A Deep Learning-Based Assessment Pipeline for Intraepithelial and Stromal Tumor-Infiltrating Lymphocytes in High-Grade Serous Ovarian Carcinoma.

Tumor-infiltrating lymphocytes (TILs) are associated with improved survival in patients with epithelial ovarian cancer. However, TIL evaluation has not been used in routine clinical practice because of reproducibility issues. The current study developed two convolutional neural network models to detect TILs and to determine their spatial location in whole slide images, and established a spatial assessment pipeline to objectively quantify intraepithelial and stromal TILs in patients with high-grade serous ovarian carcinoma. The predictions of the established models showed a significant positive correlation with the number of CD8+ T cells and immune gene expressions. Patients with a higher density of intraepithelial TILs had a significantly prolonged overall survival and progression-free survival in multiple cohorts. On the basis of the density of intraepithelial and stromal TILs, patients were classified into three immunophenotypes: immune inflamed, excluded, and desert. The immune-desert subgroup showed the worst prognosis. Gene expression analysis showed that the immune-desert subgroup had lower immune cytolytic activity and T-cell-inflamed gene-expression profile scores, whereas the immune-excluded subgroup had higher expression of interferon-γ and programmed death 1 receptor signaling pathway. The established evaluation method provided detailed and comprehensive quantification of intraepithelial and stromal TILs throughout hematoxylin and eosin-stained slides. It has potential for clinical application for personalized treatment of patients with ovarian cancer.

Measurement of the Acetabular Cup Orientation after Total Hip Arthroplasty Based on Three-Dimensional Reconstruction from a Single X-ray Image Using Generative Adversarial Networks.

BACKGROUND: The purpose of this study was to reconstruct three-dimensional (3D) computed tomography (CT) images from single anteroposterior (AP) postoperative total hip arthroplasty (THA) X-ray images using a deep learning algorithm known as generative adversarial networks (GANs) and to validate the accuracy of cup angle measurement on GAN-generated CT. METHODS: We used two GAN-based models, CycleGAN and X2CT-GAN, to generate 3D CT images from X-ray images of 386 patients who underwent primary THAs using a cementless cup. The training dataset consisted of 522 CT images and 2,282 X-ray images. The image quality was validated using the peak signal-to-noise ratio (PSNR) and the structural similarity index measure (SSIM). The cup anteversion and inclination measurements on the GAN-generated CT images were compared with the actual CT measurements. Statistical analyses of absolute measurement errors were performed using Mann-Whitney U tests and nonlinear regression analyses. RESULTS: The study successfully achieved 3D reconstruction from single AP postoperative THA X-ray images using GANs, exhibiting excellent PSNR (37.40) and SSIM (0.74). The median absolute difference in radiographic anteversion (RA) was 3.45° and the median absolute difference in radiographic inclination (RI) was 3.25°, respectively. Absolute measurement errors tended to be larger in cases with cup malposition than in those with optimal cup orientation. CONCLUSION: This study demonstrates the potential of GANs for 3D reconstruction from single AP postoperative THA X-ray images to evaluate cup orientation. Further investigation and refinement of this model are required to improve its performance.

[An Intraductal Papillary Mucinous Neoplasm of the Pancreas with Metachronous Quadruple Cancers].

The patient described herein was diagnosed with left breast, endometrial, and early gastric cancers at 49, 53, and 57 years of age, respectively. Magnetic resonance cholangiopancreatography performed when she was undergoing treatment for cholecystitis at 50 years of age showed local pancreatic duct dilatation in the pancreatic head. She was followed in the Department of Gastroenterology at our hospital for an intraductal papillary mucinous neoplasm(IPMN). An abdominal computed tomography scan obtained at 59 years of age revealed dilation of the main pancreatic duct in the pancreas body and tail, therefore an endoscopic ultrasound-guided fine needle aspiration was performed. She was diagnosed with pancreatic cancer and underwent a laparoscopic distal pancreatectomy. The postoperative course was uneventful; however, the pancreatic cancer recurred and she died approximately 14 months postoperatively. Reports of multiple cancers associated with IPMNs are rare, yet we managed a patient with a pancreatic head IPMN complicated by metachronous quadruple carcinomas( breast, endometrial, gastric, and pancreatic cancers).

Human papillomavirus vaccine effectiveness by age at first vaccination among Japanese women.

