Kyoung Doo Song
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Kyoung Doo Song has a total of 141 co-authors. Here are the top co-authors by number of shared publications:
- Tae Wook Kang
Department of Radiology and Center for Imaging Science, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Korea
19 shared publications - Min Woo Lee
Department of Radiology, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea
18 shared publications - Hyunchul Rhim
Department of Radiology and Center for Imaging Science, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Korea
11 shared publications - Dong Ik Cha
Department of Radiology and Center for Imaging Science, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea
10 shared publications - Hyunchul Rhim
Department of Radiology and Center for Imaging Science, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea
7 shared publications - Soohyun Ahn
Ajou University
6 shared publications
Latest Publications
- Open Access
- 1 January 2026
- Published by Korean Society of Ultrasound in Medicine in Ultrasonography
- Vol. 45 (1) , 59-68
Prospective evaluation of the safety and efficacy of microwave ablation for hepatocellular carcinoma in Korean patients
Purpose: This study evaluated the 3-year safety and efficacy of microwave ablation (MWA) as an alternative to radiofrequency ablation in Korean patients with hepatocellular carcinoma (HCC).Methods: In this prospective single-center study, 33 adults with a single HCC lesion (2–5 cm; Barcelona Clinic Liver Cancer stage A) were enrolled between December 2017 and May 2020. All underwent ultrasound-guided MWA using the NEUWAVE system. One patient did not complete post-treatment evaluation and two had protocol deviations, yielding a full analysis set of 30 patients and a safety analysis set of 33 patients. Study endpoints included technical success (TS), technique efficacy (TE), local tumor progression (LTP), overall survival (OS), and adverse events (AEs) over a 3-year follow-up period.Results: All 33 patients were included in the safety analysis, while 30 were included in the efficacy analysis. Most patients were male (83.3%), with a median age of 64 years and a mean tumor size of 2.5±0.6 cm. Median follow-up was 36.2 months. TS and TE were achieved in all cases. The 3-year cumulative LTP rate was 7.3%, and the 3-year OS rate was 90.0%. Minor AEs occurred in 54.5% of patients, most commonly post-ablation pain and fever. Two patients (6.1%) experienced major AEs, including pyrexia and duodenal thermal injury. No procedure- or device-related deaths occurred.Conclusion: MWA appears to be a safe and effective treatment for early-stage HCC, demonstrating durable tumor control and a low incidence of major complications over 3 years.
- 1 January 2026
- Published by Korean Society of Ultrasound in Medicine in Ultrasonography
- Vol. 45 (1) , 59-68
Open AccessProspective evaluation of the safety and efficacy of microwave ablation for hepatocellular carcinoma in Korean patients
Purpose: This study evaluated the 3-year safety and efficacy of microwave ablation (MWA) as an alternative to radiofrequency ablation in Korean patients with hepatocellular carcinoma (HCC).Methods: In this prospective single-center study, 33 adults with a single HCC lesion (2–5 cm; Barcelona Clinic Liver Cancer stage A) were enrolled between December 2017 and May 2020. All underwent ultrasound-guided MWA using the NEUWAVE system. One patient did not complete post-treatment evaluation and two had protocol deviations, yielding a full analysis set of 30 patients and a safety analysis set of 33 patients. Study endpoints included technical success (TS), technique efficacy (TE), local tumor progression (LTP), overall survival (OS), and adverse events (AEs) over a 3-year follow-up period.Results: All 33 patients were included in the safety analysis, while 30 were included in the efficacy analysis. Most patients were male (83.3%), with a median age of 64 years and a mean tumor size of 2.5±0.6 cm. Median follow-up was 36.2 months. TS and TE were achieved in all cases. The 3-year cumulative LTP rate was 7.3%, and the 3-year OS rate was 90.0%. Minor AEs occurred in 54.5% of patients, most commonly post-ablation pain and fever. Two patients (6.1%) experienced major AEs, including pyrexia and duodenal thermal injury. No procedure- or device-related deaths occurred.Conclusion: MWA appears to be a safe and effective treatment for early-stage HCC, demonstrating durable tumor control and a low incidence of major complications over 3 years.
