1
|
Kadota T, Hayashi H, Bise R, Tanaka K, Uchida S. Deep Bayesian active learning-to-rank with relative annotation for estimation of ulcerative colitis severity. Med Image Anal 2024; 97:103262. [PMID: 38986351 DOI: 10.1016/j.media.2024.103262] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 11/03/2023] [Revised: 05/04/2024] [Accepted: 06/27/2024] [Indexed: 07/12/2024]
Abstract
Automatic image-based severity estimation is an important task in computer-aided diagnosis. Severity estimation by deep learning requires a large amount of training data to achieve a high performance. In general, severity estimation uses training data annotated with discrete (i.e., quantized) severity labels. Annotating discrete labels is often difficult in images with ambiguous severity, and the annotation cost is high. In contrast, relative annotation, in which the severity between a pair of images is compared, can avoid quantizing severity and thus makes it easier. We can estimate relative disease severity using a learning-to-rank framework with relative annotations, but relative annotation has the problem of the enormous number of pairs that can be annotated. Therefore, the selection of appropriate pairs is essential for relative annotation. In this paper, we propose a deep Bayesian active learning-to-rank that automatically selects appropriate pairs for relative annotation. Our method preferentially annotates unlabeled pairs with high learning efficiency from the model uncertainty of the samples. We prove the theoretical basis for adapting Bayesian neural networks to pairwise learning-to-rank and demonstrate the efficiency of our method through experiments on endoscopic images of ulcerative colitis on both private and public datasets. We also show that our method achieves a high performance under conditions of significant class imbalance because it automatically selects samples from the minority classes.
Collapse
Affiliation(s)
- Takeaki Kadota
- Department of Advanced Information Technology, Kyushu University, 744, Motooka, Nishi-ku, Fukuoka-shi, Fukuoka, 819-0395, Japan.
| | - Hideaki Hayashi
- Institute for Datability Science, Osaka University, 2-8, Yamadaoka, Suita-shi, Osaka, 565-0871, Japan
| | - Ryoma Bise
- Department of Advanced Information Technology, Kyushu University, 744, Motooka, Nishi-ku, Fukuoka-shi, Fukuoka, 819-0395, Japan; Research Center for Medical Bigdata, National Institute of Informatics, 2-1-2, Hitotsubashi, Chiyoda-ku, Tokyo, 101-8430, Japan
| | - Kiyohito Tanaka
- Department of Gastroenterology, Kyoto Second Red Cross Hospital, 355-5, Haruobicho Kamigyo-ku, Kyoto-shi, Kyoto, 602-8026, Japan
| | - Seiichi Uchida
- Department of Advanced Information Technology, Kyushu University, 744, Motooka, Nishi-ku, Fukuoka-shi, Fukuoka, 819-0395, Japan; Research Center for Medical Bigdata, National Institute of Informatics, 2-1-2, Hitotsubashi, Chiyoda-ku, Tokyo, 101-8430, Japan
| |
Collapse
|
2
|
Takabayashi K, Kobayashi T, Matsuoka K, Levesque BG, Kawamura T, Tanaka K, Kadota T, Bise R, Uchida S, Kanai T, Ogata H. Artificial intelligence quantifying endoscopic severity of ulcerative colitis in gradation scale. Dig Endosc 2024; 36:582-590. [PMID: 37690125 DOI: 10.1111/den.14677] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 05/26/2023] [Accepted: 09/05/2023] [Indexed: 09/12/2023]
Abstract
OBJECTIVES Existing endoscopic scores for ulcerative colitis (UC) objectively categorize disease severity based on the presence or absence of endoscopic findings; therefore, it may not reflect the range of clinical severity within each category. However, inflammatory bowel disease (IBD) expert endoscopists categorize the severity and diagnose the overall impression of the degree of inflammation. This study aimed to develop an artificial intelligence (AI) system that can accurately represent the assessment of the endoscopic severity of UC by IBD expert endoscopists. METHODS A ranking-convolutional neural network (ranking-CNN) was trained using comparative information on the UC severity of 13,826 pairs of endoscopic images created by IBD expert endoscopists. Using the trained ranking-CNN, the UC Endoscopic Gradation Scale (UCEGS) was used to express severity. Correlation coefficients were calculated to ensure that there were no inconsistencies in assessments of severity made using UCEGS diagnosed by the AI and the Mayo Endoscopic Subscore, and the correlation coefficients of the mean for test images assessed using UCEGS by four IBD expert endoscopists and the AI. RESULTS Spearman's correlation coefficient between the UCEGS diagnosed by AI and Mayo Endoscopic Subscore was approximately 0.89. The correlation coefficients between IBD expert endoscopists and the AI of the evaluation results were all higher than 0.95 (P < 0.01). CONCLUSIONS The AI developed here can diagnose UC severity endoscopically similar to IBD expert endoscopists.
