Search

Introduction: In the 21st century's era of rapid technological advancement, ChatGPT-3.5, an artificial intelligence (AI) language model, is scrutinized for its application in dermatology. Using 119 questions from the National Specialist Examination (PES), we assess ChatGPT-3.5's performance by comparing it to human skills and addressing ethical implications. Our primary aim is to evaluate ChatGPT-3.5's proficiency in responding to 119 dermatology questions from the PES. The study emphasizes ethical considerations and compares the model's knowledge and skills to those of human dermatologists. Material and methods: Utilizing the 2023 PES question database, questions were categorized by Bloom's taxonomy and thematic content. ChatGPT-3.5, version of 3 August 2023, answered 119 questions in five sessions, allowing for a probabilistic evaluation. Statistical analyses, conducted using R Studio, assessed correctness, confidence, and difficulty. Results: ChatGPT-3.5 achieved a 49.58% correct response rate, below the 60% passing threshold. No significant differences in difficulty or correlations between difficulty and certainty were observed. Varied performance across question types highlighted strengths and weaknesses. Despite suboptimal results, ChatGPT-3.5's differential performance offers insights, suggesting future improvements. The study advocates for ongoing research into AI integration in dermatology, envisioning a promising role for AI in assisting dermatologists. Conclusions: Ethical considerations are crucial for effective AI introduction, minimizing errors, and enhancing dermatological healthcare quality, fostering optimism for AI's evolving role in dermatology. [ABSTRACT FROM AUTHOR]

Introduction: The rapid development of artificial intelligence (AI) has sparked a desire to analyse its potential applications in medicine. The aim of this article is to present the effectiveness of the ChatGPT advanced language model in the context of the pass rate of the polish National Specialty Examination (PES) in nuclear medicine. It also aims to identify its strengths and limitations through an in-depth analysis of the issues raised in the exam questions. Methods: The PES exam provided by the Centre for Medical Examinations in Łódź, consisting of 120 questions, was used for the study. The questions were asked using the openai.com platform, through which free access to the GPT-3.5 model is available. All questions were classified according to Bloom's taxonomy to determine their complexity and difficulty, and according to two authors' subcategories. To assess the model's confidence in the validity of the answers, each questions was asked five times in independent sessions. Results: ChatGPT achieved 56%, which means it did not pass the exam. The pass rate is 60%. Of the 117 questions asked, 66 were answered correctly. In the percentage of each type and subtype of questions answered correctly, there were no statistically significant differences. Conclusion: Further testing is needed using the questions provided by Centre for Medical Examinations from the nuclear medicine specialty exam to evaluate the utility of the ChatGPT model. This opens the door for further research on upcoming improved versions of the ChatGPT. [ABSTRACT FROM AUTHOR]

In recent years, deep neural networks have enabled countless innovations in the field of image classification. Encouraged by success in this field, researchers worldwide have demonstrated how to use Convolutional Neural Network techniques in medical imaging problems. In this article, the results were obtained through the use of the EfficientNet in the task of classifying 14 different diseases based on chest X-ray images coming from the NIH (National Institutes of Health) ChestX-ray14 dataset. The approach addresses dataset imbalances by introducing a custom split to ensure fair representation. Binary cross entropy loss is utilized to handle the multi-label difficulty. The model architecture comprises an EfficientNet backbone for feature extraction, succeeded by sequential layers including GlobalAveragePooling, Dense, and BatchNormalization. The main contribution of this paper is a proposed solution that outperforms previous state-of-the-art deep learning models average area under the receiver operating characteristic curve—AUC-ROC (score: 84.28%). The usage of the transfer-learning technique and traditional deep learning engineering techniques was shown to enable us to obtain such results on consumer-class GPUs (graphics processing units). [ABSTRACT FROM AUTHOR]