Hidden Flaws Behind Expert-Level Accuracy of Multimodal GPT-4 Vision in Medicine.

Recent studies indicate that Generative Pre-trained Transformer 4 with Vision (GPT-4V) outperforms human physicians in medical challenge tasks. However, these evaluations primarily focused on the accuracy of multi-choice questions alone. Our study extends the current scope by conducting a comprehensive analysis of GPT-4V's rationales of image comprehension, recall of medical knowledge, and step-by-step multimodal reasoning when solving New England Journal of Medicine (NEJM) Image Challenges - an imaging quiz designed to test the knowledge and diagnostic capabilities of medical professionals. Evaluation results confirmed that GPT-4V performs comparatively to human physicians regarding multi-choice accuracy (81.6% vs. 77.8%). GPT-4V also performs well in cases where physicians incorrectly answer, with over 78% accuracy. However, we discovered that GPT-4V frequently presents flawed rationales in cases where it makes the correct final choices (35.5%), most prominent in image comprehension (27.2%). Regardless of GPT-4V's high accuracy in multi-choice questions, our findings emphasize the necessity for further in-depth evaluations of its rationales before integrating such multimodal AI models into clinical workflows.

A computer-aided polyp detection system in screening and surveillance colonoscopy: an international, multicentre, randomised, tandem trial.

BACKGROUND: Studies on the effect of computer-aided detection (CAD) in a daily clinical screening and surveillance colonoscopy population practice are scarce. The aim of this study was to evaluate a novel CAD system in a screening and surveillance colonoscopy population. METHODS: This multicentre, randomised, controlled trial was done in ten hospitals in Europe, the USA, and Israel by 31 endoscopists. Patients referred for non-immunochemical faecal occult blood test (iFOBT) screening or surveillance colonoscopy were included. Patients were randomomly assigned to CAD-assisted colonoscopy or conventional colonoscopy; a subset was further randomly assigned to undergo tandem colonoscopy: CAD followed by conventional colonoscopy or conventional colonoscopy followed by CAD. Primary objectives included adenoma per colonoscopy (APC) and adenoma per extraction (APE). Secondary objectives included adenoma miss rate (AMR) in the tandem colonoscopies. The study was registered at ClinicalTrials.gov, NCT04640792. FINDINGS: A total of 916 patients were included in the modified intention-to-treat analysis: 449 in the CAD group and 467 in the conventional colonoscopy group. APC was higher with CAD compared with conventional colonoscopy (0·70 vs 0·51, p=0·015; 314 adenomas per 449 colonoscopies vs 238 adenomas per 467 colonoscopies; poisson effect ratio 1·372 [95% CI 1·068-1·769]), while showing non-inferiority of APE compared with conventional colonoscopy (0·59 vs 0·66; p<0·001 for non-inferiority; 314 of 536 extractions vs 238 of 360 extractions). AMR in the 127 (61 with CAD first, 66 with conventional colonoscopy first) patients completing tandem colonoscopy was 19% (11 of 59 detected during the second pass) in the CAD first group and 36% (16 of 45 detected during the second pass) in the conventional colonoscopy first group (p=0·024). INTERPRETATION: CAD increased adenoma detection in non-iFOBT screening and surveillance colonoscopies and reduced adenoma miss rates compared with conventional colonoscopy, without an increase in the resection of non-adenomatous lesions. FUNDING: Magentiq Eye.

Diagnosis and management of common gastrointestinal tract infectious diseases in ulcerative colitis and Crohn's disease patients.

Management of inflammatory bowel disease (IBD), including ulcerative colitis and Crohn's disease, stretches beyond control of flares. Some infections of the gastrointestinal tract are more commonly seen in patients with IBD. Work from the Human Microbiome Project has been instrumental in our understanding of the interplay between the vast gut microbiota and host immune responses. Patients with IBD may be more prone to infectious complications based on their underlying inflammatory disease and variations in their microbiome. Immunosuppressant medications commonly used to treat patients with Crohn's and colitis also play a role in predisposing these patients to acquire these infections. Here, we present a detailed review of the data focusing on the most common infections of the gastrointestinal tract in patients with IBD: Clostridium difficile infections (CDI) and cytomegalovirus (CMV). We will discuss appropriate diagnostic tools and treatment options for these infections. Other less common infections will also be reviewed briefly. Studying the various infections of the gastrointestinal tract in these patients could enhance our understanding of the pathophysiology of IBD.

