Integrating spatial transcriptomic data with deep features of tissue images

Hongyoon Choi 대한종양외과학회 2022 대한임상종양학회 학술대회지 Vol.2022 No.2

18F-Fluorodeoxyglucose and 11C-methionine positron emission tomography in relation to methyl-guanine methyltransferase promoter methylation in high-grade gliomas

Choi, Hongyoon,Bang, Ji-In,Cheon, Gi Jeong,Kim, Yong Hwy,Park, Chul-Kee,Park, Sung-Hye,Kang, Keon Wook,Chung, June-Key,Kim, Euishin E.,Lee, Dong Soo Wolters Kluwer Health | Lippincott Williams Wilkin 2015 Nuclear medicine communications Vol.36 No.3

INTRODUCTION: Methylation status of the methyl-guanine methyltransferase (MGMT) promoter is associated with a favorable response to a DNA alkylating agent in high-grade gliomas. We analyzed PET scans of patients with high-grade gliomas to determine whether the MGMT methylation status affects the tumor metabolic characteristics. PATIENTS AND METHODS: Twenty-three patients with high-grade glioma, who were initially examined with C-methionine (MET) and F-fluorodeoxyglucose (FDG) PET, were retrospectively enrolled. MET and FDG PET images were coregistered to each other and quantitative uptake of MET or FDG was assessed using tumor-to-normal uptake ratio of the cortex (TNR). TNRs for MET and FDG PET were compared between the two groups classified by MGMT promoter methylation status. RESULTS: Maximum TNRFDG of the MGMT methylated group was significantly higher than that of the MGMT unmethylated group (1.80±0.90 vs. 1.29±0.19; P=0.02). The MGMT methylated group also showed a trend for increased mean TNRFDG compared with the unmethylated group (0.85±0.21 vs. 0.72±0.11; P=0.10). There was no significant difference in TNRMET between the groups. In subgroup analyses with WHO grade 3 and 4, a trend for higher maximum TNRFDG was found in the MGMT methylated group compared with the unmethylated group. CONCLUSION: The MGMT methylated group showed higher glucose metabolism compared with the unmethylated group, whereas MET uptake did not show a significant difference. This suggests that MGMT methylation in high-grade gliomas could affect the tumor glucose metabolism. Thus, MGMT methylation status can cause a discrepancy in the prognostic prediction of high-grade gliomas by FDG PET, especially in patients scheduled for DNA alkylating chemotherapeutics.

Imaging of Myocardial Ischemia–Reperfusion Injury Using Sodium [ ¹⁸ F]Fluoride Positron Emission Tomography/Computed Tomography in Rats and Humans

Choi, Hongyoon,Han, Jeong Hee,Lim, Sue Yeon,Lee, Inki,Cho, Young-Seok,Chun, Eun Ju,Lee, Won Woo MIT PRESS 2017 MOLECULAR IMAGING Vol.16 No.-

<P>Positron emission tomography (PET)/computed tomography (CT) using sodium [<SUP>18</SUP>F]fluoride (Na[<SUP>18</SUP>F]F) has been proven to be a promising hot-spot imaging modality for myocardial infarction (MI). We investigated Na[<SUP>18</SUP>F]F uptake in ischemia–reperfusion injury (IRI) of rats and humans. Sodium [<SUP>18</SUP>F]fluoride PET/CT was performed in Sprague-Dawley rats that had IRI surgery, and it readily demonstrated prominent Na[<SUP>18</SUP>F]F uptake in the infarct area post-IRI. Sodium [<SUP>18</SUP>F]fluoride uptake was matched with negative 2,3,5-triphenyl-2<I>H</I>-tetrazolium chloride staining results, accompanied by myocardial apoptosis and associated with positive calcium staining results. Furthermore, area at risk was negative for Na[<SUP>18</SUP>F]F uptake. Cyclosporine A (CysA) treatment reduced standardized uptake value of <SUP>18</SUP>F over the infarct area, and a significant decrease in infarct size was also observed by the CysA treatment. In humans, Na[<SUP>18</SUP>F]F PET/CT readily demonstrated increased Na[<SUP>18</SUP>F]F uptake in the 2 patients with MI post-percutaneous coronary intervention. In conclusion, this study sheds light on the potential utility of Na[<SUP>18</SUP>F]F PET/CT as a hot-spot imaging modality for myocardial IRI.</P>

