 0.17 [0.05- 0.59], p = 0.005). In contrast, no significant EVE effect was found in the AI group (HR = 1.56 [0.92 2.64], p = 0.1). Evening/Night intake of EVE plus tamoxifen was independently associated with DFS (HR 0.13 [0.03-0.6], p = 0.002) after adjustment for age, nodal status, tumor size, ET timing intake, and EVE dose (2.5, 5 or 10 mg). Conclusions: These exploratory results are consistent with a role of circadian rhythm in the efficacy of EVE when given with tamoxifen in pts with high risk early breast cancer. 1. S. Giacchetti et al, Abst 546, ASCO 2023. Clinical trial information: NCT01805271 .

Laparotomy for Traumatic Splenic Injury: Should we Increase our Efforts to Preserve the Spleen?

Abstract Background In trauma laparotomy the spleen is removed very liberally, and often little effort is made for splenic preservation. Aims The aim of this study was to assess the effect of surgical management (splenic repair vs. splenectomy) on outcomes in patients undergoing trauma laparotomy. Methods This is a trauma registry (TQIP) study (2013–2019), including adult patients with severe splenic injuries (grades III-V) undergoing trauma laparotomy. Characteristics and outcomes of patients who underwent splenic repair vs. splenectomy within 6 hours of admission were compared using different statistical approaches, i.e. 1:1 exact matching with consecutive conditional logistic regression analysis as primary analysis, multivariable logistic regression, propensity score matching and inverse-probability weighting as sensitivity analyses. The primary outcome was in-hospital mortality. Results A total of 11,247 patients with a severe splenic injury undergoing trauma laparotomy within 6 hours of admission were identified. Of these, 10,820 (96.2%) patients underwent splenectomy and 427 (3.8%) patients splenic repair within 6 hours of admission. Among patients who underwent an initial splenic salvage procedure, 23 (5.3%) patients underwent a splenectomy during the subsequent hospital stay. 400 patients with splenic preservation, were matched with 400 patients who underwent splenectomy (matched for age, sex, hypotension, trauma mechanism as well as AIS spleen grades 3,4,5 and AIS groups (AIS 0-2, AIS 3, AIS 4-5) for head, face, neck, thorax, spine, and lower &amp; upper extremity). The mortality was significantly lower in the group of patients with splenic repair (6.5% vs 12.8%, p=0.002). The mortality benefit was subsequently verified by conditional regression analysis (aOR 0.4, CI 95% 0.2; 0.9). Multivariable logistic regression, propensity score matching, and inverse-probability weighting confirmed the lower mortality in the splenic repair group. Conclusion In trauma laparotomy splenic repair was independently associated with lower mortality compared with splenectomy. Splenectomy was required in approximately 5% of the patients after initial splenic repair. The findings provide evidence that efforts for spleen preservation in trauma laparotomy should be considered in selected cases.

Assessing the Impact of AI-Enhanced Diagnostic Tools on the Treatment Planning of Orthodontic Cases: An RCT

ABSTRACT Background: Orthodontic treatment planning involves the precise assessment of dental and skeletal anomalies, which can be facilitated by AI-enhanced diagnostic tools. Materials and Methods: A total of 100 orthodontic cases were included in this RCT. Patients were randomly assigned to two groups: an AI-enhanced diagnostic group and a traditional diagnostic group. The AI-enhanced diagnostic group underwent orthodontic assessment with the aid of AI-powered software, which provided automated cephalometric analysis, 3D model evaluations, and treatment suggestions. The traditional diagnostic group received conventional diagnostic assessments by orthodontists. The primary outcome measures included treatment planning accuracy, treatment time, and patient satisfaction. Secondary outcomes included the number of appointments required and treatment cost. Results: The AI-enhanced diagnostic group demonstrated a significantly higher accuracy in treatment planning compared to the traditional diagnostic group (P &lt; 0.05). The AI group also required fewer appointments (mean ± SD: 10.2 ± 2.1 vs. 12.8 ± 3.4) and had a shorter treatment time (mean ± SD: 14.6 ± 3.2 months vs. 18.9 ± 4.5 months) (P &lt; 0.001 for both comparisons). Additionally, patient satisfaction scores were higher in the AI group (mean ± SD: 9.2 ± 0.6 vs. 8.1 ± 0.8) (P &lt; 0.001). However, the AI-enhanced diagnostic group had a slightly higher treatment cost. Conclusion: AI-enhanced diagnostic tools significantly enhance the accuracy of treatment planning in orthodontic cases, leading to reduced treatment time, fewer appointments, and increased patient satisfaction.