- 23 September 2025
- Published by Springer Nature in European Radiology
- Vol. 36 (3) , 1864-1873
MRI features of lateral lymph nodes predicting lateral local recurrence and prognostic impact of lateral lymph node dissection in low rectal cancer
Objectives To evaluate MRI features of lateral lymph nodes (LLNs) associated with lateral local recurrence (LLR) and assess the prognostic value of lateral lymph node dissection (LLND) in low rectal cancer. Materials and methods This retrospective study included consecutive patients with cT3-4 rectal cancer located ≤ 8 cm from the anal verge who underwent radical surgery following neoadjuvant chemoradiotherapy. LLN location, short-axis diameter, and malignant features (internal heterogeneity or irregular borders) were assessed. LLR rates were estimated using Kaplan–Meier analysis and compared with the log-rank test. Cox proportional hazards models identified factors associated with LLR. Results A total of 878 patients (mean age, 57 ± 12 years; 574 males) were included. The cumulative 5-year LLR rate was 4.2%. Patients with LLNs ≥ 7 mm had higher LLR rates than those with LLNs < 7 mm or no LLNs (17.2% vs. 2.6% vs. 3.1% at 5 years, p < 0.001). Malignant features on pretreatment MRI were associated with higher LLR rates than LLNs without malignant features or no LLNs (28.2% vs. 2.6% vs. 3.1% at 5 years, p < 0.001). In multivariable analysis, malignant features remained independent predictors of LLR (hazard ratio = 9.0 [1.9–43.9], p = 0.006). LLND reduced LLR in patients with internal iliac lymph nodes with malignant features (14.2% vs. 52.9% at 5 years, p = 0.01). Conclusion Malignant features of LLNs on pretreatment MRI are associated with increased risk of LLR. LLND significantly reduces LLR in patients with internal iliac lymph nodes with malignant features. Key Points Question The optimal therapeutic approach for rectal cancer patients with abnormal lateral lymph nodes on the staging MRI remains controversial. Findings Malignant features in lateral lymph nodes independently predicted lateral local recurrence, and lateral lymph node dissection reduced recurrence for internal iliac lymph nodes with malignant features. Clinical relevance Identifying malignant features of lateral lymph nodes on pretreatment MRI enables risk stratification for lateral local recurrence and helps guide the need for lateral lymph node dissection, potentially improving oncologic outcomes in patients with low rectal cancer. Graphical Abstract
- 23 September 2025
- Published by Springer Nature in European Radiology
- Vol. 36 (3) , 1864-1873
MRI features of lateral lymph nodes predicting lateral local recurrence and prognostic impact of lateral lymph node dissection in low rectal cancer
Objectives To evaluate MRI features of lateral lymph nodes (LLNs) associated with lateral local recurrence (LLR) and assess the prognostic value of lateral lymph node dissection (LLND) in low rectal cancer. Materials and methods This retrospective study included consecutive patients with cT3-4 rectal cancer located ≤ 8 cm from the anal verge who underwent radical surgery following neoadjuvant chemoradiotherapy. LLN location, short-axis diameter, and malignant features (internal heterogeneity or irregular borders) were assessed. LLR rates were estimated using Kaplan–Meier analysis and compared with the log-rank test. Cox proportional hazards models identified factors associated with LLR. Results A total of 878 patients (mean age, 57 ± 12 years; 574 males) were included. The cumulative 5-year LLR rate was 4.2%. Patients with LLNs ≥ 7 mm had higher LLR rates than those with LLNs < 7 mm or no LLNs (17.2% vs. 2.6% vs. 3.1% at 5 years, p < 0.001). Malignant features on pretreatment MRI were associated with higher LLR rates than LLNs without malignant features or no LLNs (28.2% vs. 2.6% vs. 3.1% at 5 years, p < 0.001). In multivariable analysis, malignant features remained independent predictors of LLR (hazard ratio = 9.0 [1.9–43.9], p = 0.006). LLND reduced LLR in patients with internal iliac lymph nodes with malignant features (14.2% vs. 52.9% at 5 years, p = 0.01). Conclusion Malignant features of LLNs on pretreatment MRI are associated with increased risk of LLR. LLND significantly reduces LLR in patients with internal iliac lymph nodes with malignant features. Key Points Question The optimal therapeutic approach for rectal cancer patients with abnormal lateral lymph nodes on the staging MRI remains controversial. Findings Malignant features in lateral lymph nodes independently predicted lateral local recurrence, and lateral lymph node dissection reduced recurrence for internal iliac lymph nodes with malignant features. Clinical relevance Identifying malignant features of lateral lymph nodes on pretreatment MRI enables risk stratification for lateral local recurrence and helps guide the need for lateral lymph node dissection, potentially improving oncologic outcomes in patients with low rectal cancer. Graphical Abstract