Generation of Structural MR Images from Amyloid PET: Application to MR-Less Quantification

Choi, Hongyoon,Lee, Dong Soo Society of Nuclear Medicine 2018 The Journal of nuclear medicine Vol.59 No.7

<P>Structural MR images concomitantly acquired with PET images can provide crucial anatomic information for precise quantitative analysis. However, in the clinical setting, not all the subjects have corresponding MR images. Here, we developed a model to generate structural MR images from amyloid PET using deep generative networks. We applied our model to quantification of cortical amyloid load without structural MR. <B>Methods:</B> We used florbetapir PET and structural MR data from the Alzheimer Disease Neuroimaging Initiative database. The generative network was trained to generate realistic structural MR images from florbetapir PET images. After the training, the model was applied to the quantification of cortical amyloid load. PET images were spatially normalized to the template space using the generated MR, and then SUV ratio (SUVR) of the target regions was measured by predefined regions of interest. A real MR-based quantification was used as the gold standard to measure the accuracy of our approach. Other MR-less methods—a normal PET template–based, a multiatlas PET template–based, and a PET segmentation–based normalization/quantification—were also tested. We compared the performance of quantification methods using generated MR with that of MR-based and MR-less quantification methods. <B>Results:</B> Generated MR images from florbetapir PET showed signal patterns that were visually similar to the real MR. The structural similarity index between real and generated MR was 0.91 ± 0.04. The mean absolute error of SUVR of cortical composite regions estimated by the generated MR-based method was 0.04 ± 0.03, which was significantly smaller than other MR-less methods (0.29 ± 0.12 for the normal PET template, 0.12 ± 0.07 for the multiatlas PET template, and 0.08 ± 0.06 for the PET segmentation–based methods). Bland–Altman plots revealed that the generated MR-based SUVR quantification was the closest to the SUVRs estimated by the real MR-based method. <B>Conclusion:</B> Structural MR images were successfully generated from amyloid PET images using deep generative networks. Generated MR images could be used as templates for accurate and precise amyloid quantification. This generative method might be used to generate multimodal images of various organs for further quantitative analyses.</P>

Deep Learning in Nuclear Medicine and Molecular Imaging: Current Perspectives and Future Directions

Choi, Hongyoon 대한핵의학회 2018 핵의학 분자영상 Vol.52 No.2

Recent advances in deep learning have impacted various scientific and industrial fields. Due to the rapid application of deep learning in biomedical data, molecular imaging has also started to adopt this technique. In this regard, it is expected that deep learning will potentially affect the roles of molecular imaging experts as well as clinical decision making. This review firstly offers a basic overview of deep learning particularly for image data analysis to give knowledge to nuclear medicine physicians and researchers. Because of the unique characteristics and distinctive aims of various types of molecular imaging, deep learning applications can be different from other fields. In this context, the review deals with current perspectives of deep learning in molecular imaging particularly in terms of development of biomarkers. Finally, future challenges of deep learning application for molecular imaging and future roles of experts in molecular imaging will be discussed.

Segmentation-Based MR Attenuation Correction Including Bones Also Affects Quantitation in Brain Studies: An Initial Result of ¹⁸F-FP-CIT PET/MR for Patients with Parkinsonism

Choi, Hongyoon,Cheon, Gi Jeong,Kim, Han-Joon,Choi, Seung Hong,Lee, Jae Sung,Kim, Yong-il,Kang, Keon Wook,Chung, June-Key,Kim, E. Edmund,Lee, Dong Soo Society of Nuclear Medicine 2014 The Journal of nuclear medicine Vol.55 No.10

