Background The latest large language models (LLMs) solve unseen problems via user-defined text prompts without the need for retraining, offering potentially more efficient information extraction from free-text medical records than manual annotation. Purpose To compare the performance of the LLMs ChatGPT and GPT-4 in data mining and labeling oncologic phenotypes from free-text CT reports on lung cancer by using user-defined prompts. Materials and Methods This retrospective study included patients who underwent lung cancer follow-up CT between September 2021 and March 2023. A subset of 25 reports was reserved for prompt engineering to instruct the LLMs in extracting lesion diameters, labeling metastatic disease, and assessing oncologic progression. This output was fed into a rule-based natural language processing pipeline to match ground truth annotations from four radiologists and derive performance metrics. The oncologic reasoning of LLMs was rated on a five-point Likert scale for factual correctness and accuracy. The occurrence of confabulations was recorded. Statistical analyses included Wilcoxon signed rank and McNemar tests. Results On 424 CT reports from 424 patients (mean age, 65 years ± 11 [SD]; 265 male), GPT-4 outperformed ChatGPT in extracting lesion parameters (98.6% vs 84.0%, P < .001), resulting in 96% correctly mined reports (vs 67% for ChatGPT, P < .001). GPT-4 achieved higher accuracy in identification of metastatic disease (98.1% [95% CI: 97.7, 98.5] vs 90.3% [95% CI: 89.4, 91.0]) and higher performance in generating correct labels for oncologic progression (F1 score, 0.96 [95% CI: 0.94, 0.98] vs 0.91 [95% CI: 0.89, 0.94]) (both P < .001). In oncologic reasoning, GPT-4 had higher Likert scale scores for factual correctness (4.3 vs 3.9) and accuracy (4.4 vs 3.3), with a lower rate of confabulation (1.7% vs 13.7%) than ChatGPT (all P < .001). Conclusion When using user-defined prompts, GPT-4 outperformed ChatGPT in extracting oncologic phenotypes from free-text CT reports on lung cancer and demonstrated better oncologic reasoning with fewer confabulations. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Hafezi-Nejad and Trivedi in this issue.

“Just Accepted” papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review ...

Purpose: Apoptotic dysregulation, redox adaptive mechanisms, and resilience to hypoxia are major causes of glioblastoma (GBM) resistance to therapy. Commonly known as crucial factors in energy metabolism, OCTN2 (SLC22A5) and its substrate L-carnitine (LC) are increasingly recognized as actors in cytoprotection. This study provides a comprehensive expression and survival analysis of the OCTN2/LC system in GBM and clarifies the system's impact on GBM progression. Experimental Design: OCTN2 expression and LC content were measured in 121 resected human GBM specimens and 10 healthy brain samples and analyzed for prognostic significance. Depending on LC administration, the effects of hypoxic, metabolic, and cytotoxic stress on survival and migration of LN18 GBM cells were further studied in vitro. Finally, an orthotopic mouse model was employed to investigate inhibition of the OCTN2/LC system on in vivo GBM growth. Results: Compared with healthy brain, OCTN2 expression was increased in primary and even more so in recurrent GBM on mRNA and protein level. High OCTN2 expression was associated with a poor overall patient survival; the unadjusted HR for death was 2.7 (95% CI, 1.47–4.91; P < 0.001). LC administration to GBM cells increased their tolerance toward cytotoxicity, whereas siRNA-mediated OCTN2 silencing led to a loss of tumor cell viability. In line herewith, OCTN2/LC inhibition by meldonium resulted in reduced tumor growth in an orthotopic GBM mouse model. Conclusions: Our data indicate a potential role of the OCTN2/LC system in GBM progression and resistance to therapy, and suggests OCTN2 as a prognostic marker in patients with primary GBM.

