Cross-domain learning in the medical field is a research challenge as many conditions like in oncology imaging use different imaging modalities. Moreover, in most of the medical learning applications, the sample training size is relatively small. Although few-shot learning (FSL) through the use of Siamese neural network was able to be trained on small sample with remarkable accuracy, FSL fails to be effective for the use in multiple domains as their convolution weights are set for task-specific application. In this paper, we are addressing this problem by enabling FSL to posses the ability to shift cross domains through designing two-layer FSL network that can learn individually from each domain and produce shared features map with extra modulation to be used at the second layer that can recognize important targets from mix domains. Our initial experimentations based on mixed medical dataset like the Medical-MNIST reveals promising results. We aim to continue this research to perform full scale-analytics for testing our cross-domain FSL learning.

Cataract image severity classification according to contrast sensitivity chart like VTSC represents a challenge to classical machine learning. Typically cataract images hide significant retina features because of the fogy nature of these images. Contrast sensitivity measurement provides important information on the functioning of the visual system that cannot simple detected by the visual acuity tests. In this paper, we are introducing a framework for segmenting fundus images into regions of salient contrast. The framework start by converting the RGB color channels into joint entropy where it can be used to categorized anchor images into contrast based classes. It has been found that classes like normal fundus, fundus with cataract or fundus having cataract with additional complications like availability of cotton-wool spots or fibrosis. We found that contrast sensitivity scoring correlate nicely with the joint entropy scoring. For this purpose, our last level of the framework employs a triplet loss Siamese neural network that has been trained on few contrast classified fundus images to detect the severity of the cataract. We used a similar contrast sensitivity scale like those used by the optometrist with classification rate that reaches 72%.

Practicing evidence based medicine requires establishing relevant and focused clinical questions that allow physicians to seek appropriate answers from rigor of the research and medical practice. These researchable question is not only important for diagnosis, prognosis, treatment and therapy but also for compiling clinical case reports when clinician encounters unanswered issues in current clinical practice. However, clinicians face multiple challenges in formulating the questions manually, searching medical literature as well as summarizing the methods and outcome from the finding of the most relevant articles. This paper extends our QL4POMR to provide a PICO clinical questioning wrapper as well as literature finding summarizers based on two pre-trained models like the BART and Bio-BERT. We also managed to integrate a clinical case report generator that can synthesis information on the clinical case based on its SOAP(s) description as well as the PICO questions and the clinical finding literature summary related to these PICO questions. Compiling the variations of the PICO questions and the integration harmonization with the SOAP cases has been done using the GraphQL API and the flexibility of the Neo4J graph-based representations for both the SOAP cases and the PubMed literature objects.

Capsule endoscopy (CE) is an established noninvasive diagnostic modality in investigating small bowel disease. CE has a pivotal role in assessing patients with suspected bleeding or identifying evidence of active Crohn’s disease in the small bowel. However, CE produces lengthy videos with at least eighty thousand frames, with a frequency rate of 2 frames per second. Gastroenterologists cannot dedicate 8 to 15 hours for reading the CE video frames to arrive at a diagnosis. This is why the issue of analyzing CE videos based on modern artificial intelligence techniques becomes a necessity. However, machine learning including deep learning has failed to report robust results because of the lack of large samples to train its neural nets. In this paper we are describing a thick data approach that learns from few anchor images provided by sound data sets like KVASIR and CrohnIPI to filter candidate frames that include interesting anomalies at any CE video and allow these candidate frames to a feature extraction process to provide representative measures on the anomaly, like the size of the anomaly and the color contrast compared to the image background, and later feed these features to a decision tree that can classify the candidate frames as having a condition like the Crohn's Disease. Our thick data approach as reported accuracy of detecting Crohn’s Disease based on availability of ulcer areas at the candidate frames for KVASIR was 89.9% and for the CrohnIPI was 83.3%. We are continuing our research to fine tune our approach by adding more thick data methods for enhancing the diagnosis accuracy.

Data integration of health through connected services can save lives in the event of a medical emergency or to provide efficient and effective interventions for the benefit of the patients through the integration of bedside and bench side clinical research. Such integration will support all wind of change in healthcare for being predictive, pre-emptive, personalized, problem-oriented and participatory. Prototyping a healthcare system that enables data integration is a big challenge for healthcare for a long time. However, an innovative solution started to emerge by focusing on problem lists where everything can connect the problem list forming a growing graph. This notion has been introduced by Dr. Lawrence Weed in early 70’s but the enabling technologies weren’t mature enough to provide a successful implementation prototype. In this article we are describing our efforts in prototyping Dr. Lawrence Weed problem-oriented medical record (POMR) and his patient case schema (SOAP) to shape a prototype for connected health. For this we are using the TypeGraphQL API and our enterprise-based QL4POMR to describe a Web-Based gateway for healthcare services connectivity. Our prototype has reported success in connecting to the HL7 FHIR medical record and the OpenTarget biomedical repositories.

Machine learning including deep learning cannot effectively result in reliable and robust accuracy in classifying medical images unless the neural net is trained on extremely large annotated samples. With many medical practices like the inflammatory bowel diseases (IBD), the annotated training data is relatively small requiring more additive techniques to bring focus to the context and then to predict important indicators like identifying acute ulcerative colitis regions of interest. This paper introduced the notion of thick data that can bring the required context by adding the expert heuristics in the form of anchor images and along with the use of Siamese neural network that uses triple ranking loss, it can rate the ulcerative colitis severity level according to the Mayo endoscopic (MES) scoring system. Many other thick data techniques can be added as cascading filters on this core model. We are anticipating many other results to follow in our future research.

Physicians and Clerkships follow the SOAP note in describing and presenting patient cases. The SOAP note was originated from the problem-oriented medical record (POMR) developed nearly 60 years ago by Lawrence Weed, MD. However, the POMR/SOAP is not commonly found in electronic medical records (EMR) used today due to the flexible nature of building patient cases that requires complex harmonization with variety of patient case schemas. In this article we attempted to use the GraphQL API for harmonizing patient case data communicated with different care interface and providing the query on the data. This harmonization interface is called QL4POMR as a GraphQL implementation to the POMR SOAP note. Physicians can used this interface to describe and present any patient case for the purpose of diagnosis and prognosis with a varying backend. The QL4POMR implemented a mapping module to map graphs from POMR to HL7 FHIR and vise versa.

Every healthcare unit is experiencing tremendous pressure to improve its practice quality across several dimensions. Multiple bodies of literature support the importance of establishing community of practice (CoP) to enrich the professional practice and add the expert context on the patient cases. The CoP emphasizes the importance of qualitative social learning and connectivity as preferred sources of knowledge updates to guide the practice rather than using the mere direct quantitative evidence. Social learning and connectivity in CoP is a complex sociotechnical process that takes an abstract idea through a cycle of participation and reification to derive more thickened context and refined knowledge that will help largely the accuracy of decision making. This process is not a straightforward one requiring the use of suitable hyper structure for representing the contextual evolving knowledge as well as a flexible infrastructure to enable CoP learning from experts, agents and connected services as well as other sources of data and knowledge. This article focuses on using the notion of workflow as the hyper structure and Node-RED as the platform that can facilitate CoP learning and connectivity. The focus is on using the CoP Node-RED workflows in healthcare settings to provide basic collaboration and connectivity as well as extensions to facilitate higher participation, learning, and connectivity to arrive at reification of the practice experience. With Node-RED workflows ideas can be represented as flows and sub flows where it can be shared with other CoP members as JSON hyper structure for further improvement, analytics and decision making.