Artificial Intelligence Group Research Articles

Effective abdominal ultrasonographic detection of pancreatic cystic lesions using artificial intelligence‐assisted noise reduction

AbstractIntroductionPatients with intraductal papillary mucinous neoplasms (IPMNs) exhibit a high incidence of pancreatic cancer. An IPMN presents as a pancreatic cystic lesion (PCL). Artificial intelligence (AI) can be used to enhance ultrasonographic images. We retrospectively evaluated the utility of AI‐assisted ultrasonography in terms of PCL detection.MethodsIn total, 969 patients who underwent artificial narrow intelligence‐assisted transabdominal ultrasound examinations (AI group) were compared with 943 patients who underwent non‐AI‐assisted examinations (control group). The morphological changes and the frequencies of newly detected PCLs were compared between the groups.ResultsThe PCL detection rate was significantly higher in the AI group than in the control group (1.25 ± 0.36 and 0.21 ± 0.15%, respectively; p = .008). AI‐assisted ultrasound evaluation was independently associated with enhanced PCL detection (odds ratio: 5.83, 95% confidence interval: 1.29–26.29, p = .02). PCLs smaller than 8 mm and all PCLs in the pancreatic tail were detected in the AI group.PCL morphological changes were more common in the AI group than in the control group (63.6 ± 14.5 and 16.7 ± 15.2%, respectively; p = .04).ConclusionAI‐assisted transabdominal ultrasonography enhances the PCL detection rate. The use of AI is likely to increase the IPMN detection rate.

Efficacy of artificial intelligence in reducing miss rates of GI adenomas, polyps, and sessile serrated lesions: a meta-analysis of randomized controlled trials

This study aims to discern if the utilization of artificial intelligence (AI) in the course of endoscopic procedures can significantly diminish both the adenoma miss rate (AMR) and polyp miss rate (PMR), compared to standard endoscopy. We performed an extensive search of various databases, encompassing, PubMed, EMBASE, Cochrane Library, Web of Science, and Scopus until January 2023. The search terms used were artificial intelligence, machine learning, deep learning, transfer machine learning, computer-assisted diagnosis, convolutional neural networks (CNN), gastrointestinal (GI) endoscopy, endoscopic image analysis, polyp, adenoma, and neoplasms. Our aim was mainly to explore the impact of AI on the AMR, PMR, and sessile serrated lesion miss rate (SSLMR). A total of 7 randomized controlled trials (RCTs) were included in this meta-analysis. Pooled AMR was markedly lower in the AI group as opposed to the non-AI group (pooled RR, 0.46; 95% CI, 0.36-0.59, P<0.001). Likewise, PMR was also reduced in the AI group in contrast with the non-AI control (pooled RR: 0.43; 95% CI, 0.27-0.69, P<0.001). The results showed that AI decreased the miss rate of sessile serrated lesions (pooled RR, 0.43; 95% CI, 0.20-0.92, P<0.05) and diminutive adenomas (pooled RR, 0.49; 95% CI, 0.26-0.93) during endoscopy, but no significant effect was observed for advanced adenomas (pooled RR, 0.48; 95% CI, 0.17-1.37, P=0.17). The average number of polyps (Hedges'g=-0.486, 95% CI=-0.697 to -0.274, P=0.000) and adenomas (Hedges'g=-0.312, 95% CI=-0.551 to -0.074, P=0.01) detected during the second procedure also favored AI. However, the AI implementation did not lead to a prolonged withdrawal time (P>0.05). This meta-analysis suggested that AI technology led to significant reduction of miss rates for GI adenomas, polyps, and sessile serrated lesions (SSL) during endoscopy surveillance. These results underscore the potential of AI to improve the accuracy and efficiency of GI endoscopy procedures.

Use of Voice-Based Conversational Artificial Intelligence for Basal Insulin Prescription Management Among Patients With Type 2 Diabetes

