KNKoichi Nishitsuka

Koichi Nishitsuka

*This is an algorithmically generated author record based on publicly available information.

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


Koichi Nishitsuka has a total of 105 co-authors. Here are the top co-authors by number of shared publications:

  • Hidetoshi Yamashita

    Department of Ophthalmology and Visual Sciences, Yamagata University Faculty of Medicine, Yamagata, Japan

    26 shared publications
  • Hiroyuki Namba

    Department of Ophthalmology and Visual Sciences, Yamagata University Faculty of Medicine, Yamagata, Japan

    8 shared publications
  • Katsuhiro Nishi

    Department of Ophthalmology and Visual Sciences, Yamagata University Faculty of Medicine, Yamagata, Japan

    8 shared publications
  • Teiko Yamamoto

    Affiliation Department of Ophthalmology and Visual Sciences, Yamagata University Faculty of Medicine, Yamagata, Yamagata, Japan

    7 shared publications
  • Katsuhiro Nishi

    Department of Ophthalmology and Visual Sciences, Yamagata University Faculty of Medicine

    6 shared publications
  • Yutaka Kaneko

    Department of Ophthalmology and Visual Sciences, Yamagata University Faculty of Medicine, Yamagata, Japan

    6 shared publications

Latest Publications

  • Open Access

    Explainable Deep Learning for Lesion-Level Detection of Diabetic Retinopathy: A Segmentation Approach Using Fundus Images Graded as Mild-to-Moderate Nonproliferative Diabetic Retinopathy

    by Takumi Sato, Koichi Nishitsuka, Tohru Itoh, Toshihiro Okashita, Satoshi Wada, Atsushi Shinjo

    Deep learning has shown promise in diabetic retinopathy screening using fundus images. However, many existing models operate as “black boxes,” providing limited interpretability at the lesion level. This study aimed to develop an explainable deep learning model capable of detecting four diabetic retinopathy-related lesions—hemorrhages, hard exudates, cotton wool spots, and microaneurysms—and evaluate its performance using both conventional per-lesion metrics and a novel syntactic agreement framework. A total of 1,087 fundus images were obtained from publicly available datasets (EyePACS and APTOS), which contained 585 images graded as mild-to-moderate nonproliferative diabetic retinopathy (DR1 or DR2). All images were manually annotated for the presence of the four lesions. A U-Net–based segmentation model was trained to generate binary predictions for each lesion type. The performance of the model was evaluated using sensitivity, specificity, precision, and F1 score, along with five syntactic agreement criteria that evaluated the lesion-set consistency between the predicted and ground truth outputs at the image level. The model achieved high sensitivity and F1 scores for hemorrhages and hard exudates, showed moderate performance for cotton wool spots, and failed to detect any microaneurysms (0% sensitivity), with 92.9% of the microaneurysms cases misclassified as hemorrhages. Despite this limitation, the image-level agreement remained high, with any-lesion match and hemorrhage match rates exceeding 95%. These findings suggest that although individual lesion classification was imperfect, the model effectively recognized abnormal images, highlighting its potential as a screening tool. The proposed syntactic agreement framework offers a complementary evaluation strategy that aligns more closely with clinical interpretation and may help bridge the gap between artificial intelligence–based predictions and real-world ophthalmic decision-making.

  • Open Access

    Multi-task deep learning for predicting metabolic syndrome from retinal fundus images in a Japanese health checkup dataset

    by Tohru Itoh, Koichi Nishitsuka, Yasufumi Fukuma, Satoshi Wada

    Background: Retinal fundus images provide a noninvasive window into systemic health, offering opportunities for early detection of metabolic disorders such as metabolic syndrome (METS). Objective: This study aimed to develop a deep learning model to predict METS from fundus images obtained during routine health checkups, leveraging a multi-task learning approach. Methods: We retrospectively analyzed 5,000 fundus images from Japanese health checkup participants. Convolutional neural network (CNN) models were trained to classify METS status, incorporating fundus-specific data augmentation strategies and auxiliary regression tasks targeting clinical parameters such as abdominal circumference (AC). Model performance was evaluated using validation accuracy, test accuracy, and the area under the receiver operating characteristic curve (AUC). Results: Models employing fundus-specific augmentation demonstrated more stable convergence and superior validation accuracy compared to general-purpose augmentation. Incorporating AC as an auxiliary task further enhanced performance across architectures. The final ensemble model with test-time augmentation achieved a test accuracy of 0.696 and an AUC of 0.73178. Conclusion: Combining multi-task learning, fundus-specific data augmentation, and ensemble prediction substantially improves deep learning-based METS classification from fundus images. This approach may offer a practical, noninvasive screening tool for metabolic syndrome in general health checkup settings.