In Japan, the National Immunization Program against human papillomavirus (HPV) targets girls aged 12-16 years, and catch-up vaccination is recommended for young women up to age 26 years. Because HPV infection rates increase soon after sexual debut, we evaluated HPV vaccine effectiveness by age at first vaccination. Along with vaccination history, HPV genotyping results from 5795 women younger than 40 years diagnosed with cervical intraepithelial neoplasia grade 2-3 (CIN2-3), adenocarcinoma in situ (AIS), or invasive cervical cancer were analyzed. The attribution of vaccine-targeted types HPV16 or HPV18 to CIN2-3/AIS was 47.0% for unvaccinated women (n = 4297), but 0.0%, 13.0%, 35.7%, and 39.6% for women vaccinated at ages 12-15 years (n = 36), 16-18 years (n = 23), 19-22 years (n = 14), and older than 22 years (n = 91), respectively, indicating the greater effectiveness of HPV vaccination among those initiating vaccination at age 18 years or younger (P < .001). This finding was supported by age at first sexual intercourse; among women with CIN2-3/AIS, only 9.2% were sexually active by age 14 years, but the percentage quickly increased to 47.2% by age 16 and 77.1% by age 18. Additionally, the HPV16/18 prevalence in CIN2-3/AIS was 0.0%, 12.5%, and 40.0% for women vaccinated before (n = 16), within 3 years (n = 8), and more than 3 years after (n = 15) first intercourse, respectively (P = .004). In conclusion, our data appear to support routine HPV vaccination for girls aged 12-14 years and catch-up vaccination for adolescents aged 18 years and younger in Japan.

Tertiary lymphoid structures are associated with favorable survival outcomes in patients with endometrial cancer.

Immunotherapy has experienced remarkable growth recently. Tertiary lymphoid structures (TLSs) and B cells may play a key role in the immune response and have a survival benefit in some solid tumors, but there have been no reports about their role in endometrial cancer (EC). We investigated the clinicopathological and pathobiological characteristics of the tumor microenvironment (TME) in EC. Patients with EC at Kyoto University Hospital during 2006-2011 were retrospectively included. In 104 patients with EC who met study inclusion criteria, 81 (77.9%) had TLSs, which consisted of areas rich in CD20+ B cells, CD8+ T cells, CD4+ T cells, and CD38+ plasma cells. The absence of TLS was independently associated with tumor progression (HR, 0.154; 95% CI, 0.044-0.536; P = 0.003). Patients with TLSs that included CD23+ germinal centers had better PFS. All tumor infiltrating lymphocytes were counted in the intratumor site. The number of CD20+ B cells was significantly larger in patients with TLSs than in those without TLS (P < 0.001). CD20+ B cells numbers were positively correlated with other TLSs. The larger number of CD20+ B cell was associated with better PFS (P = 0.015). TLSs and B cell infiltration into tumors are associated with favorable survival outcomes in patients with EC. They may represent an active immune reaction of the TME in endometrial cancer.

Development of healthy lifestyle consciousness index for gynecological cancer patients.

PURPOSE: Healthy lifestyle is related to quality of life (QOL) after cancer diagnosis and prognosis. However, there are few reports on patients conscious of healthy lifestyle and patients requiring medical providers' attention regarding healthy lifestyle. We aimed to develop a healthy lifestyle consciousness index (HLCI) for cancer patients and evaluated its validity in gynecological cancer patients. METHODS: The HLCI was designed to assess degree of healthy lifestyle consciousness, including items regarding "diet," "exercise," "body weight," and "sleep." Exploratory factor analysis was performed for dimensionality of the scale; Cronbach's alpha was calculated to assess internal-consistency reliability. For criterion-based validity, we calculated proportions of stage III/IV gynecological malignancies in those with categorized HLCI scores based on tertiles. Concurrent validity was evaluated between HLCI and other quality of life (QOL) scales including European Organization for Research and Treatment of Cancer QLQ-C30 in limited patients. RESULTS: HLCI comprised five 10-point items (0-45); higher values implied improved healthy lifestyle consciousness. Data from 108 gynecological malignancy patients at Kyoto University Hospital were analyzed. The mean age of subjects was 55.8 years; 36.1% of them had uterine corpus cancer; 34.3% were at stage III/IV of gynecological malignancy. The factor analysis revealed HLCI was unidimensional; the reliability based on Cronbach's alpha was satisfactory (0.88). The proportions of stage III/IV gynecological malignancies were 25.7%, 33.3%, and 44.4% in those with first (7-24 points), second (25-30 points), and third (31-46 points) tertiles of HLCI score, respectively. For patients with other QOL scales (n = 25), the mean scores of global health status of QLQ-C30 were 33.3, 50.0, and 83.3 for first, second, and third tertiles of HLCI score, respectively. CONCLUSION: HLCI was successfully validated; thus, patients with advanced stages or higher QOL might have strong consciousness regarding healthy lifestyle. HLCI may be useful in precision care for improved lifestyles and QOL.

