Translating radiology reports into plain language using ChatGPT and GPT-4 with prompt learning:results,limitations,and potential

The large language model called ChatGPT has drawn extensively attention because of its human-like expression and reasoning abilities.In this study,we investigate the feasibility of using ChatGPT in experiments on translating radiology reports into plain language for patients and healthcare providers so that they are educated for improved healthcare.Radiology reports from 62 low-dose chest computed tomography lung cancer screening scans and 76 brain magnetic resonance imaging metastases screening scans were collected in the first half of February for this study.According to the evaluation by radiologists,ChatGPT can successfully translate radiology reports into plain language with an average score of 4.27 in the five-point system with 0.08 places of information missing and 0.07 places of misinformation.In terms of the suggestions provided by ChatGPT,they are generally relevant such as keeping following-up with doctors and closely monitoring any symptoms,and for about 37%of 138 cases in total ChatGPT offers specific suggestions based on findings in the report.ChatGPT also presents some randomness in its responses with occasionally over-simplified or neglected information,which can be mitigated using a more detailed prompt.Furthermore,ChatGPT results are compared with a newly released large model GPT-4,showing that GPT-4 can significantly improve the quality of translated reports.Our results show that it is feasible to utilize large language models in clinical education,and further efforts are needed to address limitations and maximize their potential.

Epilepsy Radiology Reports Classification Using Deep Learning Networks

The automatic and accurate classification of Magnetic Resonance Imaging(MRI)radiology report is essential for the analysis and interpretation epilepsy and non-epilepsy.Since the majority of MRI radiology reports are unstructured,the manual information extraction is time-consuming and requires specific expertise.In this paper,a comprehensive method is proposed to classify epilepsy and non-epilepsy real brain MRI radiology text reports automatically.This method combines the Natural Language Processing technique and statisticalMachine Learning methods.122 realMRI radiology text reports(97 epilepsy,25 non-epilepsy)are studied by our proposed method which consists of the following steps:(i)for a given text report our systems first cleans HTML/XML tags,tokenize,erase punctuation,normalize text,(ii)then it converts into MRI text reports numeric sequences by using indexbased word encoding,(iii)then we applied the deep learning models that are uni-directional long short-term memory(LSTM)network,bidirectional long short-term memory(BiLSTM)network and convolutional neural network(CNN)for the classifying comparison of the data,(iv)finally,we used 70%of used for training,15%for validation,and 15%for test observations.Unlike previous methods,this study encompasses the following objectives:(a)to extract significant text features from radiologic reports of epilepsy disease;(b)to ensure successful classifying accuracy performance to enhance epilepsy data attributes.Therefore,our study is a comprehensive comparative study with the epilepsy dataset obtained from numeric sequences by using index-based word encoding method applied for the deep learning models.The traditionalmethod is numeric sequences by using index-based word encoding which has been made for the first time in the literature,is successful feature descriptor in the epilepsy data set.The BiLSTM network has shown a promising performance regarding the accuracy rates.We show that the larger sizedmedical text reports can be analyzed by our proposed method.

Effect of training on resident inter-reader agreement with American College of Radiology Thyroid Imaging Reporting and Data System

BACKGROUND The American College of Radiology Thyroid Imaging Reporting and Data System(ACR TI-RADS)was introduced to standardize the ultrasound characterization of thyroid nodules.Studies have shown that ACR-TIRADS reduces unnecessary biopsies and improves consistency of imaging recommendations.Despite its widespread adoption,there are few studies to date assessing the inter-reader agreement amongst radiology trainees with limited ultrasound experience.We hypothesize that in PGY-4 radiology residents with no prior exposure to ACR TIRADS,a statistically significant improvement in inter-reader reliability can be achieved with a one hour training session.AIM To evaluate the inter-reader agreement of radiology residents in using ACR TIRADS before and after training.METHODS A single center retrospective cohort study evaluating 50 thyroid nodules in 40 patients of varying TI-RADS levels was performed.Reference standard TI-RADS scores were established through a consensus panel of three fellowship-trained staff radiologists with between 1 and 14 years of clinical experience each.Three PGY-4 radiology residents(trainees)were selected as blinded readers for this study.Each trainee had between 4 to 5 mo of designated ultrasound training.No trainee had received specialized TI-RADS training prior to this study.Each of the readers independently reviewed the 50 testing cases and assigned a TI-RADS score to each case before and after TI-RADS training performed 6 wk apart.Fleiss kappa was used to measure the pooled inter-reader agreement.The relative diagnostic performance of readers,pre-and post-training,when compared against the reference standard.RESULTS There were 33 females and 7 males with a mean age of 56.6±13.6 years.The mean nodule size was 19±14 mm(range from 5 to 63 mm).A statistically significant superior inter-reader agreement was found on the post-training assessment compared to the pre-training assessment for the following variables:1.“Shape”(k of 0.09[slight]pre-training vs 0.67[substantial]post-training,P<0.001),2.“Echogenic foci”(k of 0.28[fair]pre-training vs 0.45[moderate]post-training,P=0.004),3.‘TI-RADS level’(k of 0.14[slight]pre-training vs 0.36[fair]post-training,P<0.001)and 4.‘Recommendations’(k of 0.36[fair]pre-training vs 0.50[moderate]post-training,P=0.02).No significant differences between the preand post-training assessments were found for the variables'composition','echogenicity'and'margins'.There was a general trend towards improved pooled sensitivity with TI-RADS levels 1 to 4 for the post-training assessment while the pooled specificity was relatively high(76.6%-96.8%)for all TI-RADS level.CONCLUSION Statistically significant improvement in inter-reader agreement in the assigning TI-RADS level and recommendations after training is observed.Our study supports the use of dedicated ACR TI-RADS training in radiology residents.