  • Open Access

    Multi-Task Deep Learning for Predicting Metabolic Syndrome from Retinal Fundus Images in a Japanese Health Checkup Dataset

    by Tohru Itoh, Koichi Nishitsuka, Yasufumi Fukuma, Satoshi Wada

    Background: Retinal fundus images provide a noninvasive window into systemic health, offering opportunities for early detection of metabolic disorders such as metabolic syndrome (METS). Objective: This study aimed to develop a deep learning model to predict METS from fundus images obtained during routine health checkups, leveraging a multi-task learning approach. Methods: We retrospectively analyzed 5,000 fundus images from Japanese health checkup participants. Convolutional neural network (CNN) models were trained to classify METS status, incorporating fundus-specific data augmentation strategies and auxiliary regression tasks targeting clinical parameters such as abdominal circumference (AC). Model performance was evaluated using validation accuracy, test accuracy, and the area under the receiver operating characteristic curve (AUC). Results: Models employing fundus-specific augmentation demonstrated more stable convergence and superior validation accuracy compared to general-purpose augmentation. Incorporating AC as an auxiliary task further enhanced performance across architectures. The final ensemble model with test-time augmentation achieved a test accuracy of 0.696 and an AUC of 0.73178. Conclusion: Combining multi-task learning, fundus-specific data augmentation, and ensemble prediction substantially improves deep learning-based METS classification from fundus images. This approach may offer a practical, noninvasive screening tool for metabolic syndrome in general health checkup settings.

  • Open Access

    Differences in Ocular Biometry Between Short-Axial and Normal-Axial Eyes in the Elderly Japanese

    by Takahiro Tanaka, Koichi Nishitsuka, Hiroto Obata

    Purpose: Among patients with angle-closure glaucoma, it is common to have a short-axial eye, which also makes it difficult to select an appropriate intraocular lens. Previous studies have focused on the ocular biometry of the long-axial eye, whereas only a few reports have focused on the short-axial eye. This study aimed to clarify the characteristics of the short-axial eye on ocular biometry among the elderly Japanese. Patients and Methods: This retrospective cross-sectional study included 142 patients who consulted at our hospital. The ocular axial length (AL), central corneal thickness (CCT), average anterior corneal radius of curvature (CR), white-to-white (WTW), anterior chamber depth (ACD), and lens thickness (LT) were measured using OA-2000. The short-axial and normal-axial groups (short-axial [SA] and normal-axial [NA] groups) included those with an AL less than 22.5 mm and ranging from 22.5– 26.5 mm, respectively. Comparisons between groups were analyzed using the Wilcoxon rank sum test. Spearman’s rank correlation coefficient tests and multiple regression analyses were performed for all parameters overall and per group. Results: The parameters that were significantly different between the SA and NA groups were CR, WTW, and ACD (all P < 0.0001). In the SA group (42 patients), only ACD was significantly correlated with AL (r = 0.33, P = 0.031). On multiple regression analysis for the SA group, none of the parameters had a significant effect on AL (all P > 0.05). Conclusion: In the SA group, ACD was the only factor that positively correlated with AL; it was significantly shallower than in the NA group. Thus, in the SA eye, it is clinically important to be aware of ACD.


  • Open Access

    Correlation of Ocular Biometry with Axial Length in Elderly Japanese

    by Takahiro Tanaka, Koichi Nishitsuka, Hiroto Obata

    Purpose: Ocular parameters are not only useful for diagnosing diseases but also for guiding treatment approaches. A lot of previous studies have reported ocular parameters and its relations before cataract surgery. However, despite ethnic differences in ocular biometry, few reports have dealt with Japanese. Hence, this retrospective cross-sectional study aimed to measure parameters of preoperative cataract patients and examines the correlations between each parameter in Japanese elderly people. Patients and Methods: The 210 subjects had their ocular axial lengths measured with OA-2000. The endpoints were ocular axial length (AL), central corneal thickness (CCT), average anterior corneal radius of curvature (CR), white-to-white (WTW), anterior chamber depth (ACD), and lens thickness (LT). Our analyses utilized the eye with the longer AL in each person. Each parameter was analyzed for correlations in a round-robin manner. Regression analyses were performed on parameters correlated with AL. Results: The parameters correlated with AL were CR (r = 0.33, P < 0.0001), WTW (r = 0.29, P < 0.0001), ACD (r = 0.59, P < 0.0001), and LT (r = − 0.30, P < 0.0001). These parameters related to AL in all simple regression equations (CR (P < 0.0001), WTW (P = 0.0002), ACD (P < 0.0001), LT (P = 0.0001)). In multiple regression analyses, CR, ACD, and LT might relate to AL (CR (P = 0.0002), ACD (P < 0.0001), LT (P = 0.018)). LT tended to be thinner as AL increased, while CR, WTW, and ACD tended to increase. Conclusion: This information may be useful in developing strategies for ophthalmic surgery, as it provides information on the location of intraocular tissues. Various parameters have been used in intraocular lens (IOL) power calculations in recent years and knowledge of the interrelationship among parameters may be useful in determining IOL power according to ethnicity in the future.