Optimizing insulin therapy for patients with type 2 diabetes can be challenging given the need for frequent dose adjustments. Most patients receive suboptimal doses and do not achieve glycemic control. To examine whether a voice-based conversational artificial intelligence (AI) application can help patients with type 2 diabetes titrate basal insulin at home to achieve rapid glycemic control. In this randomized clinical trial conducted at 4 primary care clinics at an academic medical center from March 1, 2021, to December 31, 2022, 32 adults with type 2 diabetes requiring initiation or adjustment of once-daily basal insulin were followed up for 8 weeks. Statistical analysis was performed from January to February 2023. Participants were randomized in a 1:1 ratio to receive basal insulin management with a voice-based conversational AI application or standard of care. Primary outcomes were time to optimal insulin dose (number of days needed to achieve glycemic control), insulin adherence, and change in composite survey scores measuring diabetes-related emotional distress and attitudes toward health technology and medication adherence. Secondary outcomes were glycemic control and glycemic improvement. Analysis was performed on an intent-to-treat basis. The study population included 32 patients (mean [SD] age, 55.1 [12.7] years; 19 women [59.4%]). Participants in the voice-based conversational AI group more quickly achieved optimal insulin dosing compared with the standard of care group (median, 15 days [IQR, 6-27 days] vs >56 days [IQR, >29.5 to >56 days]; a significant difference in time-to-event curves; P = .006) and had better insulin adherence (mean [SD], 82.9% [20.6%] vs 50.2% [43.0%]; difference, 32.7% [95% CI, 8.0%-57.4%]; P = .01). Participants in the voice-based conversational AI group were also more likely than those in the standard of care group to achieve glycemic control (13 of 16 [81.3%; 95% CI, 53.7%-95.0%] vs 4 of 16 [25.0%; 95% CI, 8.3%-52.6%]; difference, 56.3% [95% CI, 21.4%-91.1%]; P = .005) and glycemic improvement, as measured by change in mean (SD) fasting blood glucose level (-45.9 [45.9] mg/dL [95% CI, -70.4 to -21.5 mg/dL] vs 23.0 [54.7] mg/dL [95% CI, -8.6 to 54.6 mg/dL]; difference, -68.9 mg/dL [95% CI, -107.1 to -30.7 mg/dL]; P = .001). There was a significant difference between the voice-based conversational AI group and the standard of care group in change in composite survey scores measuring diabetes-related emotional distress (-1.9 points vs 1.7 points; difference, -3.6 points [95% CI, -6.8 to -0.4 points]; P = .03). In this randomized clinical trial of a voice-based conversational AI application that provided autonomous basal insulin management for adults with type 2 diabetes, participants in the AI group had significantly improved time to optimal insulin dose, insulin adherence, glycemic control, and diabetes-related emotional distress compared with those in the standard of care group. These findings suggest that voice-based digital health solutions can be useful for medication titration. ClinicalTrials.gov Identifier: NCT05081011.

Abstract 17815: Effectiveness of Artificial Intelligence in Triaging Transmissions From Cardiac Implantable Electronic Devices

Background: Monitoring of cardiac implantable electronic devices (CIED) may cause tremendous burden on device clinics. Multiple transmissions for device alerts pose challenges in terms of triaging and clinical action. In this study we compared the effectiveness of artificial Intelligence (AI) as compared to Device technicians (DTs) in triaging the data from CIED. Methods: The study included remote transmissions from CIEDs being monitored at 78 clinics across USA from Jan to April 2023. Transmissions were initially processed by AI and were either dismissed, forwarded to the DT for review or forwarded directly to the device clinic. The data forwarded to the DT was analyzed and was either dismissed or forwarded to the device clinic with and without alerts. Transmissions were further classified based on urgency of response needed. About 15% of all transmissions processed by AI and DT were randomly examined by EP for accuracy. Results: A total of 42,534 patients with CIEDs generated 273,435 transmissions. Of these, 21% (57,345) transmissions were dismissed by AI - 17.8% (38,530) without alerts and 33.1% (18,815) with alerts. Of the transmissions dismissed, alerts for AT/AF represented the largest proportion in DT group, whereas alerts for pacing percentage was most seen with AI (table 1). Similar trends were seen with transmissions that were forwarded to clinics with pacing percentage dominating in AI group, and AT/AF detection in the DT group (table 2). Conclusion: In our study, compared to DT, AI algorithm was found to be more accurate in overall triage of remote transmission data and has the potential to improve the clinical efficiency of device clinics and care of patients with CIEDs.

Cephalometric analysis performance discrepancy between orthodontists and an artificial intelligence model using lateral cephalometric radiographs.

The purpose of the present clinical study was to compare the Ricketts and Steiner cephalometric analysis obtained by two experienced orthodontists and artificial intelligence (AI)-based software program and measure the orthodontist variability. A total of 50 lateral cephalometric radiographs from 50 patients were obtained. Two groups were created depending on the operator performing the cephalometric analysis: orthodontists (Orthod group) and an AI software program (AI group). In the Orthod group, two independent experienced orthodontists performed the measurements by performing a manual identification of the cephalometric landmarks and a software program (NemoCeph; Nemotec) to calculate the measurements. In the AI group, an AI software program (CephX; ORCA Dental AI) was selected for both the automatic landmark identification and cephalometric measurements. The Ricketts and Steiner cephalometric analyses were assessed in both groups including a total of 24 measurements. The Shapiro-Wilk test showed that the data was normally distributed. The t-test was used to analyze the data (α = 0.05). The t-test analysis showed significant measurement discrepancies between the Orthod and AI group in seven of the 24 cephalometric parameters tested, namely the corpus length (p = 0.003), mandibular arc (p < 0.001), lower face height (p = 0.005), overjet (p = 0.019), and overbite (p = 0.022) in the Ricketts cephalometric analysis and occlusal to SN (p = 0.002) and GoGn-SN (p < 0.001) in the Steiner cephalometric analysis. The intraclass correlation coefficient (ICC) between both orthodontists of the Orthod group for each cephalometric measurement was calculated. Significant discrepancies were found in seven of the 24 cephalometric measurements tested between the orthodontists and the AI-based program assessed. The intra-operator reliability analysis showed reproducible measurements between both orthodontists, except for the corpus length measurement. The artificial intelligence software program tested has the potential to automatically obtain cephalometric analysis using lateral cephalometric radiographs; however, additional studies are needed to further evaluate the accuracy of this AI-based system.