- Open Access
- 24 April 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (4) , e0321085
Correction of stomach cancer CT attenuation values for variations due to differences in CT imaging conditions through repeated CT scans
Purpose: To develop methods for correcting variations in CT attenuation values of advanced gastric cancer (AGC) due to differences in CT imaging conditions using repeated pre-treatment CT scans. Methods: A total 211 patients (146 men) with AGC who underwent pre-treatment CT twice were included in this retrospective study. The Pearson correlation between the difference in tumor attenuation values measured on both CT scans and the difference in attenuation values of other organs was analyzed. A formula to correct tumor CT attenuation values was developed using univariate linear regression analysis. Results: The Pearson correlation coefficient was the highest between the difference in tumor attenuation values and that of the main portal vein (MPV) attenuation values (0.86, P <.01). The formula to correct tumor attenuation values was as follows: calculated tumor attenuation value on CT scan 2 = tumor attenuation value on CT scan 1 - (-3.5 + 0.4 x (MPV attenuation value on CT scan 1 - MPV attenuation value on CT scan 2)). The mean difference between calculated and actual tumor attenuation values was 1.6 HU (SD, 8.7; range -22.5–24.72), with a Pearson correlation coefficient of 0.95 (P <.01). Conclusion: Utilizing the attenuation value of the MPV allows for correction of variations in tumor attenuation values caused by different CT imaging conditions, enabling the prediction of reproducible tumor attenuation in patients with AGC. Future studies are needed to validate these findings and address the study’s limitations, including its retrospective design and the absence of unenhanced CT data. Advances in knowledge: The attenuation value of the MPV can be used to predict reproducible tumor attenuation values in gastric cancer.
- 24 April 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (4) , e0321085
Open AccessCorrection of stomach cancer CT attenuation values for variations due to differences in CT imaging conditions through repeated CT scans
Purpose: To develop methods for correcting variations in CT attenuation values of advanced gastric cancer (AGC) due to differences in CT imaging conditions using repeated pre-treatment CT scans. Methods: A total 211 patients (146 men) with AGC who underwent pre-treatment CT twice were included in this retrospective study. The Pearson correlation between the difference in tumor attenuation values measured on both CT scans and the difference in attenuation values of other organs was analyzed. A formula to correct tumor CT attenuation values was developed using univariate linear regression analysis. Results: The Pearson correlation coefficient was the highest between the difference in tumor attenuation values and that of the main portal vein (MPV) attenuation values (0.86, P <.01). The formula to correct tumor attenuation values was as follows: calculated tumor attenuation value on CT scan 2 = tumor attenuation value on CT scan 1 - (-3.5 + 0.4 x (MPV attenuation value on CT scan 1 - MPV attenuation value on CT scan 2)). The mean difference between calculated and actual tumor attenuation values was 1.6 HU (SD, 8.7; range -22.5–24.72), with a Pearson correlation coefficient of 0.95 (P <.01). Conclusion: Utilizing the attenuation value of the MPV allows for correction of variations in tumor attenuation values caused by different CT imaging conditions, enabling the prediction of reproducible tumor attenuation in patients with AGC. Future studies are needed to validate these findings and address the study’s limitations, including its retrospective design and the absence of unenhanced CT data. Advances in knowledge: The attenuation value of the MPV can be used to predict reproducible tumor attenuation values in gastric cancer.
