Development of a Diagnostic Artificial Intelligence Tool for Lateral Lymph Node Metastasis in Advanced Rectal Cancer.

BACKGROUND: Metastatic lateral lymph node dissection can improve survival in patients with rectal adenocarcinoma, with or without chemoradiotherapy. However, the optimal imaging diagnostic criteria for lateral lymph node metastases remain undetermined. OBJECTIVE: To develop a lateral lymph node metastasis diagnostic artificial intelligence tool using deep learning, for patients with rectal adenocarcinoma who underwent radical surgery and lateral lymph node dissection. DESIGN: Retrospective study. SETTINGS: Multicenter study. PATIENTS: A total of 209 patients with rectal adenocarcinoma, who underwent radical surgery and lateral lymph node dissection at 15 participating hospitals, were enrolled in the study and allocated to training (n = 139), test (n = 17), or validation (n = 53) cohorts. MAIN OUTCOME MEASURES: In the neoadjuvant treatment group, images taken before pretreatment were classified as baseline images and those taken after pretreatment as presurgery images. In the upfront surgery group, presurgery images were classified as both baseline and presurgery images. We constructed 2 types of artificial intelligence, using baseline and presurgery images, by inputting the patches from these images into ResNet-18, and we assessed their diagnostic accuracy. RESULTS: Overall, 124 patients underwent surgery alone, 52 received neoadjuvant chemotherapy, and 33 received chemoradiotherapy. The number of resected lateral lymph nodes in the training, test, and validation cohorts was 2418, 279, and 850, respectively. The metastatic rates were 2.8%, 0.7%, and 3.7%, respectively. In the validation cohort, the precision-recall area under the curve was 0.870 and 0.963 for the baseline and presurgery images, respectively. Although both baseline and presurgery images provided good accuracy for diagnosing lateral lymph node metastases, the accuracy of presurgery images was better than that of baseline images. LIMITATIONS: The number of cases is small. CONCLUSIONS: An artificial intelligence tool is a promising tool for diagnosing lateral lymph node metastasis with high accuracy. DESARROLLO DE UNA HERRAMIENTA DE INTELIGENCIA ARTIFICIAL PARA EL DIAGNSTICO DE METSTASIS EN GANGLIOS LINFTICOS LATERALES EN CNCER DE RECTO AVANZADO: ANTECEDENTES:Disección de nódulos linfáticos laterales metastásicos puede mejorar la supervivencia en pacientes con adenocarcinoma del recto, con o sin quimiorradioterapia. Sin embargo, aún no se han determinado los criterios óptimos de diagnóstico por imágenes de los nódulos linfáticos laterales metastásicos.OBJETIVO:Nuestro objetivo fue desarrollar una herramienta de inteligencia artificial para el diagnóstico de metástasis en nódulos linfáticos laterales mediante el aprendizaje profundo, para pacientes con adenocarcinoma del recto que se sometieron a cirugía radical y disección de nódulos linfáticos laterales.DISEÑO:Estudio retrospectivo.AJUSTES:Estudio multicéntrico.PACIENTES:Un total de 209 pacientes con adenocarcinoma del recto, que se sometieron a cirugía radical y disección de nódulos linfáticos laterales en 15 hospitales participantes, se inscribieron en el estudio y se asignaron a cohortes de entrenamiento (n = 139), prueba (n = 17) o validación (n = 53).PRINCIPALES MEDIDAS DE RESULTADO:En el grupo de tratamiento neoadyuvante, las imágenes tomadas antes del tratamiento se clasificaron como imágenes de referencia y las posteriores al tratamiento, como imágenes previas a la cirugía. En el grupo de cirugía inicial, las imágenes previas a la cirugía se clasificaron como imágenes de referencia y previas a la cirugía. Construimos dos tipos de inteligencia artificial, utilizando imágenes de referencia y previas a la cirugía, ingresando los parches de estas imágenes en ResNet-18. Evaluamos la precisión diagnóstica de los dos tipos de inteligencia artificial.RESULTADOS:En general, 124 pacientes se sometieron a cirugía solamente, 52 recibieron quimioterapia neoadyuvante y 33 recibieron quimiorradioterapia. El número de nódulos linfáticos laterales removidos en los cohortes de entrenamiento, prueba y validación fue de 2,418; 279 y 850, respectivamente. Las tasas metastásicas fueron 2.8%, 0.7%, y 3.7%, respectivamente. En el cohorte de validación, el área de recuperación de precisión bajo la curva fue de 0.870 y 0.963 para las imágenes de referencia y antes de la cirugía, respectivamente. Aunque tanto las imágenes previas a la cirugía como las iniciales proporcionaron una buena precisión para diagnosticar metástasis en los nódulos linfáticos laterales, la precisión de las imágenes previas a la cirugía fue mejor que la de las imágenes iniciales.LIMITACIONES:El número de casos es pequeño.CONCLUSIÓN:La inteligencia artificial es una herramienta prometedora para diagnosticar metástasis en los nódulos linfáticos laterales con alta precisión. (Traducción-Dr. Aurian Garcia Gonzalez ).

Improved Object Detection Artificial Intelligence Using the Revised RetinaNet Model for the Automatic Detection of Ulcerations, Vascular Lesions, and Tumors in Wireless Capsule Endoscopy.

The use of computer-aided detection models to diagnose lesions in images from wireless capsule endoscopy (WCE) is a topical endoscopic diagnostic solution. We revised our artificial intelligence (AI) model, RetinaNet, to better diagnose multiple types of lesions, including erosions and ulcers, vascular lesions, and tumors. RetinaNet was trained using the data of 1234 patients, consisting of images of 6476 erosions and ulcers, 1916 vascular lesions, 7127 tumors, and 14,014,149 normal tissues. The mean area under the receiver operating characteristic curve (AUC), sensitivity, and specificity for each lesion were evaluated using five-fold stratified cross-validation. Each cross-validation set consisted of between 6,647,148 and 7,267,813 images from 217 patients. The mean AUC values were 0.997 for erosions and ulcers, 0.998 for vascular lesions, and 0.998 for tumors. The mean sensitivities were 0.919, 0.878, and 0.876, respectively. The mean specificities were 0.936, 0.969, and 0.937, and the mean accuracies were 0.930, 0.962, and 0.924, respectively. We developed a new version of an AI-based diagnostic model for the multiclass identification of small bowel lesions in WCE images to help endoscopists appropriately diagnose small intestine diseases in daily clinical practice.

Development and multi-institutional validation of an artificial intelligence-based diagnostic system for gastric biopsy.

To overcome the increasing burden on pathologists in diagnosing gastric biopsies, we developed an artificial intelligence-based system for the pathological diagnosis of gastric biopsies (AI-G), which is expected to work well in daily clinical practice in multiple institutes. The multistage semantic segmentation for pathology (MSP) method utilizes the distribution of feature values extracted from patches of whole-slide images (WSI) like pathologists' "low-power view" information of microscopy. The training dataset included WSIs of 4511 gastric biopsy tissues from 984 patients. In tissue-level validation, MSP AI-G showed better accuracy (91.0%) than that of conventional patch-based AI-G (PB AI-G) (89.8%). Importantly, MSP AI-G unanimously achieved higher accuracy rates (0.946 ± 0.023) than PB AI-G (0.861 ± 0.078) in tissue-level analysis, when applied to the cohorts of 10 different institutes (3450 samples of 1772 patients in all institutes, 198-555 samples of 143-206 patients in each institute). MSP AI-G had high diagnostic accuracy and robustness in multi-institutions. When pathologists selectively review specimens in which pathologist's diagnosis and AI prediction are discordant, the requirement of a secondary review process is significantly less compared with reviewing all specimens by another pathologist.

Decomposing normal and abnormal features of medical images for content-based image retrieval of glioma imaging.