Artificial intelligence aided precise detection of local recurrence on MRI for nasopharyngeal carcinoma: a multicenter cohort study

MRI is the routine examination to surveil the recurrence of nasopharyngeal carcinoma, but it has relatively lower sensitivity than PET/CT. We aimed to find if artificial intelligence (AI) could be competent pre-inspector for MRI radiologists and whether AI-aided MRI could perform better or even equal to PET/CT. This multicenter study enrolled 6916 patients from five hospitals between September 2009 and October 2020. A 2.5D convolutional neural network diagnostic model and a nnU-Net contouring model were developed in the training and test cohorts and used to independently predict and visualize the recurrence of patients in the internal and external validation cohorts. We evaluated the area under the ROC curve (AUC) of AI and compared AI with MRI and PET/CT in sensitivity and specificity using the McNemar test. The prospective cohort was randomized into the AI and non-AI groups, and their sensitivity and specificity were compared using the Chi-square test. The AI model achieved AUCs of 0.92 and 0.88 in the internal and external validation cohorts, corresponding to the sensitivity of 79.5% and 74.3% and specificity of 91.0% and 92.8%. It had comparable sensitivity to MRI (e.g., 74.3% vs. 74.7%, P=0.89) but lower sensitivity than PET/CT (77.9% vs. 92.0%, P<0.0001) at the same individual-specificities. The AI model achieved moderate precision with a median dice similarity coefficient of 0.67. AI-aided MRI improved specificity (92.5% vs. 85.0%, P=0.034), equaled PET/CT in the internal validation subcohort, and increased sensitivity (81.9% vs. 70.8%, P=0.021) in the external validation subcohort. In the prospective cohort of 1248 patients, the AI group had higher sensitivity than the non-AI group (78.6% vs. 67.3%, P=0.23), albeit nonsignificant. In future randomized controlled trials, a sample size of 3943 patients in each arm would be required to demonstrate the statistically significant difference. The AI model equaled MRI by expert radiologists, and AI-aided MRI by expert radiologists equaled PET/CT. A larger randomized controlled trial is warranted to demonstrate the AI's benefit sufficiently. The Sun Yat-sen University Clinical Research 5010 Program (2015020), Guangdong Basic and Applied Basic Research Foundation (2022A1515110356), and Guangzhou Science and Technology Program (2023A04J1788).

Evaluating AI rib fracture detections using follow-up CT scans

PurposeThis study compares the results of Artificial Intelligence (AI) diagnosis of rib fractures using initial CT and follow-up CT as the final diagnostic criteria, and studies AI-assisted diagnosis in improving the detection rate of rib fractures. MethodsA retrospective study was conducted on 113 patients who underwent initial and follow-up CT scans due to trauma. The initial and follow-up CT were used as diagnostic criteria, respectively. All images were transmitted to the AI software (V2.1.0, Huiying Medical Technology Co., Beijing, China) for rib fracture detection. The radiologist group (Group 1), AI group (Group 2), and Radiologist with AI group (Group 3) reviewed CT images at an interval of one month, recorded and compared the differences in the sensitivity and specificity for diagnosing rib fractures. Results589 and 712 rib fractures were diagnosed by the initial and follow-up CT, respectively. The initial CT diagnosis failed to detect 127 rib fractures, resulting in a missed rate of 17.84%. In addition, four normal ribs were mistakenly identified as being fractured. The follow-up CT was regarded as the diagnostic standard for rib fractures. The sensitivity and specificity were 82.16% and 99.80% for Group 1, 79.35% and 84.90% for Group 2, and 91.57% and 99.70% for Group 3. The sensitivity of Group 3 was higher than that of Group 1 and Group 2 (p < 0.05). The specificity was lower for Group 2 compared with Group 1 and Group 3 (p < 0.05). ConclusionAI-assisted diagnosis improved the detection rate of rib fractures, the follow-up CT should be used for the diagnosis standard of rib fractures, and AI misdiagnoses can be greatly reduced when a radiologist reviews the diagnosis.