Relationship between routine multi-detector cardiac computed tomographic angiography prior to reoperative cardiac surgery, length of stay, and hospital charges.

While multi-detector cardiac computed tomography angiography (MDCCTA) prior to reoperative cardiac surgery (RCS) has been associated with improved clinical outcomes, its impact on hospital charges and length of stay remains unclear. We studied 364 patients undergoing RCS at Washington Hospital Center between 2004 and 2008, including 137 clinically referred for MDCCTA. Baseline demographics, procedural data, and perioperative outcomes were recorded at the time of the procedure. The primary clinical endpoint was the composite of perioperative death, myocardial infarction (MI), stroke, and hemorrhage-related reoperation. Secondary clinical endpoints included surgical procedural variables and the perioperative volume of bleeding and transfusion. Length of stay was determined using the hospital's electronic medical record. Cost data were extracted from the hospital's billing summary. Analysis was performed on individual categories of care, as well as on total hospital charges. Data were compared between subjects with and without MDCCTA, after adjustment for the Society of Thoracic Surgeons score. Baseline characteristics were similar between the two groups. MDCCTA was associated with shorter procedural times, shorter intensive care unit stays, fewer blood transfusions, and less frequent perioperative MI. There was additionally a trend towards a lower incidence of the primary endpoint (17.5 vs. 24.2 %, p = 0.13) primarily due to a lower incidence of perioperative MI (0 vs. 5.7 %, p = 0.002). MDCCTA was also associated with lower median recovery room [$1,325 (1,250-3,302) vs. $3,217 (1,325-5,353) p < 0.001] and nursing charges [$6,335 (3,623-10,478) vs. $6,916 (3,915-14,499) p = 0.03], although operating room charges were higher [$24,100 (22,300-29,700) vs. $23,500 (19,900-27,700) p < 0.05]. Median total charges [$127,000 (95,000-188,000) vs. $123,000 (86,800-226,000) p = 0.77] and length of stay [9 days (6-19) vs. 11 days (7-19), p = 0.21] were similar. Means analysis demonstrated a strong trend towards lower mean total hospital charges [$163,000 (108,426) vs. $192,000 (181,706), p = 0.06] in the MDCCTA group. In conclusion, preoperative MDCCTA is associated with a number of improved perioperative outcomes and does not significantly effect the length of stay or total hospital charges during the index hospitalization.

Perioperative outcomes in reoperative cardiac surgery guided by cardiac multidetector computed tomographic angiography.

BACKGROUND: Preoperative evaluation with contrast-enhanced multidetector computed tomographic angiography (MDCTA) is considered an "appropriate" indication based on expert consensus. We aimed to evaluate how the presurgical evaluation with MDCTA impacts the outcomes after reoperative cardiac surgery (RCS). METHODS: We retrospectively studied 364 patients undergoing RCS between 2004 and 2008, including 137 referred for MDCTA. High-risk CT findings were defined as the presence of right ventricle or aorta <10 mm from the sternum or a bypass graft <10 mm from the sternum crossing the midline. The primary clinical end point was the composite of perioperative death, myocardial infarction (MI), stoke, and hemorrhage-related reoperation. Secondary end points included surgical procedural variables and the perioperative volume of bleeding and of red blood cell (RBC) transfusion. RESULTS: Baseline clinical characteristics were similar between the 2 groups. Individuals referred for MDCTA showed a trend toward a lower incidence of the composite primary end point (17.5% vs 24.2%, P = .13), primarily related to a significantly lower incidence of perioperative MI (0% vs 5.7%, P = .002). Multidetector computed tomographic angiography was also associated with shorter perfusion (90 vs 110 minutes, P = .002), cross clamp time (63 vs 75 minutes, P = .003), and total time in intensive care unit (103 vs 148 hours, P = .04), and a lower volume of postoperative RBC transfusion (627 vs 824 mL, P = .09). These differences remained significant after adjustment for the Society of Thoracic Surgeons score and the performing surgeon. CONCLUSION: The use of MDCTA before RCS was associated with shorter perfusion and cross clamp time, shorter intensive care unit stays, and less frequent perioperative MI.