Assesment of Generative AI abilities to diagnose and propose treatment in comparison with psychiatrists from Poland and Tunisa

Introduction Increasing popularity of Generative AI systems such as GPT provides us with new dilemmas concerning the future of diagnosis and novel tools to improve daily psychiatrits’s work.ObjectivesThe aim of the study was to assess the abilities of generative AI to diagnose and propose treatment in comparison with real psychiatrists and performing a Turing test.Methods We examined the ability to diagnose and propose treatment of various Generative AI versions (CHatGPT/CHATGPTpro etc.) and then compare the results with 10 clinicians performing the same task. Then a group of 10 psychiatry specialists not involved in the first evaluation assessed wether the diagnose and treatment were established by Generative AI or a clinician.ResultsThe reults showed that the generative AI systems were able to provide valid diagnosis in most of the cases with favour to newer and most proficient version of CHATGPT. Proposed treatment results were less accurate. The comparison between human and AI group was hard to accurately acces, with tendency to favouring psychiatrists group assesment as the right decision.There is huge need to further explore the possibilites and limitations of Generative AI use in psychiatry.ConclusionsThere is huge need to further explore the possibilites and limitations of Generative AI use in psychiatry.Disclosure of InterestNone Declared

Combined detection of serum IL-6 and CEA contributes to the diagnosis of lung adenocarcinoma in situ.

Effective discrimination of lung adenocarcinoma (LUAD) in situ (AIS) from benign pulmonary nodules (BPN) is critical for the early diagnosis of AIS. Our pilot study in a small cohort of 90 serum samples has shown that serum interleukin 6 (IL-6) detection can distinguish AIS from BPN and health controls (HC). In this study, we intend to comprehensively define the diagnostic value of individual and combined detection of serum IL-6 related to the traditional tumor markers carcinoembryonic antigen (CEA) and cytokeratin 19 fragment (CYFRA21-1) for AIS. The diagnostic performance of serum IL-6 along with CEA and CYFRA21-1 were evaluated in a large cohort of 300 serum samples by a chemiluminescence immunoassay and an electrochemiluminescence immunoassay. A training set comprised of 65 AIS, 65 BPN, and 65 HC samples was used to develop the predictive model for AIS. Data obtained from an independent validation set was applied to evaluate and validate the predictive model. In the training set, the levels of serum IL-6 and CEA in the AIS group were significantly higher than those in the BPN/HC group (P < 0.05). There was no significant difference in serum CYFRA21-1 levels between the AIS group and the BPN/HC group (P> 0.05). Serum IL-6 and CEA levels for AIS patients showed an area under the curve (AUC) of 0.622 with 23.1% sensitivity at 90.7% specificity, and an AUC of 0.672 with 24.6% sensitivity at 97.6% specificity, respectively. The combination of serum IL-6 and CEA presented an AUC of 0.739, with 60.0% sensitivity at 95.4% specificity. The combination of serum IL-6 and CEA showed an AUC of 0.767 for AIS patients, with 57.1% sensitivity at 91.4% specificity in the validation set. IL-6 shows potential as a prospective serum biomarker for the diagnosis of AIS, and the combination of serum IL-6 with CEA may contribute to increased accuracy in AIS diagnosis. However, it is worth noting that further research is still necessary to validate and optimize the diagnostic efficacy of these biomarkers and to address potential sensitivity limitations.

Value of the Air Bronchogram Sign and Other Computed Tomography Findings in the Early Diagnosis of Lung Adenocarcinoma.

The present study aims to explore the application value of the air bronchogram (AB) sign and other computed tomography (CT) signs in the early diagnosis of lung adenocarcinoma (LUAD). The pathological information and CT images of 130 patients diagnosed with N 0 and M 0 solitary pulmonary nodules (diameter ≤3 cm) and treated with surgical resection in our hospital between June 2021 and June 2022 were analyzed. The patients were divided into the benign pulmonary nodule (BPN) group (14 cases), the AIS group (30 cases), the MIA group (10 cases), and the IAC group (76 cases). Among the 116 patients with AIS and LUAD, 96 showed an AB sign. Among the 14 patients with BPN, only 4 patients showed an AB sign. The average CT value and maximum diameter were significantly higher in the IAC group than in the AIS and MIA groups. In the BPN group, 5 patients had an average CT value of >80 HU. Among all LUAD-based groups, there was only 1 patient with a CT value of >60 HU. The identification of the AB sign based on CT imaging facilitates the differentiation between benign and malignant nodules. The CT value and maximum diameter of pulmonary adenocarcinoma nodules increase with the increase of the malignancy degree. The nodule type, CT value, and maximum diameter are useful for predicting the pathological type and prognosis. If the average CT value of pulmonary nodules is >80 HU, LUAD may be excluded.

Accelerated Early Childhood Growth Is Associated With the Development of Earlier Adrenarche and Puberty.