P-193 Comparison of ICSI outcomes and euploidy rates between AI and non-AI sperm selection

Abstract Study question Can artificial intelligence (AI) assisted sperm selection improve intracytoplasmic sperm injection (ICSI) outcomes and euploidy rates? Summary answer AI selected spermatozoa showed no statistical significance in various measures of ICSI outcomes but an increasing trend in euploidy rate can be observed. What is known already All ICSI practitioners would like to have an objective tool for the selection of sperm to be injected as it is one of the factors in producing successful ICSI. Sperm selection assistant (SiD: (IVF 2.0 Limited, UK) is a real-time AI spermatozoa identifier that assesses all spermatozoa in a visual field based on their motility patterns individually. Mendizabal-Ruiz et. al. (2022) have reported promising data showing SiD may reduce subjectivity in sperm selection, which can potentially be an alternative method to our current practice. Study design, size, duration From June 2022 to December 2022, 89 ICSI cycles (mean age: 35.3; age range: 26.0–47.0) were done in Alpha IVF &amp; Women’s Specialists. A sibling AI (Group A) and non-AI (Group B) assisted sperm selection study with ICSI was performed on the oocytes retrieved from these cycles. Of which, 76 cycles were planned for Preimplantation Genetic Testing for Aneuploidies (PGT-A) on utilizable blastocysts. Participants/materials, setting, methods All matured oocytes were injected using Piezo-ICSI method (Prime Tech Ltd, Japan). Half of the oocytes of each cycle were injected with SiD selected sperm whereas the other half by manual selection. All injected oocytes were cultured up to day-7 and trophectoderm biopsy for PGT-A screening (IonTorrent, USA) was done on utilizable blastocysts prior to vitrification (Cryotec, Japan). The fertilization, blastulation, utilization and euploidy rates were assessed in both groups. Main results and the role of chance In Group A, the fertilization rate (2PN), abnormal fertilization rate ( &amp;gt; 2PN), blastulation rate from 2PN, blastocyst utilization rate from 2PN and euploidy rate from utilizable blastocysts were 81.5% (419/514), 2.1% (11/514) 77.1 (323/419), 57.5% (241/419) and 45.2% (71/157) respectively. In Group B, the fertilization rate (2PN), abnormal fertilization ( &amp;gt; 2PN), blastulation rate from 2PN, blastocyst utilization rate from 2PN and euploidy rate from utilizable blastocysts were 79.7% (427/536), 2.8% (15/536), 75.9% (324/427), 59.0% (252/427) and 35.5% (60/169) respectively. No significant differences were observed in the fertilization rate (p = 0.4828), abnormal fertilization rate (p = 0.5545), blastulation rate from 2PN (p = 0.6861), blastocyst utilization rate (p = 0.7278) and euploidy rate (p = 0.0898). Although the ICSI outcomes are comparable between the two groups, AI assisted sperm selection showed an increasing trend in euploidy rates. SiD may be considered as a tool to assist embryologist in selecting better spermatozoa during ICSI procedures. Limitations, reasons for caution The patient sample size was small but further studies with a larger sample size certainly seem to be justified, particularly in association with PGTA on utilizable blastocysts. The value of SiD shall be greatly enhanced when spermatozoon morphology can also be measured. Wider implications of the findings This is more evidence suggesting that SiD may improve ICSI outcomes and euploidy rate when compared to manual selection by an embryologist. In addition, it could be used to train and assist embryologists learning ICSI procedures. Trial registration number not applicable

Blinded, randomized trial of sonographer versus AI cardiac function assessment

Artificial intelligence (AI) has been developed for echocardiography1–3, although it has not yet been tested with blinding and randomization. Here we designed a blinded, randomized non-inferiority clinical trial (ClinicalTrials.gov ID: NCT05140642; no outside funding) of AI versus sonographer initial assessment of left ventricular ejection fraction (LVEF) to evaluate the impact of AI in the interpretation workflow. The primary end point was the change in the LVEF between initial AI or sonographer assessment and final cardiologist assessment, evaluated by the proportion of studies with substantial change (more than 5% change). From 3,769 echocardiographic studies screened, 274 studies were excluded owing to poor image quality. The proportion of studies substantially changed was 16.8% in the AI group and 27.2% in the sonographer group (difference of −10.4%, 95% confidence interval: −13.2% to −7.7%, P < 0.001 for non-inferiority, P < 0.001 for superiority). The mean absolute difference between final cardiologist assessment and independent previous cardiologist assessment was 6.29% in the AI group and 7.23% in the sonographer group (difference of −0.96%, 95% confidence interval: −1.34% to −0.54%, P < 0.001 for superiority). The AI-guided workflow saved time for both sonographers and cardiologists, and cardiologists were not able to distinguish between the initial assessments by AI versus the sonographer (blinding index of 0.088). For patients undergoing echocardiographic quantification of cardiac function, initial assessment of LVEF by AI was non-inferior to assessment by sonographers.