<P>Attenuation correction (AC) with an ultrashort echo time (UTE) sequence has recently been used in combination with segmentation for cortical bone identification for brain PET/MR studies. The purpose of this study was to evaluate the quantification of <SUP>18</SUP>F-fluoropropyl-carbomethoxyiodophenylnortropane (<SUP>18</SUP>F-FP-CIT) binding in brain PET/MR, particularly focusing on effects of UTE-based AC including bone segmentation. <B>Methods:</B> Sixteen patients with initially suspected parkinsonism were prospectively enrolled. An emission scan was acquired 110 min after <SUP>18</SUP>F-FP-CIT injection on a dedicated PET/MR scanner, immediately followed by another emission scan using a PET/CT scanner 120 min after the injection. A UTE-based attenuation map was used to classify the voxels into 3 tissues: bone, soft tissue, and air. All PET images were spatially normalized, and a specific-to-nonspecific dopamine transporter (DAT) binding ratio (BR) was calculated using statistical probabilistic anatomic mapping. The level of agreement was assessed with intraclass correlation coefficients (ICCs). Voxelwise comparison between PET images acquired from PET/MR and PET/CT was performed. We compared non–attenuation-corrected images to analyze UTE-based AC effects on DAT quantification. <B>Results:</B> BR in the putamen obtained from PET/MR and PET/CT showed low interequipment variability, whereas BR in the caudate nucleus showed significant variability (ICC = 0.967 and 0.682 for putamen and caudate nucleus, respectively). BR in the caudate nucleus was significantly underestimated by PET/MR, compared with PET/CT (mean difference of BR = 0.66, <I>P</I> < 0.0001). Voxelwise analysis revealed that PET/MR showed significantly low BR in the periventricular regions, which was caused by a misclassification of the ventricle as air on the attenuation map. We also compared non-AC images, revealing low interequipment variability even in the caudate nucleus (ICC = 0.937 and 0.832 for putamen and caudate nucleus, respectively). <B>Conclusion:</B> Our data demonstrate spatial bias of the DAT BR on <SUP>18</SUP>F-FP-CIT PET/MR. Voxelwise analysis and comparison to non-AC images identified the misclassification of ventricle as air to be the cause of bias. To obtain reliable quantification for brain PET/MR studies including <SUP>18</SUP>F-FP-CIT PET, alternative and more reliable segmentation strategies are required.</P>

Voxel-based dual-time ¹⁸ F-FDG parametric imaging for rectal cancer: differentiation of residual tumor from postchemoradiotherapy changes

Choi, Hongyoon,Yoon, Hai-jeon,Kim, Tae Sung,Oh, Jae Hwan,Kim, Dae Yong,Kim, Seok-ki Lippincott WilliamsWilkins 2013 Nuclear medicine communications Vol.34 No.12

<▼1><P>Supplemental Digital Content is available in the text.</P></▼1><▼2><P><B>Introduction</B></P><P><SUP>18</SUP>F-Fluorodeoxyglucose (<SUP>18</SUP>F-FDG) PET/computed tomography (CT) has been used for evaluation of the response of rectal cancer to neoadjuvant chemoradiotherapy (CRT), but differentiating residual tumor from post-treatment changes remains a problem. We propose a voxel-based dual-time <SUP>18</SUP>F-FDG PET parametric imaging technique for the evaluation of residual rectal cancer after CRT.</P><P><B>Materials and methods</B></P><P>Eighty-six patients with locally advanced rectal cancer who underwent neoadjuvant CRT between March 2009 and February 2011 were selected retrospectively. Standard 60-min postinjection PET/CT scans followed by 90-min delayed images were coregistered by rigid-body transformation. A dual-time parametric image was generated, which divided delayed standardized uptake value (SUV) by 60-min SUV on a voxel-by-voxel basis. Maximum delayed-to-standard SUV ratios (DSR) measured on the parametric images as well as the percentage of SUV decrease from pre-CRT to post-CRT scans (pre/post-CRT response index) were obtained for each tumor and correlated with pathologic response classified by the Dworak tumor regression grade (TRG).</P><P><B>Results</B></P><P>With respect to the false-positive lesions in the nine post-CRT patients with false-positive standard <SUP>18</SUP>F-FDG scans in case groups who responded to therapy (TRG 3 or 4 tumors), eight were undetectable on dual-time parametric images (<I>P</I><0.05). The maximum DSR showed significantly higher accuracy for identification of tumor regression compared with the pre/post-CRT response index in receiver-operating characteristic analysis (<I>P</I><0.01). With a 1.25 cutoff value for the maximum DSR, 85.0% sensitivity, 95.5% specificity, and 93.0% overall accuracy were obtained for identification of good response.</P><P><B>Conclusion</B></P><P>A voxel-based dual-time parametric imaging technique for evaluation of post-CRT rectal cancer holds promise for differentiating residual tumor from treatment-related nonspecific <SUP>18</SUP>F-FDG uptake.</P></▼2>