Impact of efforts to prevent maternal deaths due to obstetric hemorrhage on trends in epidemiology and management of severe postpartum hemorrhage in Japan: a nationwide retrospective study.

BACKGROUND: The Japan Society of Obstetrics and Gynecology and the Japan Association of Obstetricians and Gynecologists have issued the guidelines and recommendations on postpartum hemorrhage since 2010 and have been conducted widespread educational activities from 2012. The aim of this study was to investigate the impact of these efforts by the Societies to prevent maternal deaths due to obstetric hemorrhage on trends in epidemiology and management of severe postpartum hemorrhage in Japan. METHODS: A national retrospective cohort study was conducted using the national database of health insurance claims for the period 2012 and 2018. The subjects were all insured women who received a blood transfusion for postpartum hemorrhage. The primary endpoints of this study were hysterectomy and maternal mortality. The etiology of hemorrhage, treatment facility, type of procedure, and blood transfusion volume were tabulated. RESULTS: Women with postpartum hemorrhage that underwent transfusion increased from 3.5 to 5.5 per 1000 deliveries between 2012 and 2018. The most common cause of postpartum hemorrhage was atonic hemorrhage. After insurance coverage in 2013, the intrauterine balloon tamponade use increased to 20.3% of postpartum hemorrhages treated with transfusion in 2018, while the proportion of hysterectomy was decreased from 7.6% (2013-2015) to 6.4% (2016-2018) (p < 0.0001). The proportion of postpartum hemorrhage in maternal deaths decreased from 21.1% (2013-2015) to 14.1% (2016-2018) per all maternal deaths cases (p = 0.14). Cases with postpartum hemorrhage managed in large referral hospitals was increased (65.9% in 2012 to 70.4% in 2018) during the study period (p < 0.0001). CONCLUSIONS: The efforts by the Societies to prevent maternal mortality due to obstetric hemorrhage resulted in a significant decrease in the frequency of hysterectomies and a downward trend in maternal mortality due to obstetric hemorrhage.

Association between non-anemic iron deficiency in early pregnancy and perinatal mental health: A retrospective pilot study.

AIM: Postpartum anemia and iron deficiency are reportedly involved in postpartum depression, but the association between perinatal depression and iron deficiency with or without anemia is poorly documented. This pilot study retrospectively investigated the relationship between non-anemic iron deficiency (NAID) in early pregnancy and perinatal depressive symptoms. METHODS: This study included 31 non-anemic women among patients who received perinatal care with preserved residual serum from routine antenatal checkups in Kyoto University Hospital. All participants gave informed consent for research. The ferritin concentration in their preserved serum was measured. The hemoglobin (Hb) and ferritin in early pregnancy, as well as the Edinburgh Postpartum Depression Scale (EPDS) at mid-pregnancy and 1 month after childbirth were analyzed. Iron deficiency was defined as a serum ferritin concentration < 30 ng/mL. RESULTS: Based on the ferritin level in early pregnancy, 13 women (41.9%) had NAID, whereas 18 were normal. The mean Hb and ferritin were 12.7 ± 0.6 g/dL and 18.5 ± 5.8 ng/mL in the NAID group and 12.8 ± 0.9 g/dL and 74.7 ± 39.2 ng/mL in the normal group, respectively. The median EPDS scores at mid-pregnancy and 1 month postpartum, respectively, were 2.0 (2.0-3.3) and 5.0 (4.0-6.6) in the NAID group and 4.5 (2.3-7.3) and 4.5 (2.3-5.7) in the normal group. EPDS scores increased significantly from mid-pregnancy to 1 month postpartum in the NAID group only. CONCLUSION: NAID in early pregnancy was highly prevalent and was suggested to reduce resilience to depression during the perinatal period.

