Dataset Information

"KAIZEN" method realizing implementation of deep-learning models for COVID-19 CT diagnosis in real world hospitals.

ABSTRACT: Numerous COVID-19 diagnostic imaging Artificial Intelligence (AI) studies exist. However, none of their models were of potential clinical use, primarily owing to methodological defects and the lack of implementation considerations for inference. In this study, all development processes of the deep-learning models are performed based on strict criteria of the "KAIZEN checklist", which is proposed based on previous AI development guidelines to overcome the deficiencies mentioned above. We develop and evaluate two binary-classification deep-learning models to triage COVID-19: a slice model examining a Computed Tomography (CT) slice to find COVID-19 lesions; a series model examining a series of CT images to find an infected patient. We collected 2,400,200 CT slices from twelve emergency centers in Japan. Area Under Curve (AUC) and accuracy were calculated for classification performance. The inference time of the system that includes these two models were measured. For validation data, the slice and series models recognized COVID-19 with AUCs and accuracies of 0.989 and 0.982, 95.9% and 93.0% respectively. For test data, the models' AUCs and accuracies were 0.958 and 0.953, 90.0% and 91.4% respectively. The average inference time per case was 2.83 s. Our deep-learning system realizes accuracy and inference speed high enough for practical use. The systems have already been implemented in four hospitals and eight are under progression. We released an application software and implementation code for free in a highly usable state to allow its use in Japan and globally.

SUBMITTER: Okada N

PROVIDER: S-EPMC10799049 | biostudies-literature | 2024 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

"KAIZEN" method realizing implementation of deep-learning models for COVID-19 CT diagnosis in real world hospitals.

Okada Naoki N Umemura Yutaka Y Shi Shoi S Inoue Shusuke S Honda Shun S Matsuzawa Yohsuke Y Hirano Yuichiro Y Kikuyama Ayano A Yamakawa Miho M Gyobu Tomoko T Hosomi Naohiro N Minami Kensuke K Morita Natsushiro N Watanabe Atsushi A Yamasaki Hiroyuki H Fukaguchi Kiyomitsu K Maeyama Hiroki H Ito Kaori K Okamoto Ken K Harano Kouhei K Meguro Naohito N Unita Ryo R Koshiba Shinichi S Endo Takuro T Yamamoto Tomonori T Yamashita Tomoya T Shinba Toshikazu T Fujimi Satoshi S

Scientific reports 20240119 1

Numerous COVID-19 diagnostic imaging Artificial Intelligence (AI) studies exist. However, none of their models were of potential clinical use, primarily owing to methodological defects and the lack of implementation considerations for inference. In this study, all development processes of the deep-learning models are performed based on strict criteria of the "KAIZEN checklist", which is proposed based on previous AI development guidelines to overcome the deficiencies mentioned above. We develop ...[more]

PMID: 38243054

Similar Datasets

Project description:BackgroundKaizen-a management technique increasingly employed in health care-enables employees, regardless of their hierarchy level, to contribute to the improvement of their organization. The approach puts special emphasis on frontline employees because it represents one of their main opportunities to participate directly in decision making. In this study, we aimed to (1) understand the experiences of nurses in two hospitals that had recently implemented kaizen, and (2) identify factors affecting the implementation of the technique.MethodsBy means of purposeful sampling, we selected 30 nurses from different units in two private acute care hospitals in Switzerland in May 2018. We used the Organizational Transformation Model to conduct semi-structured interviews and perform qualitative content analysis. Lastly, originating from Herzberg's motivation theory, we suggest two types of factor influencing the implementation of kaizen-hygiene factors that may prevent nurses from getting demotivated, and motivational factors that may boost their motivation.ResultsNurses generally experienced kaizen as a positive practice that enabled them to discuss work-related activities in a more comprehensive manner. In some cases, however, a lack of visible improvement in the workplace lowered nurses' motivation to make suggestions. Nurses' attitudes towards kaizen differed across both hospitals depending on the available managerial support, resources such as infrastructure and staffing levels.ConclusionsFrom our findings, we derived several coping strategies to help health practitioners implement kaizen for the benefit of their organization and employees: Strong managerial support, appropriate use of kaizen tools, and a greater sense of team cohesion, among other factors, can influence how effectively hospital teams implement kaizen. To reap the benefits of kaizen, hospital managers should promote the exchange of opinions across hierarchy levels, allocate the necessary resources in terms of personnel and infrastructure, and show nurses how the technique can help them improve their workplace.

Project description:BackgroundThere is much interest in virtual consultations using video technology. Randomized controlled trials have shown video consultations to be acceptable, safe, and effective in selected conditions and circumstances. However, this model has rarely been mainstreamed and sustained in real-world settings.ObjectiveThe study sought to (1) define good practice and inform implementation of video outpatient consultations and (2) generate transferable knowledge about challenges to scaling up and routinizing this service model.MethodsA multilevel, mixed-method study of Skype video consultations (micro level) was embedded in an organizational case study (meso level), taking account of national context and wider influences (macro level). The study followed the introduction of video outpatient consultations in three clinical services (diabetes, diabetes antenatal, and cancer surgery) in a National Health Service trust (covering three hospitals) in London, United Kingdom. Data sources included 36 national-level stakeholders (exploratory and semistructured interviews), longitudinal organizational ethnography (300 hours of observations; 24 staff interviews), 30 videotaped remote consultations, 17 audiotaped face-to-face consultations, and national and local documents. Qualitative data, analyzed using sociotechnical change theories, addressed staff and patient experience and organizational and system drivers. Quantitative data, analyzed via descriptive statistics, included uptake of video consultations by staff and patients and microcategorization of different kinds of talk (using the Roter interaction analysis system).ResultsWhen clinical, technical, and practical preconditions were met, video consultations appeared safe and were popular with some patients and staff. Compared with face-to-face consultations for similar conditions, video consultations were very slightly shorter, patients did slightly more talking, and both parties sometimes needed to make explicit things that typically remained implicit in a traditional encounter. Video consultations appeared to work better when the clinician and patient already knew and trusted each other. Some clinicians used Skype adaptively to respond to patient requests for ad hoc encounters in a way that appeared to strengthen supported self-management. The reality of establishing video outpatient services in a busy and financially stretched acute hospital setting proved more complex and time-consuming than originally anticipated. By the end of this study, between 2% and 22% of consultations were being undertaken remotely by participating clinicians. In the remainder, clinicians chose not to participate, or video consultations were considered impractical, technically unachievable, or clinically inadvisable. Technical challenges were typically minor but potentially prohibitive.ConclusionsVideo outpatient consultations appear safe, effective, and convenient for patients in situations where participating clinicians judge them clinically appropriate, but such situations are a fraction of the overall clinic workload. As with other technological innovations, some clinicians will adopt readily, whereas others will need incentives and support. There are complex challenges to embedding video consultation services within routine practice in organizations that are hesitant to change, especially in times of austerity.

Dataset Information

"KAIZEN" method realizing implementation of deep-learning models for COVID-19 CT diagnosis in real world hospitals.

Publications

"KAIZEN" method realizing implementation of deep-learning models for COVID-19 CT diagnosis in real world hospitals.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets