Search

Cytology, a type of pathological examination, involves sampling cells from the human body and observing the morphology of the nucleus, cytoplasm, and cell arrangement. In developing classification AI technologies to support cytology, it is essential to collect and utilize a diverse range of images without bias. However, this is often challenging in practice because of the epidemiologic bias of cancer types and cellular characteristics. The main aim of this study was to develop a method to generate cytological diagnostic images from image findings using text-to-image technology in order to generate diverse images. In the proposed method, we collected Papanicolaou-stained specimens derived from the lung cells of 135 lung cancer patients, from which we extracted 472 patch images. Descriptions of the corresponding findings for these patch images were compiled to create a data set. This dataset was then utilized to finetune the Stable Diffusion (SD) v1 and v2 models. The cell images generated by this method closely resemble real images, and both cytotechnologists and cytopathologists provided positive subjective evaluations. Furthermore, SDv2 demonstrated shapes and contours of nuclei and cytoplasm that were more similar to real images compared to SDv1, showing superior performance in quantitative evaluation metrics. When the generated images were utilized in the classification tasks for cytological images, there was an improvement in classification performance. These results indicate that the proposed method may be effective for generating high-quality cytological images, which enables the image classification model to learn diverse features, thereby improving classification performance.

Abstract A 77‐year‐old male patient underwent esophagogastroduodenoscopy at his family doctor, and an easily hemorrhagic depressed lesion was noted near the anterior wall of the gastric antrum. A biopsy revealed moderately differentiated tubular adenocarcinoma > poorly differentiated adenocarcinoma, and the patient was referred to our department for further examination. A 15‐mm 0–IIc lesion is seen near the anterior wall of the gastric antrum and narrow band imaging magnifying endoscopy revealed obscured glandular duct structures and corkscrew pattern vascular structures. We diagnosed the patient with early‐stage gastric cancer [L, Ant, 15mm, cType0‐IIc, cT1(M‐SM1), cN0, cM0, cStage IA] after an esopahogastroduodenoscopy examination at our hospital, and endoscopic submucosal dissection was performed. Histopathological images with hematoxylin and eosin staining showed tumor cells with pale cytoplasm and the immunostaining for alpha‐fetoprotein, sal‐like protein 4, and Glypican3 was positive. The patient was pathologically diagnosed with gastric adenocarcinoma with enteroblastic differentiation, pT1b1 (SM, 0.4 mm), type 0–IIc, 15 mm, UL (‐), Ly0, and V0. Gastric adenocarcinoma with enteroblastic differentiation is one of the representative histological types of alpah‐fetoprotein‐producing gastric cancer. Alpha‐fetoprotein‐producing gastric cancer is infrequent, accounting for at least 3% of all gastric cancers, and is generally highly malignant. Most cases are already advanced upon diagnosis, and finding them in the early stage is rare. Therefore, pathological findings that may indicate the gastric adenocarcinoma with enteroblastic differentiation should be noted even in early gastric cancer.

Purpose: We evaluated the medium-term radiological outcomes of the interface bioactive bone cement (IBBC) technique used in the acetabular component in primary cemented total hip arthroplasty (THA). Methods: In total, 79 patients (88 hips) who underwent primary THA at our hospital between January 2004 and April 2009 were divided into the non-IBBC (NIBBC; n ​= ​39 patients, 44 hips) and IBBC (n ​= ​40 patients, 44 hips) groups. A clear zone (CZ) at the cement-bone interface around the cup was evaluated 5 and 10 years postoperatively. The CZ score, which indicates the spread of the CZ, and loosening of the cup (CZ ​> ​2 ​mm on all sides or movement of the cup >4°) were also evaluated. Results: No patient had loosening of the cup. The mean patient age at surgery was 66.5 ​± ​9.2 years in the NIBBC group and 61.9 ​± ​7.7 years in the IBBC group (p ​= ​0.012). At 5 years postoperatively, the incidence of CZ was 36.4 ​% in the NIBBC group and 18.2 ​% in the IBBC group (p ​= ​0.056). At 10 years postoperatively, the incidence of CZ was 36.4 ​% in the NIBBC group and 20.1 ​% in the IBBC group (p ​= ​0.098). Age at surgery (p ​= ​0.045) significantly affected the incidence of CZ at 5 years but not at 10 years. The IBBC technique (p ​= ​0.042) and age at surgery (p ​= ​0.028) significantly affected the CZ score at 5 years. The IBBC technique (p ​= ​0.036) significantly affected the CZ score at 10 years. Conclusions: The IBBC technique resulted in a significantly lower CZ score than the NIBBC technique in patients undergoing THA.

