Explainable Machine Learning for COVID-19 Pneumonia Classification With Texture-Based Features Extraction in Chest Radiography

Both reverse transcription-PCR (RT-PCR) and chest X-rays are used for the diagnosis of the coronavirus disease-2019 (COVID-19). However, COVID-19 pneumonia does not have a defined set of radiological findings. Our work aims to investigate radiomic features and classification models to differentiate chest X-ray images of COVID-19-based pneumonia and other types of lung patterns. The goal is to provide grounds for understanding the distinctive COVID-19 radiographic texture features using supervised ensemble machine learning methods based on trees through the interpretable Shapley Additive Explanations (SHAP) approach. We use 2,611 COVID-19 chest X-ray images and 2,611 non-COVID-19 chest X-rays. After segmenting the lung in three zones and laterally, a histogram normalization is applied, and radiomic features are extracted. SHAP recursive feature elimination with cross-validation is used to select features. Hyperparameter optimization of XGBoost and Random Forest ensemble tree models is applied using random search. The best classification model was XGBoost, with an accuracy of 0.82 and a sensitivity of 0.82. The explainable model showed the importance of the middle left and superior right lung zones in classifying COVID-19 pneumonia from other lung patterns.

Redefining and Validating Digital Biomarkers as Fluid, Dynamic Multi-Dimensional Digital Signal Patterns

“Digital biomarker” is a term broadly and indiscriminately applied and often limited in its conceptualization to mimic well-established biomarkers as defined and approved by regulatory agencies such as the United States Food and Drug Administration (FDA). There is a practical urgency to revisit the definition of a digital biomarker and expand it beyond current methods of identification and validation. Restricting the promise of digital technologies within the realm of currently defined biomarkers creates a missed opportunity. A whole new field of prognostic and early diagnostic digital biomarkers driven by data science and artificial intelligence can break the current cycle of high healthcare costs and low health quality that is being driven by today's chronic disease detection and treatment approaches. This new class of digital biomarkers will be dynamic and require developing new FDA approval pathways and next-generation gold standards.

Using Machine Learning to Predict Mortality for COVID-19 Patients on Day 0 in the ICU

Rationale: Given the expanding number of COVID-19 cases and the potential for new waves of infection, there is an urgent need for early prediction of the severity of the disease in intensive care unit (ICU) patients to optimize treatment strategies.Objectives: Early prediction of mortality using machine learning based on typical laboratory results and clinical data registered on the day of ICU admission.Methods: We retrospectively studied 797 patients diagnosed with COVID-19 in Iran and the United Kingdom (U.K.). To find parameters with the highest predictive values, Kolmogorov-Smirnov and Pearson chi-squared tests were used. Several machine learning algorithms, including Random Forest (RF), logistic regression, gradient boosting classifier, support vector machine classifier, and artificial neural network algorithms were utilized to build classification models. The impact of each marker on the RF model predictions was studied by implementing the local interpretable model-agnostic explanation technique (LIME-SP).Results: Among 66 documented parameters, 15 factors with the highest predictive values were identified as follows: gender, age, blood urea nitrogen (BUN), creatinine, international normalized ratio (INR), albumin, mean corpuscular volume (MCV), white blood cell count, segmented neutrophil count, lymphocyte count, red cell distribution width (RDW), and mean cell hemoglobin (MCH) along with a history of neurological, cardiovascular, and respiratory disorders. Our RF model can predict patient outcomes with a sensitivity of 70% and a specificity of 75%. The performance of the models was confirmed by blindly testing the models in an external dataset.Conclusions: Using two independent patient datasets, we designed a machine-learning-based model that could predict the risk of mortality from severe COVID-19 with high accuracy. The most decisive variables in our model were increased levels of BUN, lowered albumin levels, increased creatinine, INR, and RDW, along with gender and age. Considering the importance of early triage decisions, this model can be a useful tool in COVID-19 ICU decision-making.

Saving Time for Patient Care by Optimizing Physician Note Templates: A Pilot Study

Background: At times, electronic medical records (EMRs) have proven to be less than optimal, causing longer hours behind computers, shorter time with patients, suboptimal patient safety, provider dissatisfaction, and physician burnout. These concerning healthcare issues can be positively affected by optimizing EMR usability, which in turn would lead to substantial benefits to healthcare professionals such as increased healthcare professional productivity, efficiency, quality, and accuracy. Documentation issues, such as non-standardization of physician note templates and tedious, time-consuming notes in our mother-baby unit (MBU), were discussed during meetings with stakeholders in the MBU and our hospital's EMR analysts.Objective: The objective of this study was to assess physician note optimization on saving time for patient care and improving provider satisfaction.Methods: This quality improvement pilot investigation was conducted in our MBU where four note templates were optimized: History and Physical (H and P), Progress Note (PN), Discharge Summary (DCS), and Hand-Off List (HOL). Free text elements documented elsewhere in the EMR (e.g., delivery information, maternal data, lab result, etc.) were identified and replaced with dynamic links that automatically populate the note with these data. Discrete data pick lists replaced necessary elements that were previously free texts. The new note templates were given new names for ease of accessibility. Ten randomly chosen pediatric residents completed both the old and new note templates for the same control newborn encounter during a period of one year. Time spent and number of actions taken (clicks, keystrokes, transitions, and mouse-keyboard switches) to complete these notes were recorded. Surveys were sent to MBU providers regarding overall satisfaction with the new note templates.Results: The ten residents' average time saved was 23 min per infant. Reflecting this saved time on the number of infants admitted to our MBU between January 2016 and September, 2019 which was 9373 infants; resulted in 2.6 hours saved per day, knowing that every infant averages two days length of stay. The new note templates required 69 fewer actions taken than the old ones (H and P: 11, PN: 8, DCS: 18, HOL: 32). The provider surveys were consistent with improved provider satisfaction.Conclusion: Optimizing physician notes saved time for patient care and improved physician satisfaction.