- Open Access
- 1 January 2025
- Published by XMLink in Journal of the Korean Society of Radiology
- Vol. 86 (6) , 882-889
Past, Present, and Future of the Radiology Board Examination: Transition Toward Assessment of Actual Reading Ability
이인숙, et al. J Korean Soc Radiol. 2025 Nov;86(6):882-889. https://doi.org/10.3348/jksr.2025.0086
- 1 January 2025
- Published by XMLink in Journal of the Korean Society of Radiology
- Vol. 86 (6) , 882-889
Open AccessPast, Present, and Future of the Radiology Board Examination: Transition Toward Assessment of Actual Reading Ability
이인숙, et al. J Korean Soc Radiol. 2025 Nov;86(6):882-889. https://doi.org/10.3348/jksr.2025.0086
- Open Access
- 17 August 2024
- Published by Springer Nature in Scientific Reports
- Vol. 14 (1) , 1-9
Synthesis and quality assessment of combined time-series and static medical data using a real-world time-series generative adversarial network
This study addresses challenges related to privacy issues in utilizing medical data, particularly the protection of personal information. To overcome this obstacle, the research focuses on data synthesis using real-world time-series generative adversarial networks (RTSGAN). A total of 53,005 data were synthesized using the dataset of 15,799 patients with colorectal cancer. The results of the quantitative evaluation of the synthetic data’s quality are as follows: the Hellinger distance ranged from 0 to 0.25; the train on synthetic, test on real (TSTR) and train on real, test on synthetic (TRTS) results showed an average area under the curve of 0.99 and 0.98; a propensity mean squared error was 0.223. The synthetic and real data were similar in the qualitative methods including t-SNE and histogram analyses. The application of synthetic data in predicting five-year survival in colorectal cancer patients demonstrates comparable performance to models based on real data. This study employs distance to closest records and membership inference test to assess potential privacy exposure, revealing minimal risk. This study demonstrated that it is feasible to synthesize medical data, including time-series data, using the RTSGAN, and the synthetic data can be evaluated to accurately reflect the characteristics of real data through quantitative and qualitative methods as well as by utilizing real-world artificial intelligence models.
- 17 August 2024
- Published by Springer Nature in Scientific Reports
- Vol. 14 (1) , 1-9
Open AccessSynthesis and quality assessment of combined time-series and static medical data using a real-world time-series generative adversarial network
This study addresses challenges related to privacy issues in utilizing medical data, particularly the protection of personal information. To overcome this obstacle, the research focuses on data synthesis using real-world time-series generative adversarial networks (RTSGAN). A total of 53,005 data were synthesized using the dataset of 15,799 patients with colorectal cancer. The results of the quantitative evaluation of the synthetic data’s quality are as follows: the Hellinger distance ranged from 0 to 0.25; the train on synthetic, test on real (TSTR) and train on real, test on synthetic (TRTS) results showed an average area under the curve of 0.99 and 0.98; a propensity mean squared error was 0.223. The synthetic and real data were similar in the qualitative methods including t-SNE and histogram analyses. The application of synthetic data in predicting five-year survival in colorectal cancer patients demonstrates comparable performance to models based on real data. This study employs distance to closest records and membership inference test to assess potential privacy exposure, revealing minimal risk. This study demonstrated that it is feasible to synthesize medical data, including time-series data, using the RTSGAN, and the synthetic data can be evaluated to accurately reflect the characteristics of real data through quantitative and qualitative methods as well as by utilizing real-world artificial intelligence models.