Objective: The early detection of cardiac disease is important because the disease can lead to sudden death and poor prognosis. Electrocardiograms (ECG) are used to screen for cardiac diseases and are useful for the early detection and determination of treatment strategies. However, the ECG waveforms of cardiac care unit (CCU) patients with severe cardiac disease are often complicated by comorbidities and patient conditions, making it difficult to predict the severity of further cardiac disease. Therefore, this study predicts the short-term prognosis of CCU patients to detect further deterioration in CCU patients at an early stage. Methods: The ECG data (II, V3, V5, aVR induction) of CCU patients were converted to image data. The transformed ECG images were used to predict short-term prognosis with a two-dimensional convolutional neural network (CNN). Results: The prediction accuracy was 77.3%. Visualization by GradCAM showed that the CNN tended to focus on the shape and regularity of waveforms, such as heart failure and myocardial infarction. Conclusion: These results suggest that the proposed method may be useful for short-term prognosis prediction using the ECG waveforms of CCU patients. Clinical impact: The proposed method could be used to determine the treatment strategy and choose the intensity of treatment after admission to the CCU.

Abstract Glycosylation of proteins and lipids occurs in vertebrates, usually terminating with sialylation, which regulates the physicochemical and biological properties of these glycoconjugates. Although less commonly known, sialic acid residues also undergo various modifications, such as acetylation, methylation, and sulfation. However, except for acetylation, the enzymes or functions of the other modification processes are unknown. To the best of our knowledge, this study is the first to demonstrate the ubiquitous occurrence of sulfated sialic acids and two genes encoding the sialate: O-sulfotransferases 1 and 2 in vertebrates. These two enzymes showed about 50% amino acid sequence identity, and appeared to be complementary to each other in acceptor substrate preferences. Gene targeting experiments showed that the deficiency of these genes was lethal for medaka fish during young fry development and accompanied by different phenotypes. Thus, the sulfation of sialic acids is essential for the vertebrate development.

Abstract Artificial intelligence (AI) applications in medical imaging continue facing the difficulty in collecting and using large datasets. One method proposed for solving this problem is data augmentation using fictitious images generated by generative adversarial networks (GANs). However, applying a GAN as a data augmentation technique has not been explored, owing to the quality and diversity of the generated images. To promote such applications by generating diverse images, this study aims to generate free-form lesion images from tumor sketches using a pix2pix-based model, which is an image-to-image translation model derived from GAN. As pix2pix, which assumes one-to-one image generation, is unsuitable for data augmentation, we propose StylePix2pix, which is independently improved to allow one-to-many image generation. The proposed model introduces a mapping network and style blocks from StyleGAN. Image generation results based on 20 tumor sketches created by a physician demonstrated that the proposed method can reproduce tumors with complex shapes. Additionally, the one-to-many image generation of StylePix2pix suggests effectiveness in data-augmentation applications.

Heart failure (HF) with preserved left ventricular ejection fraction (LVEF) is a heterogeneous syndrome. An LVEF of 50% is widely used to categorize patients with HF; however, this is controversial. Previously, we have reported that patients with an LVEF of ≥ 58% have good prognoses. Further, cardiac sympathetic nervous system (SNS) activation is a feature of HF. In this retrospective, observational study, the cardiac SNS activity of HF patients (n = 63, age: 78.4 ± 9.6 years; male 49.2%) with LVEF ≥ 58% (n = 15) and LVEF < 58% (n = 48) were compared using 123I-metaiodobenzylguanidine scintigraphy. During the follow-up period (median, 3.0 years), 18 all-cause deaths occurred. The delayed heart/mediastinum (H/M) ratio was significantly higher in the LVEF ≥ 58% group than in the LVEF < 58% group (2.1 ± 0.3 vs. 1.7 ± 0.4, p = 0.004), and all-cause mortality was significantly lower in patients in the former than those in the latter group (log-rank, p = 0.04). However, when these patients were divided into LVEF ≥ 50% (n = 22) and LVEF < 50% (n = 41) groups, no significant differences were found in the delayed H/M ratio, and the all-cause mortality did not differ between the groups (log-rank, p = 0.09). In conclusion, an LVEF of 58% is suitable for reclassifying patients with HF according to cardiac SNS activity.