Report from "International Workshop for Junior Fellows 2: Tocolytic treatment for prevention of preterm birth" at the 73rd Annual Congress of the Japan Society of Obstetrics and Gynecology.

At the 73rd Annual Congress of the Japan Society of Obstetrics and Gynecology, we discussed recent tocolytic treatments for the prevention of preterm birth with obstetricians from Korea and Taiwan. The rate of preterm birth in our countries has been increasing, and optimal tocolytic treatments are of great concern. Ritodrine hydrochloride was the first available drug for tocolysis in our countries and is often administered for over 48 h, although it is not recommended for maintenance therapy in Western countries. Meanwhile, there are differences in the use of other tocolytic drugs, based on approval of the drugs and the health insurance systems. Nifedipine and atosiban have not been considered first-line agents in Japan. The long-term use of unnecessary tocolysis should be avoided, and the introduction of other tocolytic drugs, including nifedipine or atosiban, should be discussed.

A cell-based high-throughput screening method to directly examine transthyretin amyloid fibril formation at neutral pH.

Transthyretin (TTR) is a major amyloidogenic protein associated with hereditary (ATTRm) and nonhereditary (ATTRwt) intractable systemic transthyretin amyloidosis. The pathological mechanisms of ATTR-associated amyloid fibril formation are incompletely understood, and there is a need for identifying compounds that target ATTR. C-terminal TTR fragments are often present in amyloid-laden tissues of most patients with ATTR amyloidosis, and on the basis of in vitro studies, these fragments have been proposed to play important roles in amyloid formation. Here, we found that experimentally-formed aggregates of full-length TTR are cleaved into C-terminal fragments, which were also identified in patients' amyloid-laden tissues and in SH-SY5Y neuronal and U87MG glial cells. We observed that a 5-kDa C-terminal fragment of TTR, TTR81-127, is highly amyloidogenic in vitro, even at neutral pH. This fragment formed amyloid deposits and induced apoptosis and inflammatory gene expression also in cultured cells. Using the highly amyloidogenic TTR81-127 fragment, we developed a cell-based high-throughput screening method to discover compounds that disrupt TTR amyloid fibrils. Screening a library of 1280 off-patent drugs, we identified two candidate repositioning drugs, pyrvinium pamoate and apomorphine hydrochloride. Both drugs disrupted patient-derived TTR amyloid fibrils ex vivo, and pyrvinium pamoate also stabilized the tetrameric structure of TTR ex vivo in patient plasma. We conclude that our TTR81-127-based screening method is very useful for discovering therapeutic drugs that directly disrupt amyloid fibrils. We propose that repositioning pyrvinium pamoate and apomorphine hydrochloride as TTR amyloid-disrupting agents may enable evaluation of their clinical utility for managing ATTR amyloidosis.

A novel age-related venous amyloidosis derived from EGF-containing fibulin-like extracellular matrix protein 1.

Most intractable tissue-degenerative disorders share a common pathogenic condition, so-called proteinopathy. Amyloid-related disorders are the most common proteinopathies and are characterized by amyloid fibril deposits in the brain or other organs. Aging is generally associated with the development of these amyloid-related disorders, but we still do not fully understand how functional proteins become pathogenic amyloid deposits during the human aging process. We identified a novel amyloidogenic protein, named epidermal growth factor-containing fibulin-like extracellular matrix protein 1 (EFEMP1), in massive venous amyloid deposits in specimens that we obtained from an autopsied patient who died of gastrointestinal bleeding. Our postmortem analyses of additional patients indicate that EFEMP1 amyloid deposits frequently developed in systemic venous walls of elderly people. EFEMP1 was highly expressed in veins, and aging enhanced venous EFEMP1 expression. In addition, biochemical analyses indicated that these venous amyloid deposits consisted of C-terminal regions of EFEMP1. In vitro studies showed that C-terminal regions formed amyloid fibrils, which inhibited venous tube formation and cell viability. EFEMP1 thus caused a novel age-related venous amyloid-related disorder frequently found in the elderly population. Understanding EFEMP1 amyloid formation provides new insights into amyloid-related disorders occurring during the aging process. Copyright © 2018 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Novel intrauterine balloon tamponade systems for postpartum hemorrhage.