- 12 June 2024
- Published by Elsevier in European Journal of Radiology
- Vol. 177, 111560
Gastrointestinal tract perforation after radiofrequency ablation for hepatic tumor: Incidence and risk factors
- 12 June 2024
- Published by Elsevier in European Journal of Radiology
- Vol. 177, 111560
Gastrointestinal tract perforation after radiofrequency ablation for hepatic tumor: Incidence and risk factors
- Open Access
- 24 May 2024
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 19 (5) , e0304352
Value of Sonazoid-enhanced ultrasonography in characterizing indeterminate focal liver lesions on gadoxetic acid-enhanced liver MRI in patients without risk factors for hepatocellular carcinoma
Purpose: To evaluate the added value of contrast-enhanced ultrasonography (CEUS) using Sonazoid in characterizing focal liver lesions (FLLs) with indeterminate findings on gadoxetic acid-enhanced liver MRI in patients without risk factors for hepatocellular carcinoma (HCC). Methods: Patients who underwent CEUS using Sonazoid for characterizing indeterminate FLLs on gadoxetic acid-enhanced liver MRI were. The indeterminate FLLs were classified according to the degree of malignancy on a 5-point scale on MRI and combined MRI and CEUS. The final diagnosis was made either pathologically or based on more than one-year follow-up. The diagnostic performance was assessed using a receiver operating characteristic (ROC) curve analysis, and the net reclassification improvement (NRI) was calculated. Results: A total of 97 patients (mean age, 49 years ± 16, 41 men, 80 benign and 17 malignant lesions) were included. When CEUS was added to MRI, the area under the ROC curve increased, but the difference was not statistically significant (0.87 [95% confidence interval {CI}, 0.77–0.98] for MRI vs 0.93 [95% CI, 0.87–0.99] for CEUS added to MRI, P = 0.296). The overall NRI was 0.473 (95% CI, 0.100–0.845; P = 0.013): 33.8% (27/80) of benign lesions and 41.2% (7/17) of malignant lesions were appropriately reclassified, whereas 10.0% (8/80) of benign lesions and 17.6% (3/17) of malignant lesions were incorrectly reclassified. Conclusions: Although performing CEUS with Sonazoid did not significantly improve the overall diagnostic performance in characterizing indeterminate FLLs on gadoxetic acid-enhanced liver MRI in patients without risk factors for HCC, it may increase radiologist’s confidence in classifying FLLs.
- 24 May 2024
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 19 (5) , e0304352
Open AccessValue of Sonazoid-enhanced ultrasonography in characterizing indeterminate focal liver lesions on gadoxetic acid-enhanced liver MRI in patients without risk factors for hepatocellular carcinoma
Purpose: To evaluate the added value of contrast-enhanced ultrasonography (CEUS) using Sonazoid in characterizing focal liver lesions (FLLs) with indeterminate findings on gadoxetic acid-enhanced liver MRI in patients without risk factors for hepatocellular carcinoma (HCC). Methods: Patients who underwent CEUS using Sonazoid for characterizing indeterminate FLLs on gadoxetic acid-enhanced liver MRI were. The indeterminate FLLs were classified according to the degree of malignancy on a 5-point scale on MRI and combined MRI and CEUS. The final diagnosis was made either pathologically or based on more than one-year follow-up. The diagnostic performance was assessed using a receiver operating characteristic (ROC) curve analysis, and the net reclassification improvement (NRI) was calculated. Results: A total of 97 patients (mean age, 49 years ± 16, 41 men, 80 benign and 17 malignant lesions) were included. When CEUS was added to MRI, the area under the ROC curve increased, but the difference was not statistically significant (0.87 [95% confidence interval {CI}, 0.77–0.98] for MRI vs 0.93 [95% CI, 0.87–0.99] for CEUS added to MRI, P = 0.296). The overall NRI was 0.473 (95% CI, 0.100–0.845; P = 0.013): 33.8% (27/80) of benign lesions and 41.2% (7/17) of malignant lesions were appropriately reclassified, whereas 10.0% (8/80) of benign lesions and 17.6% (3/17) of malignant lesions were incorrectly reclassified. Conclusions: Although performing CEUS with Sonazoid did not significantly improve the overall diagnostic performance in characterizing indeterminate FLLs on gadoxetic acid-enhanced liver MRI in patients without risk factors for HCC, it may increase radiologist’s confidence in classifying FLLs.