Abstract Objective Among standard treatments for infantile spasms, adrenocorticotropic hormone (ACTH) is reported as the best treatment, but ACTH is ineffective in one‐half of the patients. To establish precision medicine, we examined pharmacoresistance of focal epileptic spasms (ES), generalized ES, and generalized ES combined with focal seizures, diagnosed based on the revised seizure classification of ILAE in 2017. Methods We conducted a retrospective nationwide study in Japan on the long‐term seizure outcome of ES. Long‐term seizure outcome was evaluated by seizure‐free rate, seizure‐free period, and Kaplan‐Meier curve. Seizure‐free was defined as seizure control for longer than 2 months. Results From the medical history of 501 patients, 325 patients had generalized ES only (GES group) at the start of the first treatment, 125 patients had generalized ES after focal seizure onset (FS‐GES group), seven patients had focal ES after focal seizure onset (FS‐FES group), and 24 patients had generalized ES combined with focal seizures after focal seizure onset (FS‐GES + FS group). Seizure‐free period of ES (generalized ES and focal ES) [mean (95% confidence interval)] was 2.7 (0.0‐5.4) months in GES group, 1.1 (0.1‐2.2) months in FS‐GES group, 1.0 (0.2‐1.9) months in FS‐GES + FS group, and 0.1 (−0.2‐0.5) months in FS‐FES group. Seizure‐free rate, seizure‐free period, and Kaplan‐Meier curve of generalized ES were almost the same in GES group and FS‐GES group, with characteristics of superior response to ACTH. Mean seizure‐free period of generalized ES combined with focal seizures was significantly shorter in FS‐GES + FS group than in GES group. Mean seizure‐free period of focal ES in FS‐FES group was extremely short with exceedingly early relapse. Significance Pharmacoresistance was different in generalized ES, focal ES, and generalized ES combined with focal seizures. ES with focal features or with focal seizures may have focal lesions, thus consider surgical options earlier in the course.

Abstract In cytological examination, suspicious cells are evaluated regarding malignancy and cancer type. To assist this, we previously proposed an automated method based on supervised learning that classifies cells in lung cytological images as benign or malignant. However, it is often difficult to label all cells. In this study, we developed a weakly supervised method for the classification of benign and malignant lung cells in cytological images using attention-based deep multiple instance learning (AD MIL). Images of lung cytological specimens were divided into small patch images and stored in bags. Each bag was then labeled as benign or malignant, and classification was conducted using AD MIL. The distribution of attention weights was also calculated as a color map to confirm the presence of malignant cells in the image. AD MIL using the AlexNet-like convolutional neural network model showed the best classification performance, with an accuracy of 0.916, which was better than that of supervised learning. In addition, an attention map of the entire image based on the attention weight allowed AD MIL to focus on most malignant cells. Our weakly supervised method automatically classifies cytological images with acceptable accuracy based on supervised learning without complex annotations.

Abstract Background Socket fixation with bone grafting for dysplastic hips is technically demanding, and inadequate coverage of the socket may cause poor results in patients with severely dysplastic hips. An accurate technique to form a bone graft to fit into the defect is necessary. We aim to introduce the simple method of bone grafting, “inverted reamer technique” in cemented total hip arthroplasty (cTHA). Methods After acetabular preparation with a normal acetabular reamer, the bone graft was prepared from the resected femoral head with the inverted reamer. The graft can be press-fit into the defect of the acetabulum with good compatibility through this method. Then, the bone graft was fixed with 1–3 screws and the socket was implanted with bone cement. Results The “inverted reamer technique” can easily and automatically create a well-fit graft. This method is simple and technically less demanding; it can be performed by every surgeon, including trainee and inexperienced surgeons. Conclusion This method can improve the outcome of cTHA for dysplastic hips by preserving bone stock and increasing bone coverage of the socket implanted in the anatomic position.

