Deep Learning Enables Prostate MRI Segmentation: A Large Cohort Evaluation With Inter-Rater Variability Analysis

Whole-prostate gland (WPG) segmentation plays a significant role in prostate volume measurement, treatment, and biopsy planning. This study evaluated a previously developed automatic WPG segmentation, deep attentive neural network (DANN), on a large, continuous patient cohort to test its feasibility in a clinical setting. With IRB approval and HIPAA compliance, the study cohort included 3,698 3T MRI scans acquired between 2016 and 2020. In total, 335 MRI scans were used to train the model, and 3,210 and 100 were used to conduct the qualitative and quantitative evaluation of the model. In addition, the DANN-enabled prostate volume estimation was evaluated by using 50 MRI scans in comparison with manual prostate volume estimation. For qualitative evaluation, visual grading was used to evaluate the performance of WPG segmentation by two abdominal radiologists, and DANN demonstrated either acceptable or excellent performance in over 96% of the testing cohort on the WPG or each prostate sub-portion (apex, midgland, or base). Two radiologists reached a substantial agreement on WPG and midgland segmentation (κ = 0.75 and 0.63) and moderate agreement on apex and base segmentation (κ = 0.56 and 0.60). For quantitative evaluation, DANN demonstrated a dice similarity coefficient of 0.93 ± 0.02, significantly higher than other baseline methods, such as DeepLab v3+ and UNet (both p values < 0.05). For the volume measurement, 96% of the evaluation cohort achieved differences between the DANN-enabled and manual volume measurement within 95% limits of agreement. In conclusion, the study showed that the DANN achieved sufficient and consistent WPG segmentation on a large, continuous study cohort, demonstrating its great potential to serve as a tool to measure prostate volume.

Operationalizing GPT-3 in Healthcare: An outlook of compliance, trust, and access for pretrained large AI linguistic models (Preprint)

UNSTRUCTURED Generative Pre-trained Transformer (GPT) models have been popular recently with their enhanced capability and performance. In contrast to many existing Artificial Intelligence (AI) models, GPT can perform with very limited training data. GPT-3 is one of the latest releases in this pipeline, demonstrating human-like logical and intellectual responses to prompts: some examples are including writing essays, complex question answering, matching pronouns to their noun, and sentiment analysis. However, its implementation in healthcare is still a question mark in terms of operationalization and its use in clinical practice and research. In this viewpoint paper, we outlined three major operational factors that drive the adoption of GPT-3 in healthcare: (1) Health Insurance Portability and Accountability Act (HIPAA) compliance, (2) building trust with healthcare providers, and (3) establishing the broader access to the GPT-3 tools.

Taking control of your own pervasive mHealth Data: A Proof of Concept using a configurable Blockchain and mHealth application on iOS and Android devices (Preprint)

BACKGROUND Background: Blockchains offer a promising new distributed technology to address the challenges of data standardization, system interoperability, security, privacy, and accessibility for all data. However, integrating pervasive computing with blockchain’s ability to store privacy-protected mHealth data while providing HIPAA compliance is a challenge. Patients use a multitude of devices, apps, and services to collect and store mHealth data. Before the advent of blockchains, providing anonymized privacy controlled single point of access for different data sources for each user was a challenging problem. We present the design of an IoT-based configurable blockchain with different mHealth applications on iOS and Android collecting the same user’s data. We discuss the advantages of using such a blockchain architecture and demonstrate two things – the ease with which users can retain full control of their pervasive mHealth data and the ease with which HIPAA compliance can be accomplished by provider(s) who choose to access user data. We also allude to the future of shareable and tradeable data with our paper. OBJECTIVE Objective: The purpose of this paper is to design, evaluate and test IoT-based mHealth data using wearable devices using an efficient configurable blockchain designed and implemented ground up to store such data. The purpose of this paper is to demonstrate the privacy-preserving and HIPAA-compliant nature of pervasive computing-based personalized healthcare systems that give users total control of their own data. METHODS Methods: This paper followed the methodical design science approach adapted in information systems wherein we evaluate prior designs, propose enhancements with a Blockchain design pattern published by the same author(s), and use the design to support IoT transactions. We prototype both the blockchain and the IoT-based mHealth applications in different devices and test all use cases that formed the design goals for such a system. Specifically, we validate the design goals for our system using the HIPAA checklist for businesses and prove compliance of our architecture for mHealth data on pervasive computing devices. RESULTS Results: Blockchain-based personalized healthcare systems provide several advantages over traditional systems. They support the following features: provide and support extreme privacy protection, ability to share personalized data, provide the ability to delete data upon request, and support the ability to work on data. CONCLUSIONS Conclusions: We conclude that blockchain(s) and specifically the CHASM architecture presented in this paper, with configurable module(s) and a Software as a service Model provide many advantages for patients using pervasive devices that store mHealth data on the blockchain. Among them, is the ability to store, retrieve and modify one(s) generated healthcare data with a single private key across devices. This data is transparent and stored perennially and provides patients the privacy and pseudo-anonymity in addition to very strong encryption for data access. Firms and Device manufacturers would be benefited from such an approach wherein they relinquish user data control, while giving users the ability to select and offer their own mHealth data on data marketplaces. We show that such an architecture complies with the stringent requirements of HIPAA for patient data access.