- Open Access
- 30 April 2024
- Published by Korean Association for the Study of Intestinal Diseases in Intestinal Research
- Vol. 22 (2) , 117-118
Achieving high-quality magnetic resonance enterography is critical for assessing Crohn’s disease activity
Thicker. Thus, achieving optimal bowel distension is essential for a precise evaluation of bowel wall thickness. This necessitates choosing the right oral contrast agent and fine-tuning the dosage and administration method. MRE involves ingesting a contrast agent orally, often resulting in insufficient distension of the proximal jejunum. Bohra et al. ' s findings echoed this issue. Although magnetic resonance enteroclysis can more reliably distend the intestines than MRE, its requirement for additional intubation can cause patient discomfort. Since Crohn' s disease more commonly affects the ileum than the jejunum, MRE remains the preferred method for most cases of Crohn' s disease. The irregular peristaltic movement of the small intestine can significantly degrade image quality, a concern for MRE given the lengthy image acquisition time. Antiperistaltic agents like glucagon and hyoscine butylbromide are used to mitigate peristalsis. The onset and duration of their effects vary based on the medication type and administration method. 6 While it typically takes more than 30 minutes to acquire a complete MRE, bowel movement begins to reappear after a mean period of 6.8 ± 5.3 minutes in hyoscine butylbromide intravenous (IV) group and after 18.3 ± 7.0 minutes in glucagon IV group. 7 This means that using a single medication may result in lower image quality towards the latter part of the image acquisition due to peristalsis. To overcome this, combination of 2 medications can be used, or the medication can be split and administered both at the start and midway through the examination. The type, amount, and method of administering oral con
- 30 April 2024
- Published by Korean Association for the Study of Intestinal Diseases in Intestinal Research
- Vol. 22 (2) , 117-118
Open AccessAchieving high-quality magnetic resonance enterography is critical for assessing Crohn’s disease activity
Thicker. Thus, achieving optimal bowel distension is essential for a precise evaluation of bowel wall thickness. This necessitates choosing the right oral contrast agent and fine-tuning the dosage and administration method. MRE involves ingesting a contrast agent orally, often resulting in insufficient distension of the proximal jejunum. Bohra et al. ' s findings echoed this issue. Although magnetic resonance enteroclysis can more reliably distend the intestines than MRE, its requirement for additional intubation can cause patient discomfort. Since Crohn' s disease more commonly affects the ileum than the jejunum, MRE remains the preferred method for most cases of Crohn' s disease. The irregular peristaltic movement of the small intestine can significantly degrade image quality, a concern for MRE given the lengthy image acquisition time. Antiperistaltic agents like glucagon and hyoscine butylbromide are used to mitigate peristalsis. The onset and duration of their effects vary based on the medication type and administration method. 6 While it typically takes more than 30 minutes to acquire a complete MRE, bowel movement begins to reappear after a mean period of 6.8 ± 5.3 minutes in hyoscine butylbromide intravenous (IV) group and after 18.3 ± 7.0 minutes in glucagon IV group. 7 This means that using a single medication may result in lower image quality towards the latter part of the image acquisition due to peristalsis. To overcome this, combination of 2 medications can be used, or the medication can be split and administered both at the start and midway through the examination. The type, amount, and method of administering oral con
- Open Access
- 4 January 2024
- Published by JMIR Publications Inc. in JMIR Preprints
Synthesis and Quality Assessment of Combined Time-Series and Static Medical Data Using a Real-world Time-Series Generative Adversarial Network (Preprint)
BACKGROUND The use of medical data often faces challenges due to personal information protection issues. To solve this problem, methods for synthesizing data using generative models are gaining attention. Synthesizing medical data using generative models is a new field, and there are no established methods for evaluating the quality of synthetic data. OBJECTIVE To synthesize medical data using real-world time-series generative adversarial networks (RTSGAN), evaluate the quality of synthetic data through quantitative and qualitative methods, apply them to real-world medical artificial intelligence (AI) models, and assess the disclosure risk of synthetic data. METHODS Data were synthesized using the RTSGAN based on a real dataset of 15,799 patients with colorectal cancer. The quality of the synthetic data was evaluated using quantitative methods such as the Hellinger distance; train on synthetic, test on real (TSTR); train on real, test on synthetic (TRTS); and propensity mean squared error (PMSE); and qualitative methods including t-SNE and histogram. We applied the synthetic data to a real-world model predicting the five-year survival of patients with colorectal cancer. Thereafter, we compared its performance with a model using real data, employing measures such as the C-index, Brier score, and integrated Brier score (IBS). Finally, we measured the distance between the synthetic and real data using DCR (Distance to Closest Records) to assess the potential for privacy exposure. RESULTS In total, 53,005 data points were obtained. The Hellinger distance ranged from 0 to 0.25. The TSTR and TRTS results showed an average area under the curve of 0.99 and 0.98 and a propensity MSE was 0.223. The synthetic and real data were confirmed to be similar in the t-SNE and histogram analyses. The C-index, Brier score, and IBS for the models using synthetic data and real data were 0.742, 0.075, 0.107, 0.777, 0.098, and 0.138, respectively. The DCR evaluation shows that the minimum distance between real data is 2.45, and the minimum distance between real and synthetic data is 3.46. CONCLUSIONS It is feasible to synthesize combined time-series and static medical data using the RTSGAN, and the synthetic data can be evaluated to accurately reflect the characteristics of real data through quantitative and qualitative methods as well as by utilizing real-world AI models. Additionally, our investigation confirms that the synthesized data poses no privacy concerns.