PET/CT can scan low-dose computed tomography (LDCT) images with morphological information and PET images with functional information. Because the whole body is targeted for imaging, PET/CT examinations are important in cancer diagnosis. However, the several images obtained by PET/CT place a heavy burden on radiologists during diagnosis. Thus, the development of computer-aided diagnosis (CAD) and technologies assisting in diagnosis has been requested. However, because FDG accumulation in PET images differs for each organ, recognizing organ regions is essential for developing lesion detection and analysis algorithms for PET/CT images. Therefore, we developed a method for automatically extracting organ regions from PET/CT images using U-Net or DenseUNet, which are deep-learning-based segmentation networks. The proposed method is a hybrid approach combining morphological and functional information obtained from LDCT and PET images. Moreover, pre-training using ImageNet and RadImageNet was performed and compared. The best extraction accuracy was obtained by pre-training ImageNet with Dice indices of 94.1, 93.9, 91.3, and 75.1% for the liver, kidney, spleen, and pancreas, respectively. This method obtained better extraction accuracy for low-quality PET/CT images than did existing studies on PET/CT images and was comparable to existing studies on diagnostic contrast-enhanced CT images using the hybrid method and pre-training.

Background: In the resting conditions, narrowing the window of coronary pressure measurements from the whole cardiac cycle to diastole improves diagnostic performance of coronary pressure–derived physiological index. However, whether this also applies to the hyperemic conditions has not yet been thoroughly evaluated. Objectives: The purpose of this study was to assess whether diastolic fractional flow reserve (diastolic FFR) has better diagnostic performance in identifying ischemia-causing coronary lesions than conventional FFR in a prospective, multicenter, and independent core laboratory–based environment. Methods: In this prospective multicenter registry at 29 Japanese centers, we compared the diagnostic performance of FFR, diastolic FFR, resting distal to aortic coronary pressure (Pd/Pa), and diastolic pressure ratio (dPR) using myocardial perfusion scintigraphy (MPS) as the reference standard in 378 patients with single-vessel coronary disease. Results: Inducible myocardial ischemia was found on MPS in the relevant myocardial territory of the target vessel in 85 patients (22%). In the receiver-operating curve analyses, diastolic FFR had comparable area under the curve (AUC) compared with FFR (AUCdiastolic FFR: 0.66; 95% confidence interval [CI]: 0.58-0.73, vs AUCFFR: 0.66; 95% CI: 0.58-0.74, P = 0.624). FFR and diastolic FFR showed significantly larger AUCs than resting Pd/Pa (0.62; 95% CI: 0.54-0.70; P = 0.033 and P = 0.046) but did not show significantly larger AUCs than dPR (0.62; 95% CI: 0.55-0.70; P = 0.102 and P = 0.113). Conclusions: Diastolic FFR showed a similar diagnostic performance to FFR as compared with MPS. This result reaffirms the use of FFR as the most accurate invasive physiological lesion assessment. (Diagnostic accuracy of diastolic fractional flow reserve (d-FFR) for functional evaluation of coronary stenosis; UMIN000015906)

Abstract Background In our institute, all elderly patients with displaced femoral neck fracture were treated with cemented bipolar hemiarthroplasty (BHA) using the modified Dall approach. To our knowledge, there are no reports on the knot position of the greater trochanter reattachment. The aim of this study was to determine influence of two knot positions (anterior or posterior) on the complications of the greater trochanter. Methods This is a prospective non-randomized study conducted on 95 elderly patients (95 hips) from September 2013 to December 2017. The knot position was changed from anterior to posterior alternately. The X-ray images obtained immediately after the operation were compared with those obtained at 3 months postoperatively; thereafter, the status of the greater trochanter was classified into three types: type A, no apparent shifting and fracture; type C, over 1-mm shifting of the fragment; and type F, fracture of the greater trochanter. Results Regarding age at operation, sex, BMI, size of the greater trochanteric fragment, stem type, and surgeon, there was no significant difference between two groups. In the anterior group, 34 hips (72.3%), 5 hips (10.6%), and 8 hips (17.0%) were classified under type A, C, and F, respectively. In the posterior group, 44 hips (91.7%), 1 hip (2.1%), and 3 hips (6.3%) were classified under type A, C, and F, respectively. There were significantly fewer greater trochanteric complications in the posterior group. Conclusions The posterior knot position improved the union of the greater trochanter after BHA compared with the anterior knot position. Trial registration We had approved IRB at our hospital clinical research review committee. Retrospectively registered.