Abstract OT3-18-03: The PRAIM study: A prospective multicenter observational study of an integrated Artificial Intelligence system with live monitoring

Abstract Background. Several retrospective studies have illustrated the potential clinical benefit of artificial intelligence (AI) systems for breast cancer screening. Some systems optimize normal mammography examination triaging, while others aim to improve cancer detection. However, no AI system has shown specificity high enough to replace human radiologists, suggesting that AI should play a different role in the breast screening pathway. The decision-referral approach is a promising alternative that has demonstrated the most potential to improve radiologist screening sensitivity and specificity while reducing workload. This collaborative human-AI approach combines AI pre-screening to triage normal examinations and post-screening to prevent missed cancers. The actual performance of decision-referral, including the interaction with human radiologists, can ideally be evaluated in a prospective real-world setting. Trial design. The PRAIM (PRospective, multicenter observational study of an integrated AI system with live monitoring to support breast cancer screening) study (German Trial Register: DRKS00027322) is a prospective controlled observational non-inferiority study to compare the use of CE-marked screening software including AI support (Vara) via the decision-referral approach, with standard screening for women participating in the German breast cancer screening program. Ethics approval was obtained from the University of Lübeck Research Ethics Committee (22-043). Examinations assessed by readers using Vara are compared to examinations without Vara (control). Eligibility criteria and target accrual. Women ages 50 to 69 years old undergoing biennial breast cancer screening within the national screening program are eligible for inclusion. We expect the inclusion of approximately 400,000 women within the inclusion period of 1.5 years. Statistical methods. The primary outcome is the screen-detected cancer rate, defined as biopsy-confirmed cancer diagnoses per 1000 screening examinations. For each screening site, rates over the prospective observation period are calculated for examinations read with AI and without. To control for systematically different screen-detected cancer rates across screening sites, a historical 5-year rate is computed for each site and subtracted from the corresponding prospective rates. Non-inferiority of the screen-detected cancer rate for the AI group compared to the control group is evaluated with a weighted, mixed-effects linear regression model. AI is considered as non-inferior if the lower bound of the two-sided 95 % confidence interval for the estimated difference in screen-detected cancer rates of AI and non-AI group is not below -10 %, which corresponds to a deviation of -0.6 screen-detected cancers per 1000 examinations. Citation Format: Danalyn Byng, Nora Eisemann, Dominik Schüler, Stefan Bunk, Christian Leibig, Moritz Brehmer, Susanne Elsner, Alexander Katalinic. The PRAIM study: A prospective multicenter observational study of an integrated Artificial Intelligence system with live monitoring [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr OT3-18-03.

Abstract WP81: Validation Of Artificial Intelligence To Limit Delays In Acute Stroke Treatment And Endovascular Therapy (VALIDATE)

Introduction: Analyses from early- and late-window thrombectomy trials for acute large vessel occlusion (LVO) stroke have consistently demonstrated a strong time-dependent treatment effect for best outcomes. The utilization of an artificial intelligence (AI)-based care coordination platform to support LVO diagnosis and treatment has the potential to assist in reducing door to puncture times and improve outcomes. Hypothesis: Utilization of an AI-based care platform in the management of LVO patients may significantly decrease door to Neurointerventionalist (NIR) notification time. Methods: Acute stroke consultations seen by TeleSpecialists, LLC physicians in the emergency department in 166 facilities (17 states) that utilized VIZ.AI software or did not have an AI software from December 1, 2021 through March 31, 2022 were extracted from the Telecare TM database. Facilities that neurology does not initiate contact with NIR were excluded. The encounters were analyzed by age, ethnicity, last known normal, arrival time, initial telemedicine stroke code activation time (TCA), candidacy for thrombolytics, door to needle time (DTN), advanced imaging, presence of LVO, time NIR accepted, premorbid modified Rankin Score, and NIHSS score. Patients were divided into two groups (utilization and non-utilization of the AI platform) and median door to NIR notification times were compared. There was a 4 minute shorter time to TCA at AI hospitals. Results: A total of 14,116 patients (8,557 AI and 5,559 non-AI) were included. The median door to NIR notification time for AI was 50 minutes (40, 82) and non-AI was 89.5 minutes (59.3, 122), p &lt;0.001. There was a small but significantly shorter time (3 minutes) from door to TCA at AI hospitals. Median DTN times for thrombolytics was also lower with AI, 40 minutes (30, 52) vs 44 minutes (32, 57.5) for non-AI, p=0.018. The AI group had more advanced imaging performed, a higher percentage of LVOs identified, and a higher percentage accepted by NIR. Conclusion: Hospital utilization of an AI-based care coordination platform was associated with a significant decrease of 39.5 minutes in the time to NIR contact, increase in patients taken for intervention, and lower DTN times for thrombolytics.

Assessment of the Role of Artificial Intelligence in the Association Between Time of Day and Colonoscopy Quality