Medical photography with a mobile phone: useful techniques, and what neurosurgeons need to know about HIPAA compliance

Medical photographs are commonly employed to enhance education, research, and patient care throughout the neurosurgical discipline. Current mobile phone camera technology enables surgeons to quickly capture, document, and share a patient scenario with colleagues. Research demonstrates that patients generally view clinical photography favorably, and the practice has become an integral part of healthcare. Neurosurgeons in satellite locations often rely on residents to send photographs of diagnostic imaging studies, neurological examination findings, and postoperative wounds. Images are also frequently obtained for research purposes, teaching and learning operative techniques, lectures and presentations, comparing preoperative and postoperative outcomes, and patient education. However, image quality and technique are highly variable. Capturing and sharing photographs must be accompanied by an awareness of the legal ramifications of the Health Insurance Portability and Accountability Act (HIPAA). HIPAA compliance is straightforward when one is empowered with the knowledge of what constitutes a patient identifier in a photograph. Little has been published to describe means of improving the accuracy and educational value of medical photographs in neurosurgery. Therefore, in this paper, the authors present a brief discussion regarding four easily implemented photography skills every surgeon who uses his or her mobile phone for patient care should know: 1) provide context, 2) use appropriate lighting, 3) use appropriate dimensionality, and 4) manage distracting elements. Details of the HIPAA-related components of mobile phone photographs and patient-protected health information are also included.

Healthcare Data Security Technology: HIPAA Compliance

Information technology (IT) plays an increasingly important and prominent role in the health sector. Data security is more important than ever to the healthcare industry and in world in general. The number of data breaches compromising confidential healthcare data is on the rise. For data security, cloud computing is very useful for securing data. Due to data storage issue, there is a need to use the electronic communication, and a number of methods have been developed for data security technology. Health Insurance Portability and Accountability Act (HIPAA) is one of the methods that can help in healthcare research. On stored database of patient in hospital or clinic, we can develop a conservational and analytical method so as to keep the medical records of the patients in a well-preserved and adequate environment. The method includes the improvement of working possibilities by delivering all the details necessary for the patient. All the information must be identified clearly. The protection of the privacy of the patients and the security of their information are the most imperative obstacles to obtain their intakes when considering the adoption of useful health data in the electronic field of healthcare industries.

An Updated “Call Database” to Support Lab Consultation and Quality Improvement Projects

Clinical laboratories may have “on-call” residents and fellows, a practice that assists the laboratory in providing quality support to clinicians and provides valuable clinical experience to trainees. A Call Database may be used as a tool to facilitate call handoffs, as an educational resource and knowledge base for trainees, and as a call logging and review tool as required by the Accreditation Council for Graduate Medical Education (ACGME). The Call Database at our institution currently contains over 51,000 entries and is used by all clinical pathology residents and fellows while “on call.” Prior versions of our Call Database (first developed in 2004 and updated in 2014) were self-hosted applications with limited feature sets that were difficult to support; moreover, they provided few guarantees around electronic protected health information (ePHI), a topic of increasing concern as HIPAA compliance among clinical laboratories is increasingly audited. A review of the current application also identified that it has limited usefulness in supporting faculty review of calls, does not easily allow structured data entry for quality improvement projects, and lacks features commonly seen in modern web applications (eg, rich text editing and file attachments). Here we describe the latest update to our existing Call Database addressing these feature limitations. In addition, we discuss our approach to using a modern cloud-based infrastructure design that address prior shortcomings in data security, user management, and ease of development. The updated Call Database is a single-page web application, compatible with a wide range of local and cloud environments. We updated the user interface with features such as auto-saving of entries, rich text entry, file attachments, and topic tagging. Several new features facilitate the faculty review process: (1) a dedicated “review mode” for rapid viewing and commenting of relevant calls, (2) automated weekly emails by topic sent to faculty, and (3) a user “mention” feature so that trainees can solicit the opinion of faculty in the text of a call or comment itself. Finally, we provide support for customizing the structured data fields of call entries, which enables support of quality improvement and monitoring projects via the call database (eg, a customized entry for monitoring use of specific send-out tests, massive transfusion reactions, etc.). Our application delivers a modern, performant, and easy to use call database for use by trainees and faculty alike. We utilize the Amazon Cloud (AWS) to host the application and have developed a specific set of compliance and risk review documents to address HIPAA compliance. Future work will incorporate user feedback and will focus on supporting the implementation of our application at other sites or clinical pathology residency programs.