Abstract. The effect of renal dysfunction on clinical outcomes following fractional flow reserve (FFR)-guided deferral of revascularization remains unelucidated. We retrospectively analyzed 224 patients with atherosclerotic coronary lesions who underwent deferred revascularization based on an FFR of >0.80. The median follow-up interval was 28.1 months. Patients were divided into 2 groups: the hemodialysis (HD) and the non-HD group. The non-HD group was further classified into 2 subgroups according to their estimated glomerular filtration rate (eGFR) level: eGFR

BackgroundSurgery for intertrochanteric fractures using intramedullary hip nails (IHNs) is among the most common surgical procedures in the orthopedic field. Although IHNs provide good overall outcomes, they sometimes cause complications, such as loss of reduction and cut-out. Here, we investigated the usefulness of IHNs with an anterior offset (Best Fit Nail® [BFN]) in maintaining fragment reduction and ensuring proper lag screw position compared with conventional non-offset nails (Proximal Femoral Nail Antirotation® [PFNA]), using postoperative computed tomography (CT).MethodsFifty consecutive patients with intertrochanteric fractures who underwent surgery with BFNs (BFN group) and 50 patients who underwent surgery with PFNAs (PFNA group) were retrospectively analyzed. Indices evaluated by postoperative CT were displacement distance of proximal fragment relative to distal fragment, reduction status (intramedullary, anatomical, and extramedullary types), lag screw direction, and angle between lag screw and femoral neck axis (deviation angle).ResultsMedian [interquartile range] displacement distance was significantly smaller in the BFN group (0 [0, 0] mm) compared with the PFNA group (5.2 [3.6, 7.1] mm) (pConclusionsIntertrochanteric fracture surgery using offset BFNs exhibited significantly smaller displacement distance, better reduction status, and higher frequency of no deviation with central lag screw position, compared with surgery using non-offset PFNAs.

Interstitial pneumonia of uncertain cause is referred to as idiopathic interstitial pneumonia (IIP). Among the various types of IIPs, the prognosis of cases of idiopathic pulmonary fibrosis (IPF) is extremely poor, and accurate differentiation between IPF and non-IPF pneumonia is critical. In this study, we consider deep learning (DL) methods owing to their excellent image classification capabilities. Although DL models require large quantities of training data, collecting a large number of pathological specimens is difficult for rare diseases. In this study, we propose an end-to-end scheme to automatically classify IIPs using a convolutional neural network (CNN) model. To compensate for the lack of data on rare diseases, we introduce a two-step training method to generate pathological images of IIPs using a generative adversarial network (GAN). Tissue specimens from 24 patients with IIPs were scanned using a whole slide scanner, and the resulting images were divided into patch images with a size of 224 × 224 pixels. A progressive growth GAN (PGGAN) model was trained using 23,142 IPF images and 7817 non-IPF images to generate 10,000 images for each of the two categories. The images generated by the PGGAN were used along with real images to train the CNN model. An evaluation of the images generated by the PGGAN showed that cells and their locations were well-expressed. We also obtained the best classification performance with a detection sensitivity of 97.2% and a specificity of 69.4% for IPF using DenseNet. The classification performance was also improved by using PGGAN-generated images. These results indicate that the proposed method may be considered effective for the diagnosis of IPF.

In computed tomography colonography (CTC), an electric cleansing technique is used to mix barium with residual fluid, and colon residue is removed by image processing. However, a nonhomogenous mixture of barium and residue may not be properly removed. We developed an electronic cleansing method using CycleGAN, a deep learning technique, to assist diagnosis in CTC. In this method, an original computed tomography (CT) image taken during a CTC examination and a manually cleansed image in which the barium area was manually removed from the original CT image were prepared and converted to an image in which the barium was removed from the original CT image using CycleGAN. In the experiment, the electric cleansing images obtained using the conventional method were compared with those obtained using the proposed method. The average barium cleansing rates obtained by the conventional and proposed methods were 72.3% and 96.3%, respectively. A visual evaluation of the images showed that it was possible to remove only barium without removing the intestinal tract. Furthermore, the extraction of colorectal polyps and early stage cancerous lesions in the colon was performed as in the conventional method. These results indicate that the proposed method using CycleGAN may be useful for accurately visualizing the colon without barium.