Time of day was associated with a decline in adenoma detection during colonoscopy. Artificial intelligence (AI) systems are effective in improving the adenoma detection rate (ADR), but the performance of AI during different times of the day remains unknown. To validate whether the assistance of an AI system could overcome the time-related decline in ADR during colonoscopy. This cohort study is a secondary analysis of 2 prospective randomized controlled trials (RCT) from Renmin Hospital of Wuhan University. Consecutive patients undergoing colonoscopy were randomly assigned to either the AI-assisted group or unassisted group from June 18, 2019, to September 6, 2019, and July 1, 2020, to October 15, 2020. The ADR of early and late colonoscopy sessions per half day were compared before and after the intervention of the AI system. Data were analyzed from March to June 2022. Conventional colonoscopy or AI-assisted colonoscopy. Adenoma detection rate. A total of 1780 patients (mean [SD] age, 48.61 [13.35] years, 837 [47.02%] women) were enrolled. A total of 1041 procedures (58.48%) were performed in early sessions, with 357 randomized into the unassisted group (34.29%) and 684 into the AI group (65.71%). A total of 739 procedures (41.52%) were performed in late sessions, with 263 randomized into the unassisted group (35.59%) and 476 into the AI group (64.41%). In the unassisted group, the ADR in early sessions was significantly higher compared with that of late sessions (13.73% vs 5.70%; P = .005; OR, 2.42; 95% CI, 1.31-4.47). After the intervention of the AI system, as expected, no statistically significant difference was found (22.95% vs 22.06%, P = .78; OR, 0.96; 95% CI; 0.71-1.29). Furthermore, the AI systems showed better assistance ability on ADR in late sessions compared with early sessions (odds ratio, 3.81; 95% CI, 2.10-6.91 vs 1.60; 95% CI, 1.10-2.34). In this cohort study, AI systems showed higher assistance ability in late sessions per half day, which suggests the potential to maintain high quality and homogeneity of colonoscopies and further improve endoscopist performance in large screening programs and centers with high workloads.

Comparison of artificial intelligence-assisted informed consent obtained before coronary angiography with the conventional method: Medical competence and ethical assessment.

At the time of informed consent (IC) for coronary angiography (CAG), patients' knowledge of the process is inadequate. Time constraints and a lack of personalization of consent are the primary causes of inadequate information. This procedure can be enhanced by obtaining IC using a chatbot powered by artificial intelligence (AI). In the study, patients who will undergo CAG for the first time were randomly divided into two groups, and IC was given to one group using the conventional method and the other group using an AI-supported chatbot, chatGPT3. They were then evaluated with two distinct questionnaires measuring their satisfaction and capacity to understand CAG risks. While the satisfaction questionnaire was equal between the two groups (p = 0.581), the correct understanding of CAG risk questionnaire was found to be significantly higher in the AI group (<0.001). AI can be trained to support clinicians in giving IC before CAG. In this way, the workload of healthcare professionals can be reduced while providing a better IC.

Performance and comparison of artificial intelligence and human experts in the detection and classification of colonic polyps

ObjectiveThe main aim of this study was to analyze the performance of different artificial intelligence (AI) models in endoscopic colonic polyp detection and classification and compare them with doctors with different experience.MethodsWe searched the studies on Colonoscopy, Colonic Polyps, Artificial Intelligence, Machine Learning, and Deep Learning published before May 2020 in PubMed, EMBASE, Cochrane, and the citation index of the conference proceedings. The quality of studies was assessed using the QUADAS-2 table of diagnostic test quality evaluation criteria. The random-effects model was calculated using Meta-DISC 1.4 and RevMan 5.3.ResultsA total of 16 studies were included for meta-analysis. Only one study (1/16) presented externally validated results. The area under the curve (AUC) of AI group, expert group and non-expert group for detection and classification of colonic polyps were 0.940, 0.918, and 0.871, respectively. AI group had slightly lower pooled specificity than the expert group (79% vs. 86%, P < 0.05), but the pooled sensitivity was higher than the expert group (88% vs. 80%, P < 0.05). While the non-experts had less pooled specificity in polyp recognition than the experts (81% vs. 86%, P < 0.05), and higher pooled sensitivity than the experts (85% vs. 80%, P < 0.05).ConclusionThe performance of AI in polyp detection and classification is similar to that of human experts, with high sensitivity and moderate specificity. Different tasks may have an impact on the performance of deep learning models and human experts, especially in terms of sensitivity and specificity.

Impact of Artificial Intelligence on Colonoscopy Surveillance After Polyp Removal: A Pooled Analysis of Randomized Trials

Artificial intelligence (AI) tools aimed at improving polyp detection have been shown to increase the adenoma detection rate during colonoscopy. However, it is unknown how increased polyp detection rates by AI affect the burden of patient surveillance after polyp removal. We conducted a pooled analysis of 9 randomized controlled trials (5 in China, 2 in Italy, 1 in Japan, and 1 in the United States) comparing colonoscopy with or without AI detection aids. The primary outcome was the proportion of patients recommended to undergo intensive surveillance (ie, 3-year interval). We analyzed intervals for AI and non-AI colonoscopies for the U.S. and European recommendations separately. We estimated proportions by calculating relative risks using the Mantel-Haenszel method. A total of 5796 patients (51% male, mean 53 years of age) were included; 2894 underwent AI-assisted colonoscopy and 2902 non-AI colonoscopy. When following U.S. guidelines, the proportion of patients recommended intensive surveillance increased from 8.4% (95% CI, 7.4%-9.5%) in the non-AI group to 11.3% (95% CI, 10.2%-12.6%) in the AI group (absolute difference, 2.9% [95% CI, 1.4%-4.4%]; risk ratio, 1.35 [95% CI, 1.16-1.57]). When following European guidelines, it increased from 6.1% (95% CI, 5.3%-7.0%) to 7.4% (95% CI, 6.5%-8.4%) (absolute difference, 1.3% [95% CI, 0.01%-2.6%]; risk ratio, 1.22 [95% CI, 1.01-1.47]). The use of AI during colonoscopy increased the proportion of patients requiring intensive colonoscopy surveillance by approximately 35% in the United States and 20% in Europe (absolute increases of 2.9% and 1.3%, respectively). While this may contribute to improved cancer prevention, it significantly adds patient burden and healthcare costs.

