Chimeric antigen receptor (CAR) T-cells have been recently developed and are producing impressive outcomes in patients with hematologic malignancies. However, there is no standardized method for cell trafficking and in vivo CAR T-cell monitoring. We assessed the feasibility of real-time in vivo ⁸⁹Zr-p-Isothiocyanatobenzyl-desferrioxamine (Df-Bz-NCS, DFO) labeled CAR T-cell trafficking using positron emission tomography (PET). The ⁸⁹Zr-DFO radiolabeling efficiency of Jurkat/CAR and human peripheral blood mononuclear cells (hPBMC)/CAR T-cells was 70%–79%, and cell radiolabeling activity was 98.1–103.6 kBq/10⁶ cells. Cell viability after radiolabeling was >95%. Cell proliferation was not significantly different during the early period after radiolabeling, compared with unlabeled cells; however, the proliferative capacity decreased over time (day 7 after labeling). IL-2 or IFN-γ secretion was not significantly different between unlabeled and labeled CAR T-cells. PET/magnetic resonance imaging in the xenograft model showed that most of the ⁸⁹Zr-DFO-labeled Jurkat/CAR T-cells were distributed in the lung (24.4% ± 3.4%ID) and liver (22.9% ± 5.6%ID) by one hour after injection. The cells gradually migrated from the lung to the liver and spleen by day 1, and remained stable in these sites until day 7 (on day 7: lung 3.9% ± 0.3%ID, liver 36.4% ± 2.7%ID, spleen 1.4% ± 0.3%ID). No significant accumulation of labeled cells was identified in tumors. A similar pattern was observed in ex vivo biodistributions on day 7 (lung 3.0% ± 1.0%ID, liver 19.8% ± 2.2%ID, spleen 2.3% ± 1.7%ID). ⁸⁹Zr-DFO-labeled hPBMC/CAR T-cells showed a similar distribution, compared with Jurkat/CAR T-cells, on serial PET images. CAR T cell distribution was cross-confirmed by flow cytometry, Alu polymerase chain reaction, and immunohistochemistry. Real-time in vivo cell trafficking is feasible using PET imaging of ⁸⁹Zr-DFO-labeled CAR T-cells. This can be used to investigate cellular kinetics, initial in vivo biodistribution, and safety profiles in future CAR T-cell development.

Bone scans are often used to identify bone metastases, but their low specificity may necessitate further studies. Deep learning models may improve diagnostic accuracy but require both medical and programming expertise. Therefore, we investigated the feasibility of constructing a deep learning model employing ChatGPT for the diagnosis of bone metastasis in bone scans and to evaluate its diagnostic performance. We examined 4626 consecutive cancer patients (age, 65.1 ± 11.3 years; 2334 female) who had bone scans for metastasis assessment. A nuclear medicine physician developed a deep learning model using ChatGPT 3.5 (OpenAI). We employed ResNet50 as the backbone network and compared the diagnostic performance of four strategies (original training set, original training set with 1:10 class weight, 10-fold data augmentation for positive images only, and 10-fold data augmentation for all images) to address the class imbalance. We used a class activation map algorithm for visualization. Among the four strategies, the deep learning model with 10-fold data augmentation for positive cases only, using a batch size of 16 and an epoch size of 150, achieved the area under curve of 0.8156, the sensitivity of 56.0 %, and specificity of 88.7 %. The class activation map indicated that the model focused on disseminated bone metastases within the spine but might confuse them with benign spinal lesions or intense urinary activity. Our study illustrates that a clinical physician with rudimentary programming skills can develop a deep learning model for medical image analysis, such as diagnosing bone metastasis in bone scans using ChatGPT. Model visualization may offer guidance in enhancing deep learning model development, including preprocessing, and potentially support clinical decision-making processes.

Yoon HY, et al. J Korean Med Sci. 2021 Sep;36(41):e257. https://doi.org/10.3346/jkms.2021.36.e257

Background: Acute exacerbation (AE) is the most lethal postoperative complication in idiopathic pulmonary fibrosis (IPF); however, prediction before surgery is difficult. We investigated the role of ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG PET/CT) in predicting postoperative AE in IPF. Method: Clinical data of 48 IPF patients who underwent 18F-FDG PET/CT before thoracic surgery were retrospectively analyzed. Mean and maximal standardized uptake values (SUV_mean and SUV_max, respectively) were measured in the fibrotic area. Additionally, adjusted values-SUV ratio (SUVR, defined as SUV_max-to-liver SUV_mean ratio), tissue fraction-corrected SUV_mean (SUV_meanTF), and SUVR (SUVR_TF)-were calculated. Results: The mean age of the subjects was 67.8 years and 91.7% were male. After thoracic surgery, 21 (43.8%) patients experienced postoperative complications including prolonged air leakage (29.2%), death (14.6%), and AE (12.5%) within 30 days. Patients who experienced AE showed higher SUV_max, SUVR, SUV_meanTF, and SUVR_TF than those who did not, but other clinical parameters were not different between patients with and without AE. The SUV parameters did not differ for other complications. The SUVR (odds ratio [OR] 29.262; P = 0.030), SUV_meanTF (OR 3.709; P = 0.041) and SUVR_TF (OR 20.592; P = 0.017) were significant predicting factors for postoperative AE following a multivariate logistic regression analysis. On receiver operating characteristics curve analysis, SUVR_TF had the largest area under the curve (0.806, P = 0.007) for predicting postoperative AE among SUV parameters. Conclusions: Our findings suggest that ¹⁸F-FDG PET/CT may be useful in predicting postoperative AE in IPF patients and among SUVs, SUVR_TF is the best parameter for predicting postoperative AE in IPF patients.

Primary central nervous system lymphoma (PCNSL) typically shows a strong uptake of ¹⁸F-fludeoxyglucose (FDG) imaged by positron emission tomography (PET). Uncommonly, PCNSL demonstrates a low uptake on FDG PET. We investigated the clinicopathological characteristics of the unusual cases of PCNSL with low FDG uptake. We retrospectively enrolled 104 consecutive patients with newly diagnosed PCNSL who underwent baseline brain FDG PET. The degree of FDG uptake of PCNSL was visually scored by 4 grades (0, ≤contralateral white matter; 1, >contralateral white matter and contralateral gray matter). Grades 0–2 were considered as PCNSL with low uptake. We investigated association of low uptake of PCNSL with the following clinicopathological factors: age, sex, steroid treatment, lactate dehydrogenase level, cerebrospinal fluid protein level, condition of PET scanning, immunohistochemical markers (cluster of differentiation 10 [CD10], B-cell lymphoma 6 [BCL-6], B-cell lymphoma 2 [BCL-2], multiple myeloma oncogene 1 [MUM1], Epstein-Barr virus [EBV] protein, and Ki67), location of lesions, tumor size, multiplicity of lesions, involvement of deep brain structures, and cystic or necrotic appearance of lesions. Of the 104 patients with PCNSL, 14 patients (13.5%) showed PCNSL with low FDG uptake on PET. Among various clinicopathological factors, MUM1 negativity was the only factor associated with low FDG uptake PCNSL by univariate (P = .002) and multivariate analysis (P = .007). This study suggests that the different clinicopathological characteristics between patients with high uptake and low uptake of PCNSL on FDG PET is closely associated with lack of MUM1, a protein known to be a crucial regulator of B-cell development and tumorigenesis.