Collapse
Affiliation(s)
- Kaoru Takabayashi
- Center for Diagnostic and Therapeutic Endoscopy, Keio University School of Medicine, Tokyo, Japan
| | - Taku Kobayashi
- Center for Advanced IBD Research and Treatment, Kitasato University Kitasato Institute Hospital, Tokyo, Japan
| | - Katsuyoshi Matsuoka
- Division of Gastroenterology and Hepatology, Department of Internal Medicine, Toho University Sakura Medical Center, Chiba, Japan
| | - Barrett G Levesque
- Division of Gastroenterology, Los Angeles County/University of Southern California Medical Center, Los Angeles, USA
| | - Takuji Kawamura
- Department of Gastroenterology, Kyoto Second Red Cross Hospital, Kyoto, Japan
| | - Kiyohito Tanaka
- Department of Gastroenterology, Kyoto Second Red Cross Hospital, Kyoto, Japan
| | - Takeaki Kadota
- Department of Advanced Information Technology, Kyushu University, Fukuoka, Japan
| | - Ryoma Bise
- Research Center for Medical Bigdata, National Institute of Informatics, Tokyo, Japan
- Department of Advanced Information Technology, Kyushu University, Fukuoka, Japan
| | - Seiichi Uchida
- Research Center for Medical Bigdata, National Institute of Informatics, Tokyo, Japan
- Department of Advanced Information Technology, Kyushu University, Fukuoka, Japan
| | - Takanori Kanai
- Division of Gastroenterology and Hepatology, Department of Internal Medicine, Keio University School of Medicine, Tokyo, Japan
| | - Haruhiko Ogata
- Center for Diagnostic and Therapeutic Endoscopy, Keio University School of Medicine, Tokyo, Japan
| |
Collapse
|
3
|
Iboshi Y, Nakamura K, Fukaura K, Iwasa T, Ogino H, Sumida Y, Ihara E, Akiho H, Harada N, Nakamuta M. Increased IL-17A/IL-17F expression ratio represents the key mucosal T helper/regulatory cell-related gene signature paralleling disease activity in ulcerative colitis. J Gastroenterol 2017; 52:315-326. [PMID: 27178567 DOI: 10.1007/s00535-016-1221-1] [Citation(s) in RCA: 22] [Impact Index Per Article: 3.1] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 09/23/2015] [Accepted: 04/26/2016] [Indexed: 02/07/2023]
Abstract
BACKGROUND T helper (Th) and regulatory T (Treg) cell-related cytokines are implicated in inflammatory bowel diseases, including ulcerative colitis (UC). While these cytokines are generally upregulated in inflamed mucosae, the key cytokine profile explaining disease severity has not been determined. METHODS The Rachmilewitz endoscopic index (REI) was assessed in 61 UC patients undergoing colonoscopy. Biopsies obtained from inflamed (REI 3-12) and noninflamed (REI 0-2) areas were analyzed by quantitative PCR for expression of mRNAs encoding cytokines and transcription factors related to Th1 (TNF-α, IFN-γ, IL-12p35, IL-12p40, and T-bet), Th2 (IL-4, IL-13, IL-33, and GATA3), Th17 (IL-17A, IL-17F, IL-21, IL-22, IL-23p19, IL-6, and RORC), Th9 (IL-9, IRF4, and PU.1), and Treg (TGF-β and Foxp3). Expression patterns associated with higher REI were determined by univariate and multivariate analyses. RESULTS Despite general upregulation, none of these mRNAs showed univariate correlation with REI in inflamed samples. Multiple regression analysis, however, found that joint expression of IL-17A, IL-17F, IL-21, RORC, and TGF-β was significantly predictive of REI (P < 0.0002, R2 = 0.380), with major individual contributions by IL-17A (P < 0.0001) and IL-17F (P < 0.0001), which were associated with increased and decreased REI, respectively. Partial correlation analysis, validating this model, indicated differences between IL-17A and IL-17F in correlating with other targets. The IL-17A/IL-17F ratio showed a significant correlation with REI (r = 0.5124, P < 0.0001), whereas no other mRNAs were essentially predictive of REI. CONCLUSIONS Mucosal IL-17A/IL-17F ratio significantly correlates with endoscopic score in UC patients, accompanied by their disparate interactions with other Th/Treg-related genes.