The study on artificial intelligence (AI) colonoscopy in affecting the rate of polyp detection in colonoscopy: A single centre retrospective study

AbstractAimThe aim of this study was to evaluate if the application of Artificial Intelligence (AI) Colonoscopy CLN (ENDO‐AID) could increase the polyp detection rate (PDR).Methods and MaterialsA single center retrospective study was performed in Tin Shui Wai Hospital. PDR in CLN from 11/2020 to 03/2021 after the application of ENDO‐AID (AI group) was compared to the cases from 12/2019 to 11/2020 before the application of the practice (non‐AI group). Procedures were performed by a single endoscopist with experience in performing &gt; 3,000 CLN. Variables, such as patients' demographic data, indications, incidence of PDR, Boston Bowel Preparation Scale BBPS, withdrawal time, post CLN complication rate between the 2 groups, were compared. Categorical and continuous variables were analyzed by using the Chi‐Square test (Fisher exact test if appropriate) and Mann‐Whitney test respectively. Results were considered to be significant if p‐value &lt; 0.05.ResultsTotal 234 patients were recruited. 115 patients (49.1%) were in the non‐AI group while 119 patients (50.9%) were in the AI group. The mean age of the non‐AI was higher than the AI group (65.3 vs 59.2, p&lt; 0.001*), otherwise, there was no significant difference in sex (p = 0.05), percentage of smokers (20.8% vs 27.7%, p = 0.22), past medical history of IBD (0 vs 0, p = 1.0), family history of colorectal cancer (9 vs 9, p = 0.94), indications for CLN (e.g. follow up CLN for polyp/ cancer, per‐rectal bleeding, altered bowel habit etc. p &gt; 0.05), BBPS (7.88 vs 8.04, p = 0.217), withdrawal time (7.65 min vs 7.48 min, p = 0.935), completion rate (95.6% vs 98.3%, p = 0.27) and complication rate (0% in both groups,p=1.0) between groups. In the contrary, PDR was significantly higher in the AI group than the non‐AI group (64.7% vs 46.0%, p = 0.003*). Besides, adenoma detection rate was also found significantly higher in the AI group than the non‐AI group (52.9% vs 37.4%, p = 0.017*).ConclusionsAI CLN can improve the PDR.

Diagnostic Value of Artificial Intelligence Based on CT Image in Benign and Malignant Pulmonary Nodules.

Objective To evaluate the diagnostic value of artificial intelligence-assisted CT imaging in benign and malignant pulmonary nodules. Methods The CT scan screening of pulmonary nodules from November 2018 to November 2020 was retrospectively collected. The diagnosis of pulmonary nodules and surgical treatment were performed. A total of 194 nodules in 152 patients with clear pathological results were observed. All patients underwent CT examination to analyze the consistency of the results of artificial intelligence, physician reading according to imaging features, multidisciplinary team work (MDT) diagnosis, and postoperative pathological results; the diagnostic efficacy of different diagnostic methods for solitary pulmonary nodules and the differences of ROC curve and AUC were analyzed. The accuracy, specificity, sensitivity, positive predictive value, negative predictive value, false negative rate, and false positive rate of different diagnostic methods for pulmonary nodules were calculated, and the ROC curves of different diagnostic methods were plotted. Results The accuracy, sensitivity, specificity, and Youden index of artificial intelligence (AI) were 89.69%, 92.98%, 65.22%, and 58.20%; the accuracy, sensitivity, specificity, and Youden index of physician reading were 85.57%, 88.30%, 65.22%, and 53.52%; the accuracy, sensitivity, specificity, and Youden index of MDT were 96.91%, 98.25%, 86.96%, and 85.21%, respectively. The kappa values of artificial intelligence, physician reading, and MDT were 0.541, 0.437, and 0.852, and the AUC was 0.768, 0.791, and 0.926, respectively (P < 0.001). The average detection time of pulmonary nodules in the AI group, manual reading group, and MAT group was (145 ± 97) s, (534 ± 297) s, and (421 ± 128) s (P < 0.001). Conclusion Artificial intelligence pulmonary nodule detection system can improve the coincidence rate and accuracy of early diagnosis of lung cancer, shorten the average detection time, and provide more accurate information for clinical decision-making.