  • Open Access

    Ocular biometric differences between long and not long-axial length eyes

    by Takahiro Tanaka, Ayako Marubayashi, Koichi Nishitsuka, Hiroto Obata

    This retrospective cross-sectional study aimed to measure parameters of preoperative cataract patients and examines the correlations between each parameter. The 210 subjects had their ocular axial lengths measured with OA-2000. The endpoints were ocular axial length (AL), central corneal thickness (CCT), average corneal radii (CR), white-to-white(WTW), anterior chamber depth(ACD), and lens thickness(LT). AL was mainly employed for each person. Each parameter was analyzed for correlations in a round-robin manner. Regression analyses were performed on parameters correlated with AL. The parameters correlated with AL were CR (r = 0.33, p < 0.0001), WTW (r = 0.29, p < 0.0001), ACD (r = 0.59, p < 0.0001), and LT (r = −0.30, p < 0.0001). These parameters affected AL in all single regression equations (CR (P < 0.0001), WTW (P = 0.0002), ACD (P < 0.0001), LT (P = 0.0001)). In multiple regression analyses, CR, ACD, and LT might affect AL (CR (P = 0.0002), ACD (P < 0.0001), LT (P = 0.018)). LT tended to be thinner as AL increased, while CR, WTW, and ACD tended to increase. This information will be useful in developing strategies for ophthalmic surgery.

  • Open Access

    Efficacy of vitrectomy with air tamponade for rhegmatogenous retinal detachment: a prospective study

    by Katsuhiro Nishi, Madoka Nakamura, Koichi Nishitsuka

    Rhegmatogenous retinal detachment (RRD) causes a permanent decrease in visual acuity and visual field. During pars plana vitrectomy (PPV) for RRD, long acting gas have been selected for tamponade because gas stays in the eye for a long time. Recently, several studies have showed the efficacy of air tamponade for RRD treatment. Few prospective studies have analyzed the efficacy of air tamponade. We registered 194 eyes from 190 patients who consented to a prospective study of PPV with air tamponade for RRD by a single surgeon from June 2019 to November 2022. These patients were all treated with air tamponade without silicone oil and were followed for > 3 months postoperatively. Primary success rates were 97.9% (190/194) in total cases, with no discernible difference between the uncomplicated (100%:87/87) and complicated (96.3%: 103/107) RRD groups (P = 0.13). There was no considerable difference in primary success rate between upper break (97.9%:143/146) and lower break cases (97.9%:47/48). Proliferative vitreoretinopathy (PVR) grade C was associated with initial failure by multivariate analysis (P = 0.00003). Air tamponade has a sufficient therapeutic effect in cases of RRD less than PVR grade C, regardless of the location of the retinal tear.


  • Open Access

    Efficacy of Vitrectomy with Air Tamponade for Rhegmatogenous Retinal Detachment: A Prospective Study

    by Katsuhiro Nishi, Madoka Nakamura, Koichi Nishitsuka

    Rhegmatogenous retinal detachment (RRD) causes a permanent decrease in visual acuity and visual field. During pars plana vitrectomy (PPV) for RRD, long acting gas have been selected for tamponade because gas stays in the eye for a long time. Recently, several studies have showed the efficacy of air tamponade for RRD treatment. Few prospective studies have analyzed the efficacy of air tamponade. We registered 194 eyes from 190 patients who consented to a prospective study of PPV with air tamponade for RRD by a single surgeon from June 2019 to November 2022. These patients were all treated with air tamponade without silicone oil and were followed for > 3 months postoperatively. Primary success rates were 97.9% (190/194) in total cases, with no discernible difference between the uncomplicated (100%:87/87) and complicated (96.3%: 103/107) RRD groups (P = 0.13). There was no considerable difference in primary success rate between upper break (97.9%:143/146) and lower break cases (97.9%:47/48). Proliferative vitreoretinopathy (PVR) grade C was associated with initial failure by multivariate analysis (P = 0.00003). Air tamponade has a sufficient therapeutic effect in cases of RRD less than PVR grade C, regardless of the location of the retinal tear.