Background Myocardial extracellular volume fraction (ECV), measured by cardiac magnetic resonance imaging, is a useful prognostic marker for patients who have undergone aortic valve replacement (AVR) for aortic stenosis. However, the prognostic significance of ECV measurements based on computed tomography (CT) is unclear. This study evaluated the association between ECV measured with dual‐energy CT and clinical outcomes in patients with aortic stenosis who underwent transcatheter or surgical AVR. Methods and Results We retrospectively enrolled 95 consecutive patients (age, 84.0±5.0 years; 75% women) with severe aortic stenosis who underwent preprocedural CT for transcatheter AVR planning. ECV was measured using iodine density images obtained by delayed enhancement dual‐energy CT. The primary end point was a composite outcome of all‐cause death and hospitalization for heart failure after AVR. The mean ECV measured with CT was 28.1±3.8%. During a median follow‐up of 2.6 years, 22 composite outcomes were observed, including 15 all‐cause deaths and 11 hospitalizations for heart failure. In Kaplan‐Meier analysis, the high ECV group (≥27.8% [median value]) had significantly higher rates of composite outcomes than the low ECV group (

Purpose: The purpose of this study was to develop and evaluate lung cancer segmentation with a pretrained model and transfer learning. The pretrained model was constructed from an artificial dataset generated using a generative adversarial network (GAN).Materials and Methods: Three public datasets containing images of lung nodules/lung cancers were used: LUNA16 dataset, Decathlon lung dataset, and NSCLC radiogenomics. The LUNA16 dataset was used to generate an artificial dataset for lung cancer segmentation with the help of the GAN and 3D graph cut. Pretrained models were then constructed from the artificial dataset. Subsequently, the main segmentation model was constructed from the pretrained models and the Decathlon lung dataset. Finally, the NSCLC radiogenomics dataset was used to evaluate the main segmentation model. The Dice similarity coefficient (DSC) was used as a metric to evaluate the segmentation performance.Results: The mean DSC for the NSCLC radiogenomics dataset improved overall when using the pretrained models. At maximum, the mean DSC was 0.09 higher with the pretrained model than that without it.Conclusion: The proposed method comprising an artificial dataset and a pretrained model can improve lung cancer segmentation as confirmed in terms of the DSC metric. Moreover, the construction of the artificial dataset for the segmentation using the GAN and 3D graph cut was found to be feasible.

Objective(s): Positron emission tomography/computed tomography (PET/CT) examination is commonly used for the evaluation of pulmonary nodules since it provides both anatomical and functional information. However, given the dependence of this evaluation on physician’s subjective judgment, the results could be variable. The purpose of this study was to develop an automated scheme for the classification of pulmonary nodules using early and delayed phase PET/ CT and conventional CT images.Methods: We analysed 36 early and delayed phase PET/CT images in patients who underwent both PET/CT scan and lung biopsy, following bronchoscopy. In addition, conventional CT images at maximal inspiration were analysed. The images consisted of 18 malignant and 18 benign nodules. For the classification scheme, 25 types of shape and functional features were first calculated from the images. The random forest algorithm, which is a machine learning technique, was used for classification.Results: The evaluation of the characteristic features and classification accuracy was accomplished using collected images. There was a significant difference between the characteristic features of benign and malignant nodules with regard to standardised uptake value and texture. In terms of classification performance, 94.4% of the malignant nodules were identified correctly assuming that 72.2% of the benign nodules were diagnosed accurately. The accuracy rate of benign nodule detection by means of CT plus two-phase PET images was 44.4% and 11.1% higher than those obtained by CT images alone and CT plus early phase PET images, respectively.Conclusion: Based on the findings, the proposed method may be useful to improve the accuracy of malignancy analysis.

Endoscopy is widely applied in the examination of gastric cancer. However, extensive knowledge and experience are required, owing to the need to examine the lesion while manipulating the endoscope. Various diagnostic support techniques have been reported for this examination. In our previous study, segmentation of invasive areas of gastric cancer was performed directly from endoscopic images and the detection sensitivity per case was 0.98. This method has challenges of false positives and computational costs because segmentation was applied to all healthy images that were captured during the examination. In this study, we propose a cascaded deep learning model to perform categorization of endoscopic images and identification of the invasive region to solve the above challenges. Endoscopic images are first classified as normal, showing early gastric cancer and showing advanced gastric cancer using a convolutional neural network. Segmentation on the extent of gastric cancer invasion is performed for the images classified as showing cancer using two separate U-Net models. In an experiment, 1208 endoscopic images collected from healthy subjects, 533 images collected from patients with early stage gastric cancer, and 637 images from patients with advanced gastric cancer were used for evaluation. The sensitivity and specificity of the proposed approach in the detection of gastric cancer via image classification were 97.0% and 99.4%, respectively. Furthermore, both detection sensitivity and specificity reached 100% in a case-based evaluation. The extent of invasion was also identified at an acceptable level, suggesting that the proposed method may be considered useful for the classification of endoscopic images and identification of the extent of cancer invasion.