External validation study on the value of deep learning algorithm for the prediction of hematoma expansion from noncontrast CT scans

BackgroundHematoma expansion is an independent predictor of patient outcome and mortality. The early diagnosis of hematoma expansion is crucial for selecting clinical treatment options. This study aims to explore the value of a deep learning algorithm for the prediction of hematoma expansion from non-contrast computed tomography (NCCT) scan through external validation.Methods102 NCCT images of hypertensive intracerebral hemorrhage (HICH) patients diagnosed in our hospital were retrospectively reviewed. The initial computed tomography (CT) scan images were evaluated by a commercial Artificial Intelligence (AI) software using deep learning algorithm and radiologists respectively to predict hematoma expansion and the corresponding sensitivity, specificity and accuracy of the two groups were calculated and compared. Comparisons were also conducted among gold standard hematoma expansion diagnosis time, AI software diagnosis time and doctors’ reading time.ResultsAmong 102 HICH patients, the sensitivity, specificity, and accuracy of hematoma expansion prediction in the AI group were higher than those in the doctor group(80.0% vs 66.7%, 73.6% vs 58.3%, 75.5% vs 60.8%), with statistically significant difference (p < 0.05). The AI diagnosis time (2.8 ± 0.3 s) and the doctors’ diagnosis time (11.7 ± 0.3 s) were both significantly shorter than the gold standard diagnosis time (14.5 ± 8.8 h) (p < 0.05), AI diagnosis time was significantly shorter than that of doctors (p < 0.05).ConclusionsDeep learning algorithm could effectively predict hematoma expansion at an early stage from the initial CT scan images of HICH patients after onset with high sensitivity and specificity and greatly shortened diagnosis time, which provides a new, accurate, easy-to-use and fast method for the early prediction of hematoma expansion.

Augmenting the accuracy of trainee doctors in diagnosing skin lesions suspected of skin neoplasms in a real-world setting: A prospective controlled before-and-after study.

BackgroundAlthough deep neural networks have shown promising results in the diagnosis of skin cancer, a prospective evaluation in a real-world setting could confirm these results. This study aimed to evaluate whether an algorithm (http://b2019.modelderm.com) improves the accuracy of nondermatologists in diagnosing skin neoplasms.MethodsA total of 285 cases (random series) with skin neoplasms suspected of malignancy by either physicians or patients were recruited in two tertiary care centers located in South Korea. An artificial intelligence (AI) group (144 cases, mean [SD] age, 57.0 [17.7] years; 62 [43.1%] men) was diagnosed via routine examination with photographic review and assistance by the algorithm, whereas the control group (141 cases, mean [SD] age, 61.0 [15.3] years; 52 [36.9%] men) was diagnosed only via routine examination with a photographic review. The accuracy of the nondermatologists before and after the interventions was compared.ResultsAmong the AI group, the accuracy of the first impression (Top-1 accuracy; 58.3%) after the assistance of AI was higher than that before the assistance (46.5%, P = .008). The number of differential diagnoses of the participants increased from 1.9 ± 0.5 to 2.2 ± 0.6 after the assistance (P < .001). In the control group, the difference in the Top-1 accuracy between before and after reviewing photographs was not significant (before, 46.1%; after, 51.8%; P = .19), and the number of differential diagnoses did not significantly increase (before, 2.0 ± 0.4; after, 2.1 ± 0.5; P = .57).ConclusionsIn real-world settings, AI augmented the diagnostic accuracy of trainee doctors. The limitation of this study is that the algorithm was tested only for Asians recruited from a single region. Additional international randomized controlled trials involving various ethnicities are required.

Artificial intelligence application versus physical therapist for squat evaluation: a randomized controlled trial

Artificial intelligence technology is becoming more prevalent in health care as a tool to improve practice patterns and patient outcomes. This study assessed ability of a commercialized artificial intelligence (AI) mobile application to identify and improve bodyweight squat form in adult participants when compared to a physical therapist (PT). Participants randomized to AI group (n = 15) performed 3 squat sets: 10 unassisted control squats, 10 squats with performance feedback from AI, and 10 additional unassisted test squats. Participants randomized to PT group (n = 15) also performed 3 identical sets, but instead received performance feedback from PT. AI group intervention did not differ from PT group (log ratio of two odds ratios = − 0.462, 95% confidence interval (CI) (− 1.394, 0.471), p = 0.332). AI ability to identify a correct squat generated sensitivity 0.840 (95% CI (0.753, 0.901)), specificity 0.276 (95% CI (0.191, 0.382)), PPV 0.549 (95% CI (0.423, 0.669)), NPV 0.623 (95% CI (0.436, 0.780)), and accuracy 0.565 95% CI (0.477, 0.649)). There was no statistically significant association between group allocation and improved squat performance. Current AI had satisfactory ability to identify correct squat form and limited ability to identify incorrect squat form, which reduced diagnostic capabilities.Trial Registration NCT04624594, 12/11/2020, retrospectively registered.