  • Effectiveness of intraoperative optical coherence tomography on vitrectomy for proliferative diabetic retinopathy

    by Koichi Nishitsuka, Katsuhiro Nishi, Hidetoshi Yamashita

    Purpose We sought to verify the effect of intraoperative optical coherence tomography (iOCT) on vitrectomy for the treatment of proliferative diabetic retinopathy (PDR). Study design Retrospective study. Methods We reviewed 178 eyes with PDR treated with 25-gauge vitrectomy by a single surgeon between April 2013 and December 2017. In total, 98 eyes of 77 patients with PDR (mean age, 52.5 ± 11.7 years) treated with vitrectomy using iOCT (iOCT group) and 80 eyes of 60 patients with PDR (mean age, 53.1 ± 10.9 years) treated without iOCT (non-iOCT group) were included in the study. To determine the effects of combining vitrectomy with iOCT, a comparison was made of patients treated with this combination and patients treated with vitrectomy without iOCT. The effects of these treatments were assessed on the basis of intraoperative complications, reoperation ratios, postoperative complications, operation times, and postoperative visual acuity. Results The operation time was significantly reduced for the iOCT group (72.9 ± 23.9 min) when compared with the non-iOCT group (91.3 ± 31.2 min) (P = .001). The incidences of intraoperative complications, reoperation, and postoperative complications did not differ significantly between the 2 groups (P = .542, 0.258, and 0.860 respectively). Six months after surgery, the postoperative visual acuity did not differ significantly between the 2 groups (P = .508). Multiple linear regression analysis revealed that the operation time was significantly correlated with iOCT (beta [standard partial regression coefficient] = − 0.28, P < .001), the fibrovascular proliferative membrane (beta = 0.17, P = .009), cataract surgery (beta = 0.22, P = .016), preoperative retinal photocoagulation (beta = − 0.14, P = .021), intraoperative complications (beta = 0.16, P = .023), and posterior vitreous detachment (beta = − 0.14, P = .04). Conclusion Use of iOCT reduced the operation time without affecting the incidence rates of intraoperative and postoperative complications, reoperation ratios, or postoperative visual acuities in patients who underwent vitreous surgery for PDR.


  • Open Access

    Utility of Liquid-Based Cytology and Cell Block Procedure Obtained by Vitrectomy to Diagnose Ocular Sarcoidosis–The Significance of Epithelioid Granuloma and Epithelioid Cells

    by Rintaro Ohe, Yutaka Kaneko, Hiroyuki Namba, Katsuhiro Nishi, Jun-Ichi Goto, Mitsuru Futakuchi, Koichi Nishitsuka

    Purpose: The eyes are one of the most frequently involved organs in sarcoidosis in Asia, including Japan. Sarcoid uveitis is the major complaint of ocular sarcoidosis. The detection of epithelioid granuloma (EG) requires histological biopsy of the uvea for the precise diagnosis of sarcoid uveitis, because it is challenging to diagnose sarcoid uveitis without a history of systemic sarcoidosis. To diagnose sarcoid uveitis, we have established novel methods. Patients and Methods: In this study, we included 30 eyes of 21 patients with granulomatous uveitis diagnosed via slit-lamp examinations, gonioscopy, fundus photography, and fluorescein angiography. Vitrectomy was performed to remove the vitreous opacity with vision loss. To examine vitreous cell components, we used liquid-based cytology (LBC). To detect EG in an intraocular irrigating solution, we collected vitreous cell components, and then the cell pellets were embedded in the cell block procedure. Results: Here, we demonstrated the usefulness of the histological detection of EG and epithelioid cells (ECs) in LBC from vitreous body specimens and in the cell block procedure from vitreous cell components in an intraocular irrigating solution. Our results showed that the detection rates of EG were 6.3% (1/16) in LBC and 9.1% (1/11) in the cell block procedure in the sarcoid uveitis-suspected group and 7.7% (1/13) in LBC and 28.6% (2/7) in the cell block procedure in the sarcoidosis group. We would discuss the specificity of the EG/EC detection of ocular sarcoidosis. Conclusion: Our methods are helpful in the precise diagnosis of ocular sarcoidosis and the control of the development of systemic sarcoidosis.