In medical imaging, the characteristics purely derived from a disease should reflect the extent to which abnormal findings deviate from the normal features. Indeed, physicians often need corresponding images without abnormal findings of interest or, conversely, images that contain similar abnormal findings regardless of normal anatomical context. This is called comparative diagnostic reading of medical images, which is essential for a correct diagnosis. To support comparative diagnostic reading, content-based image retrieval (CBIR) that can selectively utilize normal and abnormal features in medical images as two separable semantic components will be useful. In this study, we propose a neural network architecture to decompose the semantic components of medical images into two latent codes: normal anatomy code and abnormal anatomy code. The normal anatomy code represents counterfactual normal anatomies that should have existed if the sample is healthy, whereas the abnormal anatomy code attributes to abnormal changes that reflect deviation from the normal baseline. By calculating the similarity based on either normal or abnormal anatomy codes or the combination of the two codes, our algorithm can retrieve images according to the selected semantic component from a dataset consisting of brain magnetic resonance images of gliomas. Moreover, it can utilize a synthetic query vector combining normal and abnormal anatomy codes from two different query images. To evaluate whether the retrieved images are acquired according to the targeted semantic component, the overlap of the ground-truth labels is calculated as metrics of the semantic consistency. Our algorithm provides a flexible CBIR framework by handling the decomposed features with qualitatively and quantitatively remarkable results.

Automatic detection of different types of small-bowel lesions on capsule endoscopy images using a newly developed deep convolutional neural network.

BACKGROUND : Previous computer-aided detection systems for diagnosing lesions in images from wireless capsule endoscopy (WCE) have been limited to a single type of small-bowel lesion. We developed a new artificial intelligence (AI) system able to diagnose multiple types of lesions, including erosions and ulcers, vascular lesions, and tumors. METHODS : We trained the deep neural network system RetinaNet on a data set of 167 patients, which consisted of images of 398 erosions and ulcers, 538 vascular lesions, 4590 tumors, and 34 437 normal tissues. We calculated the mean area under the receiver operating characteristic curve (AUC) for each lesion type using five-fold stratified cross-validation. RESULTS : The mean age of the patients was 63.6 years; 92 were men. The mean AUCs of the AI system were 0.996 (95 %CI 0.992 - 0.999) for erosions and ulcers, 0.950 (95 %CI 0.923 - 0.978) for vascular lesions, and 0.950 (95 %CI 0.913 - 0.988) for tumors. CONCLUSION : We developed and validated a new computer-aided diagnosis system for multiclass diagnosis of small-bowel lesions in WCE images.

[A Study on the Status of End-of-Life Chemotherapy for Solid Cancer after the Introduction of New Anti-Cancer Drugs].

BACKGROUND: In recent years, the decision to discontinue chemotherapy has become more difficult, and there is a tendency for chemotherapy to continue until just before death. We investigated the current state of end-of-life(EOL)chemotherapy for solid cancer patients. METHODS: Patients who died of cancer during hospitalization between January and November 2018 were included. Patients were divided into 2 groups, those who received EOL chemotherapy within 30 days of death(Near group: NG)and those who did not receive it(Far group: FG). The contents of each treatment were compared retrospectively. RESULTS: The number of patients were 46(32%)in the NG and 96(68%)in the FG. As EOL chemotherapy, the number of patients received cytotoxic drugs were 27(59%)and 68(71%), molecular targeted drugs were 6(13%)and 16(16%), immune-checkpoint inhibitors were 8(18%)and 12(12%), and hormone drugs were 0(0%)and 5(5%)in patients with NG and FG respectively(p<0.05). DISCUSSION: Minimally invasive drugs were often selected for EOL chemotherapy. It was suggested that the advent of new drugs has expanded the options for EOL chemotherapy.

Cost-utility analysis of aprepitant for patients who truly need it in Japan.

PURPOSE: Neurokinin-1 receptor antagonist (NK1RA) is recommended to prevent chemotherapy-induced nausea and vomiting (CINV) in patients who receive highly or moderately emetogenic chemotherapy (HEC or MEC, respectively). We previously reported that aprepitant, an NK1RA, was needed to control CINV in 43% and 12% of patients who received HEC and MEC, respectively (Support Care Cancer 23:905-912, 2015). To elucidate the cost-effectiveness of aprepitant in these patients, a cost-utility analysis according to the necessity of aprepitant was performed. METHODS: A decision-analytic model was developed according to the necessity of aprepitant and CINV responses in both acute and delayed phases of chemotherapy. Probabilities of health states and medical costs were derived from the results of the abovementioned trial. RESULT: In patients who received HEC and needed aprepitant, the incremental cost-effectiveness ratio (ICER) with aprepitant, relative to the regimen with no aprepitant, was 7912 US dollars (USD) per quality-adjusted life year (QALY) gained, which was far below the commonly accepted threshold of 50,000 USD/QALY. The ICER was 27,457 USD/QALY in patients who received MEC and needed aprepitant. In contrast, in patients who received HEC or MEC but did not need aprepitant, the ICER was 175,959 or 478,844 USD/QALY, respectively. CONCLUSION: Regardless of whether a patient received HEC or MEC, aprepitant use was highly cost-effective for patients who truly needed it. These results warrant further research to predict the necessity of NK1RA treatment before initiating emetogenic chemotherapies.