INTRODUCTION: Postpartum hemorrhage is the most common cause of maternal death worldwide. Although intrauterine balloon tamponade has been widely used as an effective procedure to control atonic postpartum hemorrhage, intrauterine balloon tamponade fails to control postpartum hemorrhage in approximately one-fifth of cases. The aim of this study was to evaluate the efficacy of novel intrauterine balloon tamponade systems for postpartum hemorrhage. MATERIAL AND METHODS: We have developed two novel intrauterine balloon tamponade systems to maintain proper balloon placement. One was a shaft cover with its fixture system and the other was "the Kyoto balloon system" designed to provide direct pressure onto the upper uterine cavity. The efficacy of the intrauterine balloon tamponade systems was evaluated using a silicone three-dimensionally printed postpartum uterine cavity model. RESULTS: Measurements of balloon displacement during inflation showed that the shaft cover significantly prevented the Bakri balloon from being displaced. The residual fluid volume in the upper uterine cavity was significantly less with the Kyoto balloon system than with the Bakri balloon system, indicating the effectiveness of the Kyoto balloon for upper uterine cavity tamponade. CONCLUSIONS: These innovative intrauterine balloon tamponade systems were effective for prevention of balloon displacement and for balloon tamponade of the upper uterine cavity in a 3D-printed postpartum-specific uterine cavity model.

Utero-ovarian varices and absent inferior vena cava in pregnancy.

Extensively distended and tortuous vessels on both sides of the uterus are rare incidental findings on transvaginal sonography in early pregnancy. A 31-year-old woman with a history of surgical repair for congenital intestinal stenosis was noted to have utero-ovarian varices on a transvaginal ultrasound examination during her first prenatal visit at 6 weeks' gestation. Magnetic resonance imaging revealed dilated ovarian veins along with infrarenal absence of inferior vena cava as well as the absence of external and common iliac veins. Despite concerns regarding spontaneous utero-ovarian vessels rupture, her antenatal course was uneventful. A vaginal delivery was successfully achieved without any postpartum complication. We also demonstrated an intraoperative view of the utero-ovarian vessels during cesarean section in her subsequent pregnancy. Absence of inferior vena cava and pelvic varices can occur in women with a history of neonatal surgery. Although the risk of utero-ovarian varices rupture remains unclear, vaginal delivery may be safely achieved.

The feasibility of an efficient drug design method with high-performance computers.

In this study, we propose a supercomputer-assisted drug design approach involving all-atom molecular dynamics (MD)-based binding free energy prediction after the traditional design/selection step. Because this prediction is more accurate than the empirical binding affinity scoring of the traditional approach, the compounds selected by the MD-based prediction should be better drug candidates. In this study, we discuss the applicability of the new approach using two examples. Although the MD-based binding free energy prediction has a huge computational cost, it is feasible with the latest 10 petaflop-scale computer. The supercomputer-assisted drug design approach also involves two important feedback procedures: The first feedback is generated from the MD-based binding free energy prediction step to the drug design step. While the experimental feedback usually provides binding affinities of tens of compounds at one time, the supercomputer allows us to simultaneously obtain the binding free energies of hundreds of compounds. Because the number of calculated binding free energies is sufficiently large, the compounds can be classified into different categories whose properties will aid in the design of the next generation of drug candidates. The second feedback, which occurs from the experiments to the MD simulations, is important to validate the simulation parameters. To demonstrate this, we compare the binding free energies calculated with various force fields to the experimental ones. The results indicate that the prediction will not be very successful, if we use an inaccurate force field. By improving/validating such simulation parameters, the next prediction can be made more accurate.

Treatment decision-making for post-partum hemorrhage using dynamic contrast-enhanced computed tomography.