- 4 January 2024
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessSynthesis and Quality Assessment of Combined Time-Series and Static Medical Data Using a Real-world Time-Series Generative Adversarial Network (Preprint)
BACKGROUND The use of medical data often faces challenges due to personal information protection issues. To solve this problem, methods for synthesizing data using generative models are gaining attention. Synthesizing medical data using generative models is a new field, and there are no established methods for evaluating the quality of synthetic data. OBJECTIVE To synthesize medical data using real-world time-series generative adversarial networks (RTSGAN), evaluate the quality of synthetic data through quantitative and qualitative methods, apply them to real-world medical artificial intelligence (AI) models, and assess the disclosure risk of synthetic data. METHODS Data were synthesized using the RTSGAN based on a real dataset of 15,799 patients with colorectal cancer. The quality of the synthetic data was evaluated using quantitative methods such as the Hellinger distance; train on synthetic, test on real (TSTR); train on real, test on synthetic (TRTS); and propensity mean squared error (PMSE); and qualitative methods including t-SNE and histogram. We applied the synthetic data to a real-world model predicting the five-year survival of patients with colorectal cancer. Thereafter, we compared its performance with a model using real data, employing measures such as the C-index, Brier score, and integrated Brier score (IBS). Finally, we measured the distance between the synthetic and real data using DCR (Distance to Closest Records) to assess the potential for privacy exposure. RESULTS In total, 53,005 data points were obtained. The Hellinger distance ranged from 0 to 0.25. The TSTR and TRTS results showed an average area under the curve of 0.99 and 0.98 and a propensity MSE was 0.223. The synthetic and real data were confirmed to be similar in the t-SNE and histogram analyses. The C-index, Brier score, and IBS for the models using synthetic data and real data were 0.742, 0.075, 0.107, 0.777, 0.098, and 0.138, respectively. The DCR evaluation shows that the minimum distance between real data is 2.45, and the minimum distance between real and synthetic data is 3.46. CONCLUSIONS It is feasible to synthesize combined time-series and static medical data using the RTSGAN, and the synthetic data can be evaluated to accurately reflect the characteristics of real data through quantitative and qualitative methods as well as by utilizing real-world AI models. Additionally, our investigation confirms that the synthesized data poses no privacy concerns.
- Open Access
- 1 January 2024
- Published by XMLink in Korean Journal of Radiology
- Vol. 25 (7) , 613-622
2023 Survey on User Experience of Artificial Intelligence Software in Radiology by the Korean Society of Radiology
Hwang EJ, et al. Korean J Radiol. 2024 Jul;25(7):613-622. https://doi.org/10.3348/kjr.2023.1246
- 1 January 2024
- Published by XMLink in Korean Journal of Radiology
- Vol. 25 (7) , 613-622
Open Access2023 Survey on User Experience of Artificial Intelligence Software in Radiology by the Korean Society of Radiology
Hwang EJ, et al. Korean J Radiol. 2024 Jul;25(7):613-622. https://doi.org/10.3348/kjr.2023.1246