Autonomous Positioning of Eye Surgical Robot Using the Tool Shadow and Kalman Filtering.

Vitreoretinal surgery is one of the most difficult surgical operations, even for experienced surgeons. Thus, a master-slave eye surgical robot has been developed to assist the surgeon in safely performing vitreoretinal surgeries; however, in the master-slave control, the robotic positioning accuracy depends on the surgeon's coordination skills. This paper proposes a new method of autonomous robotic positioning using the shadow of the surgical instrument. First, the microscope image is segmented into three regions-namely, a micropipette, its shadow, and the eye ground-using a Gaussian mixture model (GMM). The tips of the micropipette and its shadow are then extracted from the contour lines of the segmented regions. The micropipette is then autonomously moved down to the simulated eye ground until the distance between the tips of micropipette and its shadow in the microscopic image reaches a predefined threshold. To handle possible occlusions, the tip of the shadow is estimated using a Kalman filter. Experiments to evaluate the robotic positioning accuracy in the vertical direction were performed. The results show that the autonomous positioning using the Kalman filter enhanced the accuracy of robotic positioning.

Cost-effectiveness of aprepitant in Japanese patients treated with cisplatin-containing highly emetogenic chemotherapy.

Chemotherapy-induced nausea and vomiting (CINV) remains a major adverse event in cancer chemotherapy. Although aprepitant is effective in preventing CINV, an increment in financial burden for uniform use of aprepitant is a concern. The aim of the present study was to define the cost-effectiveness of aprepitant from the perspective of the Japanese National Health Insurance system. Based on the results of a randomized phase II trial comparing an aprepitant-containing regimen versus a nonaprepitant regimen in Japanese patients who received cisplatin-containing highly emetogenic chemotherapy, a decision analytic model was developed. The incremental cost-effectiveness ratio (ICER) was calculated both in the outpatient care setting (OCS) and in the inpatient care setting (ICS). The use of the aprepitant-containing regimen was associated with improved quality of life compared with the nonaprepitant regimen, with an increment in quality-adjusted life years (QALY) of 0.0016. The incremental total medical costs associated with the use of the aprepitant regimen were lower in the OCS than in the ICS, 6192 JPY (56.92 USD) and 9820 JPY (90.27 USD), respectively. The ICER was calculated as 3 906 698 JPY (35 910 USD) per QALY gained in the OCS and 6 195 781 JPY (56 952 USD) per QALY gained in the ICS. Cost-effectiveness of the aprepitant-containing antiemetic therapy was limited to the OCS, considering the threshold of willingness-to-pay commonly accepted (5 million JPY [45 960 USD] in Japan and 50 000 USD in the USA). The efficacy of aprepitant offsets the costs for revisiting clinics or rehospitalization added with rescue medications in the OCS.

Intelligent control of neurosurgical robot MM-3 using dynamic motion scaling.

OBJECTIVE Advanced and intelligent robotic control is necessary for neurosurgical robots, which require great accuracy and precision. In this article, the authors propose methods for dynamically and automatically controlling the motion-scaling ratio of a master-slave neurosurgical robotic system to reduce the task completion time. METHODS Three dynamic motion-scaling modes were proposed and compared with the conventional fixed motion-scaling mode. These 3 modes were defined as follows: 1) the distance between a target point and the tip of the slave manipulator, 2) the distance between the tips of the slave manipulators, and 3) the velocity of the master manipulator. Five test subjects, 2 of whom were neurosurgeons, sutured 0.3-mm artificial blood vessels using the MM-3 neurosurgical robot in each mode. RESULTS The task time, total path length, and helpfulness score were evaluated. Although no statistically significant differences were observed, the mode using the distance between the tips of the slave manipulators improves the suturing performance. CONCLUSIONS Dynamic motion scaling has great potential for the intelligent and accurate control of neurosurgical robots.