AIM: Post-partum hemorrhage (PPH) is the leading cause of maternal mortality. Identification of the precise bleeding site is generally important to control hemorrhage, but such an approach has not been fully established in the context of PPH. We postulated that visualization of bleeding sites could aid treatment decisions in the management of PPH. METHODS: We conducted a prospective review of 26 patients who underwent dynamic computed tomography (CT) for PPH. RESULTS: A total of 17 cases presented with uterine bleeding, eight with vaginal hematomas, and one with hemoperitoneum. Overall, dynamic CT identified contrast media extravasation in the arterial phase in 12 of 26 (46.2%) cases: the upper (n = 4) and the lower uterine segment including the cervix (n = 2), subfascial space (n = 1) and vagina (n = 5). Identification of precise arterial bleeding sites using CT provided informative guidance about where to place balloons for intractable uterine bleeding, and how to manage hemoperitoneum and vaginal hematomas. In addition, dynamic CT revealed the existence of a subtype of uterine atony, which is characterized by focal active arterial bleeding in the upper uterine segment. Furthermore, negative contrast extravasation extracted cases of PPH that were well controlled without the need for surgical or radiological intervention. No patient required emergency hysterectomy to control PPH. CONCLUSION: Dynamic CT has potential clinical utility in treatment decision-making for PPH.

Molecular dynamics simulation-based evaluation of the binding free energies of computationally designed drug candidates: importance of the dynamical effects.

The computational structure-based drug design (SBDD) mainly aims at generating or discovering new chemical compounds with sufficiently large binding free energy. In any de novo drug design methods and virtual screening methods, drug candidates are selected by approximately evaluating the binding free energy (or the binding affinity). This approximate binding free energy, usually called "empirical score," is critical to the success of the SBDD. The purpose of this work is to yield physical insight into the approximate evaluation method in comparison with an exact molecular dynamics (MD) simulation-based method (named MP-CAFEE), which can predict binding free energies accurately. We calculate the binding free energies for 58 selected drug candidates with MP-CAFEE. Here, the compounds are generated by OPMF, a novel fragment-based de novo drug design method, and the ligand-protein interaction energy is used as an empirical score. The results show that the correlation between the binding free energy and the interaction energy is not strong enough to clearly distinguish compounds with nM-affinity from those with µM-affinity. This implies that it is necessary to take into account the natural protein motion with explicitly surrounded by water molecules to improve the efficiency of the drug candidate selection procedure.

Cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy without anterior temporal pole involvement: a case report.

The location of white matter lesions, especially in the anterior temporal poles (ATP), is helpful in the diagnosis of cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL). We report a 49-year-old man with CADASIL who developed migraine with atypical aura, silent lacunar infarcts, and leukoencephalopathy without involvement of the ATP. The prevalence of migraine with aura in subjects with CADASIL is several times greater than that in the general population. Particularly in patients with CADASIL, the aura is often atypical (hemiplegic, basilar, or prolonged). A diagnosis of CADASIL should be considered in patients with lacunar infarcts, leukoencephalopathy, and migraine with atypical aura, even in the absence of white matter lesion in the ATPs.

Interpretable machine learning-based individual analysis of acute kidney injury in immune checkpoint inhibitor therapy.

BACKGROUND: Acute kidney injury (AKI) is a critical complication of immune checkpoint inhibitor therapy. Since the etiology of AKI in patients undergoing cancer therapy varies, clarifying underlying causes in individual cases is critical for optimal cancer treatment. Although it is essential to individually analyze immune checkpoint inhibitor-treated patients for underlying pathologies for each AKI episode, these analyses have not been realized. Herein, we aimed to individually clarify the underlying causes of AKI in immune checkpoint inhibitor-treated patients using a new clustering approach with Shapley Additive exPlanations (SHAP). METHODS: We developed a gradient-boosting decision tree-based machine learning model continuously predicting AKI within 7 days, using the medical records of 616 immune checkpoint inhibitor-treated patients. The temporal changes in individual predictive reasoning in AKI prediction models represented the key features contributing to each AKI prediction and clustered AKI patients based on the features with high predictive contribution quantified in time series by SHAP. We searched for common clinical backgrounds of AKI patients in each cluster, compared with annotation by three nephrologists. RESULTS: One hundred and twelve patients (18.2%) had at least one AKI episode. They were clustered per the key feature, and their SHAP value patterns, and the nephrologists assessed the clusters' clinical relevance. Receiver operating characteristic analysis revealed that the area under the curve was 0.880. Patients with AKI were categorized into four clusters with significant prognostic differences (p = 0.010). The leading causes of AKI for each cluster, such as hypovolemia, drug-related, and cancer cachexia, were all clinically interpretable, which conventional approaches cannot obtain. CONCLUSION: Our results suggest that the clustering method of individual predictive reasoning in machine learning models can be applied to infer clinically critical factors for developing each episode of AKI among patients with multiple AKI risk factors, such as immune checkpoint inhibitor-treated patients.