Collapse
Affiliation(s)
- Yoichiro Iboshi
- Department of Medicine and Bioregulatory Science, Graduate School of Medical Sciences, Kyushu University, 3-1-1 Maidashi, Fukuoka, 812-8582, Japan.,Department of Gastroenterology, Kyushu Medical Center, Clinical Research Center, National Hospital Organization, Fukuoka, Japan
| | - Kazuhiko Nakamura
- Department of Medicine and Bioregulatory Science, Graduate School of Medical Sciences, Kyushu University, 3-1-1 Maidashi, Fukuoka, 812-8582, Japan.
| | - Keita Fukaura
- Department of Medicine and Bioregulatory Science, Graduate School of Medical Sciences, Kyushu University, 3-1-1 Maidashi, Fukuoka, 812-8582, Japan
| | - Tsutomu Iwasa
- Department of Medicine and Bioregulatory Science, Graduate School of Medical Sciences, Kyushu University, 3-1-1 Maidashi, Fukuoka, 812-8582, Japan
| | - Haruei Ogino
- Department of Gastroenterology, Kitakyushu Municipal Medical Center, Kitakyushu, Japan
| | - Yorinobu Sumida
- Department of Gastroenterology, Kyushu Medical Center, Clinical Research Center, National Hospital Organization, Fukuoka, Japan
| | - Eikichi Ihara
- Department of Medicine and Bioregulatory Science, Graduate School of Medical Sciences, Kyushu University, 3-1-1 Maidashi, Fukuoka, 812-8582, Japan
| | - Hirotada Akiho
- Department of Gastroenterology, Kitakyushu Municipal Medical Center, Kitakyushu, Japan
| | - Naohiko Harada
- Department of Gastroenterology, Kyushu Medical Center, Clinical Research Center, National Hospital Organization, Fukuoka, Japan
| | - Makoto Nakamuta
- Department of Gastroenterology, Kyushu Medical Center, Clinical Research Center, National Hospital Organization, Fukuoka, Japan
| |
Collapse
|
4
|
|
5
|
Ito H, Iida M, Matsumoto T, Suzuki Y, Aida Y, Yoshida T, Takano Y, Hibi T. Direct comparison of two different mesalamine formulations for the maintenance of remission in patients with ulcerative colitis: a double-blind, randomized study. Inflamm Bowel Dis 2010; 16:1575-82. [PMID: 20049949 PMCID: PMC2972641 DOI: 10.1002/ibd.21194] [Citation(s) in RCA: 20] [Impact Index Per Article: 1.4] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Indexed: 02/01/2023]
Abstract
BACKGROUND Mesalamine has been used as the first-line medication for the treatment of ulcerative colitis (UC). We directly compared the efficacy and safety of two different mesalamine formulations in the maintenance of remission in patients with UC. METHODS In a multicenter, double-blind, randomized study, 131 patients with quiescent UC were assigned to two groups: 65 to receive a pH-dependent release formulation of mesalamine at 2.4 g/day (pH-2.4 g) and 66 to receive a time-dependent release formulation of mesalamine at 2.25 g/day (Time-2.25 g). Both formulations were administered three times daily for 48 weeks. The primary endpoint was the proportion of patients without bloody stools. RESULTS In the full analysis set (n = 130), the proportion of patients without bloody stools was 76.9% in the pH-2.4 g and 69.2% in the Time-2.25 g, demonstrating the noninferiority of pH-2.4 g to Time-2.25 g. No statistically significant difference in time to bloody stools was found between the two