Small birth size and increased postnatal growth have been associated with earlier timing of adrenarche and puberty, but it is not well known whether these factors alone or together lead to earlier maturation. This work aimed to search for different growth trajectories using a clustering approach to analyze the effects of birth size and postnatal growth on adrenarchal and pubertal development. Altogether 351 children (48% girls) were examined prospectively at ages 6 to 9 and 9 to 11 years. Birth and early-growth data were collected retrospectively. Main outcome measures included clinical signs of adrenarche and puberty, and serum androgen concentrations (dehydroepiandrosterone, dehydroepiandrosterone sulfate, androstenedione, testosterone). We detected 4 clusters with different birth sizes and postnatal growth trajectories: 1) children with average birth size and increased postnatal growth (AI), 2) children with small birth size and increased postnatal growth (SI), 3) children with average birth size and postnatal growth (AA), and 4) children with small birth size and average postnatal growth (SA). Thelarche at age 9 to 11 was most common and serum androgens at ages 6 to 9 and 9 to 11 years were highest in girls belonging to the AI and SI groups. Similar patterns in the onset of puberty and in androgen levels were not seen in the SA group. Increased early growth and weight gain predict higher serum androgen concentrations and earlier onset of puberty in girls. Adrenarche and puberty do not appear to be shifted earlier in children with small birth size who do not have catch-up growth.

Autonomous Artificial Intelligence vs Artificial Intelligence–Assisted Human Optical Diagnosis of Colorectal Polyps: A Randomized Controlled Trial

BackgroundArtificial intelligence (AI)-based optical diagnosis systems (CADx) have been developed to allow pathology prediction of colorectal polyps during colonoscopies. However, CADx systems have not yet been validated for autonomous performance. Therefore, we conducted a trial comparing Autonomous AI to AI assisted human (AI-H) optical diagnosis. MethodsWe performed a randomized non-inferiority trial of patients undergoing elective colonoscopies in one academic institution. Patients were randomized int: 1) Autonomous AI-based CADx optical diagnosis of diminutive polyps without human input; 2) endoscopists performed optical diagnosis of diminutive polyps after seeing the real-time CADx diagnosis. Primary outcome was accuracy in optical diagnosis in both arms using pathology as gold standard. Secondary outcomes included agreement with pathology for surveillance intervals. Results467 patients were randomized (238 patients/158 polyps in the Autonomous AI group; 229 patients/179 polyps in the AI-H group). Accuracy for optical diagnosis was 77.2% (95%Confidence Interval 69.7-84.7) in the Autonomous AI group and 72.1% (95%CI 65.5-78.6) in the AI-H group (p=0.86). For high confidence diagnoses, accuracy for optical diagnosis was 77.2% (95%CI 69.7-84.7) in the Autonomous AI group and 75.5% (95%CI 67.9-82.0) in the AI-H group. Autonomous AI had statistically significantly higher agreement with pathology-based surveillance intervals compared to AI-H (91.5% [95%CI 86.9-96.1] vs 82.1% [95%CI 76.5-87.7]; p=0.016). ConclusionsAutonomous AI-based optical diagnosis exhibits non-inferior accuracy to endoscopist-based diagnosis. Both Autonomous AI and AI-H exhibited relatively low accuracy for optical diagnosis, however Autonomous AI achieved higher agreement with pathology-based surveillance intervals.

A Comparative Study of Responses to Retina Questions from Either Experts, Expert-Edited Large Language Models, or Expert-Edited Large Language Models Alone

ObjectiveTo assess the quality, empathy, and safety of expert edited large language model (LLM), human expert created and LLM responses to common retina patient questions DesignRandomized, masked multicenter study ParticipantsTwenty-one common retina patient questions were randomly assigned among 13 retina specialists. Each expert created a response (Expert) and then edited a LLM (ChatGPT-4)-generated response to that question (Expert+AI), timing themselves for both tasks. Five LLMs (ChatGPT-3.5, ChatGPT-4, Claude 2, Bing, Bard) also generated responses to each question. The original question along with anonymized and randomized Expert+AI, Expert and LLM responses were evaluated by the other experts who did not write an expert response to the question. Evaluators judged quality and empathy (very poor, poor, acceptable, good, or very good) along with safety metrics (incorrect information, likelihood to cause harm, extent of harm, and missing content). Main OutcomeMean quality and empathy score, proportion of responses with incorrect information, likelihood to cause harm, extent of harm, and missing content for each response type ResultsThere were 4008 total grades collected (2608 for quality and empathy; 1400 for safety metrics), with significant differences in both quality and empathy (p<0.001, p<0.001) between LLM, Expert and Expert+AI groups. For quality, Expert+AI (3.86 +/- 0.85) performed the best overall while GPT-3.5 (3.75 +/- 0.79) was the top performing LLM. For empathy, GPT-3.5 (3.75 +/- 0.69) had the highest mean score followed by Expert+AI (3.73 +/- 0.63). By mean score, Expert placed fourth out of seven for quality and sixth out of seven for empathy. For both quality (p<0.001) and empathy (p<0.001), expert-edited LLM responses performed better than expert-created responses. There were time savings for an expert-edited LLM response vs. expert-created response (p=0.02). ChatGPT-4 performed similar to Expert for Inappropriate Content (p=0.35), Missing Content (p=0.001), Extent of Possible Harm (p=0.356), and Likelihood of Possible Harm (p=0.129). Conclusions and RelevanceIn this randomized, masked, multicenter study, LLM responses were comparable to experts in terms of quality, empathy and safety metrics, warranting further exploration of their potential benefits in clinical settings.