Objective: Papanicolaou and Giemsa stains used in cytology have different characteristics and complementary roles. In this study, we focused on cycle-consistent generative adversarial network (CycleGAN), which is an image translation technique using deep learning, and we conducted mutual stain conversion between Giemsa and Papanicolaou in cytological images using CycleGAN. Methods: A total of 191 Giemsa-stained images and 209 Papanicolaou-stained images were collected from 63 patients with lung cancer. From those images, 67 images from nine cases were used for testing and the remaining images were used for training. For data augmentation, the number of training images was increased by rotation and inversion, and the images were pipelined to CycleGAN to train the mutual conversion process involving Giemsa- and Papanicolaou-stained images. Three pathologists and three cytotechnologists performed visual evaluations of the authenticity of cell nuclei, cytoplasm, and cell layouts of the test images translated using CycleGAN. Results: As a result of converting Giemsa-stained images into Papanicolaou-stained images, the background red blood cell patterns present in Giemsa-stained images disappeared, and cell patterns that reproduced the shape and staining of the cell nuclei and cytoplasm peculiar to Papanicolaou staining were obtained. Regarding the reverse-translated results, nuclei became larger, and red blood cells that were not evident in Papanicolaou-stained images appeared. After visual evaluation, although actual images exhibited better results than converted images, the results were promising for various applications. Discussion: The stain translation technique investigated in this paper can complement specimens under conditions where only single stained specimens are available; it also has potential applications in the massive training of artificial intelligence systems for cell classification, and can also be used for training cytotechnologist and pathologists.

Patients with inactive acetaldehyde dehydrogenase 2 (ALDH2) are at high risk for esophageal squamous cell carcinoma (ESCC) and hypopharyngeal squamous cell carcinoma (HPSCC). The acetaldehyde breath test (ABT) may demonstrate ALDH2 gene polymorphisms. We evaluated the usefulness of the ABT in patients with ESCC and HPSCC. The squamous cell carcinoma (SCC) group consisted of 100 patients who were treated with endoscopic submucosal dissection (ESD) for ESCC or HPSCC, and the control group (HC) consisted of 275 healthy subjects. The SCC group comprised the "single subgroup" (n = 63), in which a single lesion was initially treated with ESD, and the "multiple subgroup" (n = 31), in which multiple lesions were initially treated with ESD. First, we compared the groups' risk factors for carcinogenesis and measured the acetaldehyde-to-ethanol (A/E) ratio. Then we tested the groups' differences in the abovementioned carcinogenic risk factors. We found that the proportion of individuals in the SCC group with inactive ALDH2 (A/E ratio ≥ 23.3) was significantly higher than that in the HC group (p = 0.035), as was the A/E ratio (p < 0.001). Also, the proportion of individuals with inactive ALDH2 in the multiple subgroup was significantly higher than that in single subgroup (p = 0.015), as was the A/E ratio (p = 0.008). In conclusion, ABT may be a potential screening tool for detecting people at risk of ESCC and HPSCC. In addition, it could be a useful tool in detecting patients at risk of multiple or double carcinomas among patients with ESCC and HPSCC. Trial registration: Trial Registration number: UMIN000040615 [https://rctportal.niph.go.jp/en/detail?trial_id=UMIN000040615], Data of Registration: 01 46 June 2020, retrospectively registered.