Practice patterns and outcomes of equivocal bone scans for patients with castration-resistant prostate cancer: Results from SEARCH

Objective:To review follow-up imaging after equivocal bone scans in men with castration resistant prostate cancer(CRPC)and examine the characteristics of equivocal bone scans that are associated with positive follow-up imaging.Methods:We identified 639 men from five Veterans Affairs Hospitals with a technetium-99m bone scan after CRPC diagnosis,of whom 99(15%)had equivocal scans.Men with equivocal scans were segregated into“high-risk”and“low-risk”subcategories based upon wording in the bone scan report.All follow-up imaging(bone scans,computed tomography[CT],magnetic resonance imaging[MRI],and X-rays)in the 3 months after the equivocal scan were reviewed.Variables were compared between patients with a positive vs.negative follow-up imaging after an equivocal bone scan.Results:Of 99 men with an equivocal bone scan,43(43%)received at least one follow-up imaging test,including 32/82(39%)with low-risk scans and 11/17(65%)with high-risk scans(p=0.052).Of follow-up tests,67%were negative,14%were equivocal,and 19%were positive.Among those who underwent follow-up imaging,3/32(9%)low-risk men had metastases vs.5/11(45%)high-risk men(p=0.015).Conclusion:While 19%of all men who received follow-up imaging had positive follow-up imaging,only 9%of those with a low-risk equivocal bone scan had metastases versus 45%of those with high-risk.These preliminary findings,if confirmed in larger studies,suggest follow-up imaging tests for low-risk equivocal scans can be delayed while high-risk equivocal scans should receive follow-up imaging.

Diagnostic value of ACR TI-RADS combined with three-dimensional shear wave elastography in ACR TI-RADS 4 and 5 thyroid nodules

Background:Three-dimensional shear wave elastography(3D-SWE)is a promising method in distinguishing benign and malignant thyroid nodules.By combining with conventional method,it may further improve the diagnostic value.The study aimed to assess the diagnostic value of American College of Radiology(ACR)thyroid imaging reporting and data system(TI-RADS)combined with 3D-SWE in ACR TI-RADS 4 and 5 thyroid nodules.Methods:All nodules were examined by conventional ultrasonography,ACR TI-RADS classification,and 3D-SWE examination.Conventional ultrasonography was used to observe the location,size,shape,margin,echogenicity,taller-than-wide sign,microcalcification,and blood flow of thyroid nodules,and then ACR TI-RADS classification was performed.The Young’s modulus values(3D-C-Emax,3D-C-Emean,and elastography standard deviation[3D-C-Esd])were measured on the reconstructed coronal plane images.According to the receiver operating characteristic(ROC)curve,the best diagnostic efficiency among 3D-C-Emax,3D-C-Emean,and 3D-C-Esd was selected and the cut-off threshold was calculated.According to the surgical pathology,they were divided into benign group and malignant group.And appropriate statistical methods such as t-test and Mann-Whitney U test were used to compare the difference between the two groups.On this basis,3D-SWE combined with conventional ACR TI-RADS was reclassified as combined ACR TI-RADS to determine benign or malignant thyroid nodules.Results:Of the 112 thyroid nodules,62 were malignant and 50 were benign.The optimal cut-off value of three-dimensional maximum Young’s modulus in coronal plane(3D-C-Emax)was 51.5 kPa and the area under the curve(AUC)was 0.798.The AUC,sensitivity,specificity,and accuracy of conventional ACR TI-RADS were 0.828,83.9%,66.0%,and 75.9%,respectively.The AUC,sensitivity,specificity,and accuracy of combined ACR TI-RADS were 0.845,90.3%,66.0%,and 79.5%,respectively.The difference between the two AUC values was statistically significant.Conclusions:Combined ACR TI-RADS has higher diagnostic efficiency than conventional ACR TI-RADS.The sensitivity and accuracy of combined ACR TI-RADS showed significant improvements.It can be used as an effective method in the diagnosis of thyroid nodules.