ABSTRACT
Introduction: Generative AI is revolutionizing patient education in healthcare, particularly through chatbots that offer personalized, clear medical information. Reliability and accuracy are vital in AI-driven patient education. Research question: How effective are Large Language Models (LLM), such as ChatGPT and Google Bard, in delivering accurate and understandable patient education on lumbar disc herniation? Material and methods: Ten Frequently Asked Questions about lumbar disc herniation were selected from 133 questions and were submitted to three LLMs. Six experienced spine surgeons rated the responses on a scale from "excellent" to "unsatisfactory," and evaluated the answers for exhaustiveness, clarity, empathy, and length. Statistical analysis involved Fleiss Kappa, Chi-square, and Friedman tests. Results: Out of the responses, 27.2% were excellent, 43.9% satisfactory with minimal clarification, 18.3% satisfactory with moderate clarification, and 10.6% unsatisfactory. There were no significant differences in overall ratings among the LLMs (p = 0.90); however, inter-rater reliability was not achieved, and large differences among raters were detected in the distribution of answer frequencies. Overall, ratings varied among the 10 answers (p = 0.043). The average ratings for exhaustiveness, clarity, empathy, and length were above 3.5/5. Discussion and conclusion: LLMs show potential in patient education for lumbar spine surgery, with generally positive feedback from evaluators. The new EU AI Act, enforcing strict regulation on AI systems, highlights the need for rigorous oversight in medical contexts. In the current study, the variability in evaluations and occasional inaccuracies underline the need for continuous improvement. Future research should involve more advanced models to enhance patient-physician communication.
ABSTRACT
BACKGROUND/OBJECTIVE: In order to monitor tissue oxygenation in patients with acute neurological disorders, probes for measurement of brain tissue oxygen tension (ptO2) are often placed non-specifically in a right frontal lobe location. To improve the value of ptO2 monitoring, placement of the probe into a specific area of interest is desirable. We present a technique using CT-guidance to place the ptO2 probe in a particular area of interest based on the individual patient's pathology. METHODS: In this retrospective cohort study, we analyzed imaging and clinical data from all patients who underwent CT-guided ptO2 probe placement at our institution between October 2017 and April 2019. Primary endpoint was successful placement of the probe in a particular area of interest rated by two independent reviewers. Secondary outcomes were complications from probe insertion, clinical consequences from ptO2 measurements, clinical outcome according to the modified Rankin Scale (mRS) as well as development of ischemia on follow-up imaging. A historical control group was selected from patients who underwent conventional ptO2 probe placement between January 2010 and October 2017. RESULTS: Eleven patients had 16 CT-guided probes inserted. In 15 (93.75%) probes, both raters agreed on the correct placement in the area of interest. Each probe triggered on average 0.48 diagnostic or therapeutic adjustments per day. Only one infarction within the vascular territory of a probe was found on follow-up imaging. Eight out of eleven patients (72.73%) reached a good outcome (mRS ≤ 3). In comparison, conventionally placed probes triggered less diagnostic and therapeutic adjustment per day (p = 0.007). Outcome was worse in the control group (p = 0.024). CONCLUSION: CT-guided probe insertion is a reliable and easy technique to place a ptO2 probe in a particular area of interest in patients with potentially reduced cerebral oxygen supply. By adjusting treatment aggressively according to this individualized monitoring data, clinical outcome may improve.