formulations (P = 0.27, log-rank test), but the time to bloody stools tended to be longer in pH-2.4 g compared to Time-2.25 g, and a similar trend was observed with regard to the time to relapse. No differences were observed between the safety profiles of the two formulations. CONCLUSIONS The pH- and time-dependent release of mesalamine formulations were similarly safe and effective. Interestingly, the remission phase tended to be longer in the group that received the pH-dependent formulation compared to the group that received the time-dependent formulation (UMIN Clinical Trials Registry, no. C000000289).
Collapse
Affiliation(s)
- Hiroaki Ito
- Digestive Disease Center of Excellence, Kitano Hospital, The Tazuke Kofukai Medical Research InstituteOsaka, Japan
| | - Mitsuo Iida
- Department of Medicine and Clinical Science, Graduate School of Medical Science, Kyushu UniversityFukuoka, Japan
| | - Takayuki Matsumoto
- Division of Lower Gastroenterology, Department of Internal Medicine, Hyogo College of MedicineHyogo, Japan
| | - Yasuo Suzuki
- Department of Internal Medicine, Toho University Sakura Medical CenterChiba, Japan
| | - Yoshiyuki Aida
- Clinical Research, ZERIA Pharmaceutical Co., Ltd.Tokyo, Japan
| | | | - Yuichi Takano
- Clinical Research, ZERIA Pharmaceutical Co., Ltd.Tokyo, Japan
| | - Toshifumi Hibi
- Division of Gastroenterology and Hepatology, Department of Internal Medicine, Keio University School of MedicineTokyo, Japan
| |
Collapse
|
6
|
Ito H, Iida M, Matsumoto T, Suzuki Y, Sasaki H, Yoshida T, Takano Y, Hibi T. Direct comparison of two different mesalamine formulations for the induction of remission in patients with ulcerative colitis: a double-blind, randomized study. Inflamm Bowel Dis 2010; 16:1567-74. [PMID: 20049950 PMCID: PMC2972638 DOI: 10.1002/ibd.21193] [Citation(s) in RCA: 44] [Impact Index Per Article: 3.1] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Indexed: 12/11/2022]
Abstract
BACKGROUND Mesalamine is the first-line drug for the treatment of ulcerative colitis (UC). We directly compared the efficacy and safety of two mesalamine formulations for the induction of remission in patients with UC. METHODS In a multicenter, double-blind, randomized study, 229 patients with mild-to-moderate active UC were assigned to 4 groups: 66 and 65 received a pH-dependent release formulation of 2.4 g/day (pH-2.4 g) or 3.6 g/day (pH-3.6 g), respectively; 65 received a time-dependent release formulation of 2.25 g/day (Time-2.25 g), and 33 received placebo (Placebo). The drugs were administered three times daily for eight weeks. The primary endpoint was a decrease in the UC disease activity index (UC-DAI). RESULTS In the full analysis set (n = 225) the decrease in UC-DAI in each group was 1.5 in pH-2.4 g, 2.9 in pH-3.6 g, 1.3 in Time-2.25 g and 0.3 in Placebo, respectively. These results demonstrate the superiority of pH-3.6 g over Time-2.25 g (P = 0.003) and the noninferiority of pH-2.4 g to Time-2.25 g. Among the patients with proctitis-type UC, a significant decrease in UC-DAI was observed in pH-2.4 g and pH-3.6 g as compared to Placebo, but not in Time-2.25 g. No differences were observed in the safety profiles. CONCLUSIONS Higher dose of the pH-dependent release formulation was more effective for induction of remission in patients with mild-to-moderate active UC. Additionally, the pH-dependent release formulation was preferable to the time-dependent release formulation for patients with proctitis-type UC (UMIN Clinical Trials Registry, no. C000000288).
