Chest x-rays are among the most common diagnostic studies used in most both inpatient and outpatient settings, and they represent a significant portion of the workload for radiologists. Many different machine learning models have been developed for the analysis of chest x-rays, including models capable of detecting and labeling the location and type of pathological findings. In addition, large language models (LLMs) such as ChatGPT have also been growing in popularity and have proven to be effective at a variety of writing tasks [2]. For this project, we will attempt to use LLMs to translate machine learning results into automatically generated radiology reports. This would provide quick pre-reads of chest x-rays which can later be corrected or validated by radiologists in a similar workflow used by cardiologists when reading electrocardiograms (ECGs). To perform this task, we will make use of the Open-I dataset of chest x-rays with associated radiology reports [1]. Additionally, we will use a top performing model from the competition on the CheXpert dataset [3, 4]. This dataset consists of multiple chest x-rays with expert-annotated bounding boxes labeling pathological findings [3]. We will use the top-performing model to label the type and location of pathological findings in the Open-I dataset [4]. Following this, we will algorithmically transform the bounding boxes into simple descriptions of the type and location of the pathological finding (i.e., consolidation lower left quadrant, atelectasis upper right quadrant, cardiomegaly). We will then train a LLM to translate these simple descriptions into a full radiology report. To evaluate the efficacy of our method, we will present a mixture of expert written and automatically generated radiology reports to volunteers to assess if the generated reports. Volunteers will be selected from a variety of expertise levels and backgrounds in medicine, including non-medical laymen, medical students, and physicians. Volunteers will be asked to evaluate whether they can distinguish between automatically generated and expert written reports and if both reports adequately convey the relevant information from the associated chest x-ray. If the LLMs can use simple descriptors of machine learning results to produce radiology reports, this would significantly improve patient care and the workload for physicians. Patients and non-radiologist physicians would benefit from immediately available results following the acquisition of a chest x-ray. Radiologists will be able to overread the chest x-rays later, either verifying the AI-generated results or providing corrections, similar to the practice of Cardiologists with ECGs.