Abstract 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.
Abstract Immune checkpoint inhibitors (ICIs) have been extensively used for the treatment of non-small cell lung cancer patients in recent years, providing a significant survival benefit. However, a major drawback of ICIs-related immunotherapy is the risk of developing post-surgical pneumonitis. In this study, we propose a deep learning-embedded, multi-modality prediction approach to assess whether patients will develop ICI-pneumonitis after receiving ICIs-based immunotherapy. This approach utilizes multi-modal data, including clinical data and pre-treatment lung screening computed tomography (CT) images. We extracted three types of features: 1) deep learning features from CT scans using a pretrained vision transformer, 2) radiomic features from CT scans using predefined radiomic algorithms, and 3) clinical features from patients’ clinical records. We then compared multiple machine learning algorithms for prediction based on these extracted features. Our results demonstrated a prediction accuracy of 0.823 and an area under the receiver operating characteristic curve of 0.895.
Importance Radiotherapy (RT) plan quality is an established predictive factor associated with cancer recurrence and survival outcomes. The addition of radiologists to the peer review (PR) process may increase RT plan quality. Objective To determine the rate of changes to the RT plan with and without radiology involvement in PR of radiation targets. Data Sources PubMed, Scopus, and Web of Science were queried for peer-reviewed articles published from inception up to March 6, 2024. Search terms included key words associated with PR of contoured targets for the purposes of RT planning with or without radiology involvement. Study Selection Studies reporting PR of contoured radiation targets with or without radiology involvement. Studies were excluded if they lacked full text, reported clinical trial–specific quality assurance, or reported PR without dedicated review of RT targets. Data Extraction and Synthesis Data were extracted according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. Titles and abstracts were screened by 2 reviewers. In the case of discordance, discussion was used to reach consensus regarding inclusion for full-text review. RT plan changes were characterized as major when the change was expected to have a substantial clinical impact, as defined by the original study. Pooled outcomes were estimated using random-effects models. Main Outcomes and Measures Primary outcome was pooled rate of RT plan changes. Secondary outcomes included pooled rates of major and minor changes to RT targets or organs at risk. Results Of 4185 screened studies, 31 reporting 39 509 RT plans were included (390 with radiology and 39 119 without). The pooled rate of plan changes was 29.0% (95% CI, 20.7%-37.2%). Radiologist participation in PR was associated with significant increases in plan change rates (49.4% [95% CI, 28.6%-70.1%] vs 25.0% [95% CI, 17.0%-33.1%]; P = .02) and in clinically relevant major changes (47.0% [95% CI, 34.1%-59.8%] vs 10.2% [95% CI, 4.6%-15.8%]; P < .001). There was no difference in minor changes (15.2% [95% CI, 9.7%-20.6%] vs 13.8% [95% CI, 9.3%-18.3%]; P = .74). Subgroup analyses identified increases in the rates of changes to the gross tumor and planning target volumes with radiology-based PR. The highest rates of plan changes were observed in head and neck or lung cancer studies, studies performing PR prior to RT planning, and prospective studies. Conclusions and Relevance In this systematic review and meta-analysis of radiation oncology PR of contoured targets, radiologist involvement in peer review was associated with a significant increase in the rate of total and clinically meaningful changes to the RT targets with no change in minor change rates. These results support the value of interdisciplinary collaboration with radiology during RT planning.
Optimal diagnostic algorithm to differentiate checkpoint inhibitor pneumonitis (CIP) from mimics is uncertain; patients with respiratory comorbidities often receive prolonged corticosteroids until diagnostic clarification. Drawbacks to empiric use of corticosteroids include decreased immunotherapy (IO) efficacy and increased infectious risk. This retrospective study systematically collected data on patients treated for lung cancer who were suspected to have severe CIP.
Methods
This single-center retrospective cohort study collected data on all lung cancer patients who received > 1 dose of an immune checkpoint inhibitor between 6/1/18 and 2/1/20 (n=210), were subsequently hospitalized and received > 1 dose of systemic corticosteroids for any indication (n=97). Data were collected on clinical factors including comorbidities, cancer stage, IO cycles, biomarkers, diagnostic work-up, antibiotics, steroids, progression, and survival. A blinded radiologist reviewed all imaging of suspected CIP cases and categorized their radiographic patterns.
Results
In our high-risk cohort of 97 patients, median follow-up was 23 months with progression in 54 patients (56%) at median 11 months and death in 67 patients (69%) at median 14mo. Twelve patients (12%) were suspected to have severe CIP after IO treatment for lung cancer; CIP was confirmed in 5/12 and ruled-out (mimics) in 7/12 after 30 and 3 median IO cycles, respectively. Most suspected patients underwent CXR, CTA chest, blood cultures, and received empiric antibiotics. Common radiographic patterns were ground-glass opacities, organizing pneumonia, hypersensitivity pneumonitis, and acute interstitial pneumonia/acute respiratory distress syndrome (AIP/ARDS) among confirmed cases (4/5) and ground-glass opacities, organizing pneumonias, bronchiolitis, AIP/ARDS among mimics (4/7). The median time to confirm CIP or rule out a mimic was 5 ± 4 days. Median time to onset of symptoms differed substantially for confirmed and mimic cases: 17 months and 1 month, respectively.
Conclusions
CIP mimics were more common than confirmed cases in routine clinical practice, particularly among patients hospitalized for respiratory symptoms <1 month after initiating immunotherapy for lung cancers. In these cases, it is reasonable to empirically cover possible CIP with shorter (~1 week) courses of steroids until diagnostic clarity is achieved. CT imaging should be obtained as it is sensitive though not specific for CIP. CIP mimics may contribute to the higher incidence of CIP reported by real-world patient registries than by clinical trials.
Ethics Approval
The study was approved by Wake Forest Baptist Medical Center's Ethics Board, IRB approval number 00044126
In this report, we present the case of a 66-year-old man who received local consolidation radiotherapy to the right lung and mediastinum for oligometastatic non-small cell lung cancer (NSCLC) following partial response to upfront chemoimmunotherapy. He continued with maintenance immunotherapy and was asymptomatic for eight months after completing radiation therapy. He then developed symptoms consistent with pneumonitis within three to five days of his first administration of the coronavirus disease 2019 (COVID-19) vaccine injection. He reported that these symptoms significantly intensified within three to five days of receiving his second dose of the vaccine. The clinical time frame and radiographic evidence raised suspicion for radiation recall pneumonitis (RRP). Patients undergoing maintenance immunotherapy after prior irradiation may be at increased risk of this phenomenon that may be triggered by the administration of the COVID-19 vaccine.