Collapse
Affiliation(s)
- Hiroaki Ito
- Digestive Disease Center of Excellence, Kitano Hospital, The Tazuke Kofukai Medical Research InstituteOsaka, Japan
| | - Mitsuo Iida
- Department of Medicine and Clinical Science, Graduate School of Medical Science, Kyushu UniversityFukuoka, Japan
| | - Takayuki Matsumoto
- Division of Lower Gastroenterology, Department of Internal Medicine, Hyogo College of MedicineHyogo, Japan
| | - Yasuo Suzuki
- Department of Internal Medicine, Toho University Sakura Medical CenterChiba, Japan
| | - Hidetaka Sasaki
- Clinical Research, ZERIA Pharmaceutical Co., Ltd.Tokyo, Japan
| | | | - Yuichi Takano
- Clinical Research, ZERIA Pharmaceutical Co., Ltd.Tokyo, Japan
| | - Toshifumi Hibi
- Division of Gastroenterology and Hepatology, Department of Internal Medicine, Keio University School of MedicineTokyo, Japan
| |
Collapse
|
7
|
Kopelman Y, Siersema PD, Nir Y, Szold A, Bapaye A, Segol O, Willenz EP, Lelcuk S, Geller A, Kopelman D. Endoluminal compression clip: full-thickness resection of the mesenteric bowel wall in a porcine model. Gastrointest Endosc 2009; 70:1146-57. [PMID: 19647245 DOI: 10.1016/j.gie.2009.05.015] [Citation(s) in RCA: 11] [Impact Index Per Article: 0.7] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 12/18/2008] [Accepted: 05/01/2009] [Indexed: 01/12/2023]
Abstract
BACKGROUND Performing a full-thickness intestinal wall resection of a sessile polyp located on the mesenteric side with a compression clip may lead to compression of mesenteric vessels. The application of such a clip may therefore cause a compromised blood supply in the particular bowel segment, leading to perforation. OBJECTIVE To evaluate the performance of a newly developed, nitinol compression clip, called the NiTi clamp, for full-thickness resection of the bowel wall, while the clip is deliberately deployed endoluminally on the mesenteric side. DESIGN Prospective animal study. Multinational, multidisciplinary; gastroenterology and general surgery, research cooperation. SETTING Animal research laboratory. INTERVENTION Six pigs were operated upon and endoscopically evaluated and then killed after 3 weeks. Linear compression closure clips based on nitinol springs were used. Three longitudinal enterotomies were performed: in the cecum, spiral colon, and proximal rectum. Four clips were deployed in each animal. MAIN OUTCOME MEASUREMENTS A total of 23 clips were deployed. The average expulsion day was 9 days. RESULTS All but 3 clips were normally expelled. One pig developed bowel ischemia due to intussusception. In endoscopic procedures, no signs of significant segmental mucosal ischemia were found. The macroscopic appearance of the compression closure lines was thin and delicate, but epithelialization was significantly delayed at 5 sites. LIMITATION Differences between porcine and human colorectal anatomy. CONCLUSION Full-thickness clamping of the bowel with the NiTi clamp, including the local mesenteric vasculature, does not significantly impair local healing of the clamp site and gives hope to further development of novel full-thickness endoscopic resection technologies.
Collapse
Affiliation(s)
- Yael Kopelman
- Gastroenterology Institute, Israel Institute of Technology, Haifa, Israel.
| | | | | | | | | | | | | | | | | | | |
Collapse
|