Background: Resistance of the highly aggressive glioblastoma multiforme (GBM) to drug therapy is a major clinical problem resulting in a poor patient’s prognosis. Beside promoter methylation of the O ⁶ -methylguanine-DNA-methyltransferase (MGMT) gene the efflux transporters ABCB1 and ABCG2 have been suggested as pivotal factors contributing to drug resistance, but the methylation of ABCB1 and ABCG2 has not been assessed before in GBM. Methods: Therefore, we evaluated the proportion and prognostic significance of promoter methylation of MGMT, ABCB1 and ABCG2 in 64 GBM patient samples using pyrosequencing technology. Further, the single nucleotide polymorphisms MGMT C-56 T (rs16906252), ABCB1 C3435T (rs1045642) and ABCG2 C421A (rs2231142) were determined using the restriction fragment length polymorphism method (RFLP). To study a correlation between promoter methylation and gene expression, we analyzed MGMT, ABCB1 and ABCG2 expression in 20 glioblastoma and 7 non-neoplastic brain samples. Results: Despite a significantly increased MGMT and ABCB1 promoter methylation in GBM tissue, multivariate regression analysis revealed no significant association between overall survival of glioblastoma patients and MGMT or ABCB1 promoter methylation. However, a significant negative correlation between promoter methylation and expression could be identified for MGMT but not for ABCB1 and ABCG2. Furthermore, MGMT promoter methylation was significantly associated with the genotypes of the MGMT C-56 T polymorphism showing a higher methylation level in the T allele bearing GBM. Conclusions: In summary, the data of this study confirm the previous published relation of MGMT promoter methylation and gene expression, but argue for no pivotal role of MGMT, ABCB1 and ABCG2 promoter methylation in GBM patients’ survival.

Patients with glioblastoma multiforme (GBM) are at high risk to develop a relapse despite multimodal therapy. Assumedly, glioma stem cells (GSCs) are responsible for treatment resistance of GBM. Identification of specific GSC markers may help to develop targeted therapies. Here, we performed expression analyses of stem cell (ABCG2, CD44, CD95, CD133, ELF4, Nanog, and Nestin) as well as differentiation and microglia markers (GFAP, Iba1, and Sparc) in GBM compared to nonmalignant brain. Furthermore, the role of these proteins for patient survival and their expression in LN18 stem-like neurospheres was analyzed. At mRNA level, ABCG2 and CD95 were reduced, GFAP was unchanged; all other investigated markers were increased in GBM. At protein level, CD44, ELF4, Nanog, Nestin, and Sparc were elevated in GBM, but only CD133 and Nestin were strongly associated with survival time. In addition, ABCG2 and GFAP expression was decreased in LN18 neurospheres whereas CD44, CD95, CD133, ELF4, Nanog, Nestin, and Sparc were upregulated. Altogether only CD133 and Nestin were associated with survival rates. This raises concerns regarding the suitability of the other target structures as prognostic markers, but makes both CD133 and Nestin candidates for GBM therapy. Nevertheless, a search for more specific marker proteins is urgently needed.

The current therapy for glioblastoma multiforme (GBM), the most aggressive and common primary brain tumor of adults, involves surgery and a combined radiochemotherapy that controls tumor progression only for a limited time window. Therefore, the identification of new molecular targets is highly necessary. Inhibition of kinases has become a standard of clinical oncology, and thus the oncogenic kinase Pim1 might represent a promising target for improvement of GBM therapy. Expression of Pim1 and associated signaling molecules was analyzed in human GBM samples, and the potential role of this kinase in patients' prognosis was evaluated. Furthermore, we analyzed the in vivo role of Pim1 in GBM cell growth in an orthotopic mouse model and examined the consequences of Pim1 inhibition in vitro to clarify underlying pathways. In comparison with normal brain, a strong upregulation of Pim1 was demonstrated in human GBM samples. Notably, patients with short overall survival showed a significantly higher Pim1 expression compared with GBM patients who lived longer than the median. In vitro experiments with GBM cells and analysis of patients' GBM samples suggest that Pim1 regulation is dependent on epidermal growth factor receptor. Furthermore, inhibition of Pim1 resulted in reduced cell viability accompanied by decreased cell numbers and increased apoptotic cells, as seen by elevated subG1 cell contents and caspase-3 and -9 activation, as well as modulation of several cell cycle or apoptosis regulatory proteins. Altogether, Pim1 could be a novel therapeutic target, which should be further analyzed to improve the outcome of patients with aggressive GBM.