Spatiotemporal characterization of glial cell activation in an Alzheimer’s disease model by spatially resolved transcriptomics

Choi Hongyoon,Lee Eun Ji,Shin Jin Seop,Kim Hyun Je,Bae Sungwoo,Choi Yoori,Lee Dong Soo 생화학분자생물학회 2023 Experimental and molecular medicine Vol.55 No.-

The molecular changes that occur with the progression of Alzheimer’s disease (AD) are well known, but an understanding of the spatiotemporal heterogeneity of changes in the brain is lacking. Here, we investigated the spatially resolved transcriptome in a 5XFAD AD model at different ages to understand regional changes at the molecular level. Spatially resolved transcriptomic data were obtained from 5XFAD AD models and age-matched control mice. Differentially expressed genes were identified using spots clustered by anatomical structures. Gene signatures of activation of microglia and astrocytes were calculated and mapped on the spatially resolved transcriptomic data. We identified early alterations in the white matter (WM) of the AD model before the definite accumulation of amyloid plaques in the gray matter (GM). Changes in the early stage of the disease involved primarily glial cell activation in the WM, whereas the changes in the later stage of pathology were prominent in the GM. We confirmed that disease-associated microglia (DAM) and astrocyte (DAA) signatures also showed initial changes in WM and that activation spreads to GM. Trajectory inference using microglial gene sets revealed the subdivision of DAMs with different spatial patterns. Taken together, these results help to understand the spatiotemporal changes associated with reactive glial cells as a major pathophysiological characteristic of AD. The heterogeneous spatial molecular changes apply to identifying diagnostic and therapeutic targets caused by amyloid accumulation in AD.

Predicting Aging of Brain Metabolic Topography Using Variational Autoencoder

Choi, Hongyoon,Kang, Hyejin,Lee, Dong Soo Frontiers Media S.A. 2018 FRONTIERS IN AGING NEUROSCIENCE Vol.10 No.-

<P>Predicting future brain topography can give insight into neural correlates of aging and neurodegeneration. Due to variability in the aging process, it has been challenging to precisely estimate brain topographical change according to aging. Here, we predict age-related brain metabolic change by generating future brain <SUP>18</SUP>F-Fluorodeoxyglucose PET. A cross-sectional PET dataset of cognitively normal subjects with different age was used to develop a generative model. The model generated PET images using age information and characteristic individual features. Predicted regional metabolic changes were correlated with the real changes obtained by follow-up data. This model was applied to produce a brain metabolism aging movie by generating PET at different ages. Normal population distribution of brain metabolic topography at each age was estimated as well. In addition, a generative model using APOE4 status as well as age as inputs revealed a significant effect of APOE4 status on age-related metabolic changes particularly in the calcarine, lingual cortex, hippocampus, and amygdala. It suggested APOE4 could be a factor affecting individual variability in age-related metabolic degeneration in normal elderly. This predictive model may not only be extended to understanding the cognitive aging process, but apply to the development of a preclinical biomarker for various brain disorders.</P>

A Case of Habitual Neck Compression Induced Electroencephalogram Abnormalities: Differentiating from Epileptic Seizures Using a Tc-99m HMPAO SPECT

Choi, Hongyoon,Seo, Min-Seok,Lee, Ho-Young,Kim, Young-Soo,Yun, Chang-Ho,Kim, Sang Eun,Park, Sung-Ho 大韓核醫學會 2014 핵의학 분자영상 Vol.48 No.2

<P>Self-induced hypoxia has been reported particularly in adolescents, and it can result in neurological injury. Here, we present a case of electroencephalogram (EEG) abnormalities induced by habitual neck compression differentiated from epileptic seizures by Tc-99m HMPAO SPECT. A 19-year-old male was admitted for evaluation of recurrent generalized tonic-clonic seizures. No interictal EEG abnormality was detected; however, abnormal slow delta waves were found immediately after habitual right neck compression. To differentiate EEG abnormalities due to a hemodynamic deficit induced by habitual neck compression from an epileptic seizure, Tc-99m HMPAO SPECT was performed immediately after right carotid artery compression. Abnormal delta waves were triggered, and cerebral hypoperfusion in the right internal carotid artery territory was detected on Tc-99m HMPAO SPECT. The slow delta wave detected on the EEG resulted from the cerebral hypoperfusion because of the habitual neck compression.</P>