Tele-Audiology: Current State and Future Directions

The importance of tele-audiology has been heightened by the current COVID-19 pandemic. The present article reviews the current state of tele-audiology practice while presenting its limitations and opportunities. Specifically, this review addresses: (1) barriers to hearing healthcare, (2) tele-audiology services, and (3) tele-audiology key issues, challenges, and future directions. Accumulating evidence suggests that tele-audiology is a viable service delivery model, as remote hearing screening, diagnostic testing, intervention, and rehabilitation can each be completed reliably and effectively. The benefits of tele-audiology include improved access to care, increased follow-up rates, and reduced travel time and costs. Still, significant logistical and technical challenges remain from ensuring a secure and robust internet connection to controlling ambient noise and meeting all state and federal licensure and reimbursement regulations. Future research and development, especially advancements in artificial intelligence, will continue to increase tele-audiology acceptance, expand remote care, and ultimately improve patient satisfaction.

Pilot Evaluation of the Impact of Lottery-Based Incentives on Engagement Levels of Male Low SES Vocational Students With an mHealth App

In general, individuals with lower socioeconomic status (SES) are less physically active and adhere to poorer diets than higher SES individuals. To promote healthier lifestyles in lower SES populations, we hosted a digital health promotion program among male vocational students at a school in The Netherlands. In a pilot study, we evaluated whether this target audience could be engaged with an mHealth app using lottery-based incentives that trigger feelings of anticipated regret. Especially, we studied the social and interpersonal aspects of regret lotteries in a within-subject experimental design. In this design, subjects either participated in a social variant (i.e., with students competing against their peers for a chance at a regret lottery), or an individual variant (i.e., with subjects solely individually engaged in a lottery). Additionally, we studied the impact of different payout schedules in a between-subject experimental design. In this design, participants were assigned to either a short-term, low-value payout schedule, or a long-term, high-value payout schedule. From a population of 72 male students, only half voluntarily participated in our 10-week program. From interviews, we learned that the main reason for neglecting the program was not related to the lottery-based incentives, nor to the prizes that were awarded. Instead, non-enrolled subjects did not join the program, because their peers were not joining. Paradoxically, it was suggested that students withheld their active participation until a larger portion of the sample was actively participating. From the subjects that enrolled in the program (N = 36, males, between 15 and 25 years of age), we found that a large proportion stopped interacting with the program over time (e.g., after roughly 4 weeks). Our results also indicated that students performed significantly more health-related activities when assigned to the social regret lottery, as opposed to the individual variant. This result was supported by interview responses from active participants: They mainly participated to compete against their peers, and not so much for the prizes. Hence, from this study, we obtained initial evidence on the impact of social and competitive aspects in lottery-based incentives to stimulate engagement levels in lower SES students with an mHealth app.

You Don't Need an App—Conducting Mobile Smoking Research Using a Qualtrics-Based Approach

With the increasing availability of smartphones, many tobacco researchers are exploring smartphone-delivered mobile smoking interventions as a disseminable means of treatment. Most effort has been focused on the development of smartphone applications (apps) to conduct mobile smoking research to implement and validate these interventions. However, developing project-specific smartphone apps that work across multiple mobile platforms (e.g., iOS and Android) can be costly and time-consuming. Here, using a hypothetical study, we present an alternate approach to demonstrate how mobile smoking cessation and outcome evaluation can be conducted without the need of a dedicated app. Our approach uses the Qualtrics platform, a popular online survey host that is used under license by many academic institutions. This platform allows researchers to conduct device-agnostic screening, consenting, and administration of questionnaires through Qualtrics's native survey engine. Researchers can also collect ecological momentary assessment data using text messaging prompts with the incorporation of Amazon Web Services' Pinpoint. Besides these assessment capabilities, Qualtrics has the potential for delivering personalized behavioral interventions through the use of JavaScript code. By customizing the question's web elements in Qualtrics (e.g., using texts, images, videos, and buttons), researchers can integrate interactive web-based interventions and complicated behavioral and cognitive tasks into the survey. In conclusion, this Qualtrics-based methodology represents a novel and cost-effective approach for conducting mobile smoking cessation and assessment research.