Cytology is the first pathological examination performed in the diagnosis of lung cancer. In our previous study, we introduced a deep convolutional neural network (DCNN) to automatically classify cytological images as images with benign or malignant features and achieved an accuracy of 81.0%. To further improve the DCNN's performance, it is necessary to train the network using more images. However, it is difficult to acquire cell images which contain a various cytological features with the use of many manual operations with a microscope. Therefore, in this study, we aim to improve the classification accuracy of a DCNN with the use of actual and synthesized cytological images with a generative adversarial network (GAN). Based on the proposed method, patch images were obtained from a microscopy image. Accordingly, these generated many additional similar images using a GAN. In this study, we introduce progressive growing of GANs (PGGAN), which enables the generation of high-resolution images. The use of these images allowed us to pretrain a DCNN. The DCNN was then fine-tuned using actual patch images. To confirm the effectiveness of the proposed method, we first evaluated the quality of the images which were generated by PGGAN and by a conventional deep convolutional GAN. We then evaluated the classification performance of benign and malignant cells, and confirmed that the generated images had characteristics similar to those of the actual images. Accordingly, we determined that the overall classification accuracy of lung cells was 85.3% which was improved by approximately 4.3% compared to a previously conducted study without pretraining using GAN-generated images. Based on these results, we confirmed that our proposed method will be effective for the classification of cytological images in cases at which only limited data are acquired.

Since recognizing the location and extent of infarction is essential for diagnosis and treatment, many methods using deep learning have been reported. Generally, deep learning requires a large amount of training data. To overcome this problem, we generated pseudo patient images using CycleGAN, which performed image transformation without paired images. Then, we aimed to improve the extraction accuracy by using the generated images for the extraction of cerebral infarction regions. First, we used CycleGAN for data augmentation. Pseudo-cerebral infarction images were generated from healthy images using CycleGAN. Finally, U-Net was used to segment the cerebral infarction region using CycleGAN-generated images. Regarding the extraction accuracy, the Dice index was 0.553 for U-Net with CycleGAN, which was an improvement over U-Net without CycleGAN. Furthermore, the number of false positives per case was 3.75 for U-Net without CycleGAN and 1.23 for U-Net with CycleGAN, respectively. The number of false positives was reduced by approximately 67% by introducing the CycleGAN-generated images to training cases. These results indicate that utilizing CycleGAN-generated images was effective and facilitated the accurate extraction of the infarcted regions while maintaining the detection rate.

Abstract Background Patients with left ventricular outflow tract obstruction who do not exhibit a dynamic pressure gradient at rest, experience pressure gradient increases of ≥ 30 mmHg only during specific situations; this is called latent left ventricular outflow tract obstruction. It is provoked by increased cardiac contraction and preload and afterload depletion. There are a few reports of patients with it developing cardiac arrest. We present a case of latent left ventricular outflow tract obstruction in which the patient with a sigmoid septum experienced refractory pulseless electrical activity due to conventional advanced cardiac life support. Case presentation A 73-year-old Asian woman on escitalopram and lorazepam was transported to our hospital for chest and back pain with altered consciousness. On arrival, she was in shock and developed pulseless electrical activity. After initiation of conventional cardiopulmonary resuscitation according to adult advanced cardiovascular life support guidelines, she could not regain spontaneous circulation. She was ultimately resuscitated via venoarterial extracorporeal membrane oxygenation initiation. The only abnormal laboratory result at admission was anemia. Her hemodynamic status stabilized after red blood cell transfusion, and venoarterial extracorporeal membrane oxygenation was subsequently terminated. Transthoracic echocardiography showed a sigmoid septum; dobutamine-infused Doppler echocardiography revealed a significant outflow gradient, and continuous monitoring showed Brockenbrough–Braunwald sign, which confirmed a diagnosis of latent left ventricular outflow tract obstruction due to a sigmoid septum. As a result, carvedilol and verapamil were initiated. A follow-up dobutamine-infused Doppler echocardiography showed a reduction of outflow gradient, and she was discharged without any sequelae. Latent left ventricular outflow tract obstruction worsened due to increasing cardiac contraction and the depletion of preload and afterload. Depleted preload occurred due to dehydration and anemia, whereas depleted afterload occurred due to the prescribed drugs, which subsequently caused pulseless electrical activity. Moreover, β-stimulation from the adrenaline probably enhanced the hypercontractile state and caused refractory pulseless electrical activity in our case. Conclusions Patients with latent left ventricular outflow tract obstruction can progress to cardiogenic shock and pulseless electrical activity due to increased cardiac contraction and depletion of preload and afterload. We should consider the patient’s underlying conditions that induced pulseless electrical activity.