Participatory Design and Process Testing to Optimize Utility, Usability, and Acceptability of a Mobile Game for Promoting Evidence-Driven Public Health Decision-Making in Resource-Constrained Settings

Innovative game-based training methods that leverage the ubiquity of cellphones and familiarity with phone-based interfaces have the potential to transform the training of public health practitioners in low-income countries such as Liberia. This article describes the design, development, and testing of a prototype of the Figure It Out mobile game. The prototype game uses a disease outbreak scenario to promote evidence-based decision-making in determining the causative agent and prescribing intervention measures to minimize epidemiological and logistical burdens in resource-limited settings. An initial prototype of the game developed by the US team was playtested and evaluated by focus groups with 20 University of Liberia Masters of Public Health (UL MPH) students. Results demonstrate that the learning objectives—improving search skills for identifying scientific evidence and considering evidence before decision-making during a public health emergency—were considered relevant and important in a setting that has repeatedly and recently experienced severe threats to public health. However, some of the game mechanics that were thought to enhance engagement such as trial-and-error and choose-your-own-path gameplay, were perceived by the target audience as distracting or too time-consuming, particularly in the context of a realistic emergency scenario. Gameplay metrics that mimicked real-world situations around lives lost, money spent, and time constraints during public health outbreaks were identified as relatable and necessary considerations. Our findings reflect cultural differences between the game development team and end users that have emphasized the need for end users to have an integral part of the design team; this formative evaluation has critically informed next steps in the iterative development process. Our multidisciplinary, cross-cultural and cross-national design team will be guided by Liberia-based public health students and faculty, as well as community members who represent our end user population in terms of experience and needs. These stakeholders will make key decisions regarding game objectives and mechanics, to be vetted and implemented by game design experts, epidemiologists, and software developers.

Automatic Segmentation of Kidneys and Kidney Tumors: The KiTS19 International Challenge

Purpose: Clinicians rely on imaging features to calculate complexity of renal masses based on validated scoring systems. These scoring methods are labor-intensive and are subjected to interobserver variability. Artificial intelligence has been increasingly utilized by the medical community to solve such issues. However, developing reliable algorithms is usually time-consuming and costly. We created an international community-driven competition (KiTS19) to develop and identify the best system for automatic segmentation of kidneys and kidney tumors in contrast CT and report the results.Methods: A training and test set of CT scans that was manually annotated by trained individuals were generated from consecutive patients undergoing renal surgery for whom demographic, clinical and outcome data were available. The KiTS19 Challenge was a machine learning competition hosted on grand-challenge.org in conjunction with an international conference. Teams were given 3 months to develop their algorithm using a full-annotated training set of images and an unannotated test set was released for 2 weeks from which average Sørensen-Dice coefficient between kidney and tumor regions were calculated across all 90 test cases.Results: There were 100 valid submissions that were based on deep neural networks but there were differences in pre-processing strategies, architectural details, and training procedures. The winning team scored a 0.974 kidney Dice and a 0.851 tumor Dice resulting in 0.912 composite score. Automatic segmentation of the kidney by the participating teams performed comparably to expert manual segmentation but was less reliable when segmenting the tumor.Conclusion: Rapid advancement in automated semantic segmentation of kidney lesions is possible with relatively high accuracy when the data is released publicly, and participation is incentivized. We hope that our findings will encourage further research that would enable the potential of adopting AI into the medical field.

CovNet: A Transfer Learning Framework for Automatic COVID-19 Detection From Crowd-Sourced Cough Sounds

Since the COronaVIrus Disease 2019 (COVID-19) outbreak, developing a digital diagnostic tool to detect COVID-19 from respiratory sounds with computer audition has become an essential topic due to its advantages of being swift, low-cost, and eco-friendly. However, prior studies mainly focused on small-scale COVID-19 datasets. To build a robust model, the large-scale multi-sound FluSense dataset is utilised to help detect COVID-19 from cough sounds in this study. Due to the gap between FluSense and the COVID-19-related datasets consisting of cough only, the transfer learning framework (namely CovNet) is proposed and applied rather than simply augmenting the training data with FluSense. The CovNet contains (i) a parameter transferring strategy and (ii) an embedding incorporation strategy. Specifically, to validate the CovNet's effectiveness, it is used to transfer knowledge from FluSense to COUGHVID, a large-scale cough sound database of COVID-19 negative and COVID-19 positive individuals. The trained model on FluSense and COUGHVID is further applied under the CovNet to another two small-scale cough datasets for COVID-19 detection, the COVID-19 cough sub-challenge (CCS) database in the INTERSPEECH Computational Paralinguistics challengE (ComParE) challenge and the DiCOVA Track-1 database. By training four simple convolutional neural networks (CNNs) in the transfer learning framework, our approach achieves an absolute improvement of 3.57% over the baseline of DiCOVA Track-1 validation of the area under the receiver operating characteristic curve (ROC AUC) and an absolute improvement of 1.73% over the baseline of ComParE CCS test unweighted average recall (UAR).