A telemedicine system allows clinicians to remotely interact with patients and other doctors. A specialist can provide consultancy via video conferencing services integrated with electronic healthcare records management. It allows us to handle non-emergency cases immediately at the ease of doctors and patients. An AI-powered telemedicine system can bring a paradigm shift in currently known methods and benefits of telemedicine. Today’s AI is majorly data driven and employs powerful machine learning (ML) tools to handle as well as process large volumes of data (big data). Continuously evolving ML algorithms and increasing computational power have made real-time data analysis possible. Big data is going to be an important attribute of the proposed telemedicine system. It would include radiological data such as CT and X-ray images, digitized pathological, clinical, and occupational data etc. AI algorithms will be used to extract relevant information from these datasets to build diagnosis and prognosis models and assist the doctors who would be remotely interacting with patients. This project aims towards (i) tailored made solution for providing telehealth solutions based on existing solutions for screening of covid19 suspects, (ii) Deployment of multiple kiosk at covid19 hotspot areas, (iii) AI-based model development and enhancement for sensor data analytics for covid19 screening.
Recently published research in ACS Chemical Neuroscience from IIT Jodhpur shows that COVID 19 has significant neurological impact such as patient loss smell and taste. These two symptoms are now inducted by ICMR as symptoms of COVID 19. Now the question is if we can measure somehow the loss of smell and taste in a gradient manner or quantitative manner, it would be extremely useful to detect asymptomatic patients. Towards this vision we are working on (i) Development of an AI based platform trained on previous data sets with a definite set of parameters to correctly aid in early diagnosis of COVID-19 and predict their likely course of treatment, so basically we are talking about the development of an artificial intelligence software that triages COVID-19 suspects based on defined parameters of loss of smell (anosmia) and taste (ageusia), (ii) Development of AI-based platform to trawl through all the current literature relating to the disease and study the host receptors and structure of the virus to identify the COVID-19 hotspots to trace suitable drug candidates.
An accurate estimate of coronavirus (COVID-19) infection progression is necessary for optimized treatment and morbidity reduction. We aim at the prognosis of COVID-19 infection based on the longitudinal studies on multimodal data including X-Ray images, X-Ray reports, CT scans, clinical data such as body temperature, and treatment given to the patients. Our main goal is to predict the severity of infection in lungs due to COVID-19 for subsequent days given the data for current days, primarily to augment the decisions of physicians. We will also consider the effect of the presence of comorbidities on COVID-19 prognosis and the impact of COVID-19 on comorbities and co-existing diseases. This project proposes to build an artificial intelligence (AI) driven hybrid model using a combination of advanced deep learning methods and belief networks. Deep learning has shown success in accurate diagnosis of many diseases such as pneumonia and also in developing prognostic biomarkers for diseases such as colorectal cancer. Very recently, studies have been reported to demonstrate effectiveness of AI tools in COVID-19 diagnosis, however, to the best of our knowledge no work has been reported to build an AI driven prognosis model of COVID-19. The first step towards building such a model is to predict the level of severity of infection. Given a multimodal dataset, we train a deep encoder- decoder neural network for level of infection prediction. The proposed network generates a compressed representation of the input data which is fed into another branch of the network (classifier) to predict the infection level. Further, using the compressed representations of the data of previous days, we train a Long Short Term Memory (LSTM) based sequence to sequence prediction network to obtain the representations of subsequent days. This would allow us to use the classifier for future infection level prediction using the compressed representations predicted by the LSTM model without requiring the actual data. Accordingly, we can estimate the progression of COVID-19 infection. As of now the data for COVID-19 is limited, however continuously growing. We will first use the available data for viral pneumonia to develop the proposed deep learning model and subsequently fine tune it for COVID- 19, which has symptoms similar to viral pneumonia. Thus, the proposed framework would be an extendable framework which can be used for other diseases in the future. Further, to take into account the co-existing diseases and estimate the severity level in presence of comorbidities, we will incorporate the prediction of the deep neural network based classifier in a Bayesian framework. We will also extend the approach to evaluate the impact of COVID-19 infection on existing comorbidities. Bayesian framework is adopted considering that the availability of a large amount of data for comorbidity analysis is unlikely. Accordingly, we will make use of articles reporting case studies on patients with comorbidities.
In the current situation, it is crucial to increase the number of individuals getting tested so that appropriate measures can be taken. The traditional way of COVID testing is a time taking process with the confirmed results taking more than a day. Due to the mismatch in the availability of kits and the number of individuals to be tested, it is essential to find additional accurate ways of testing and screening. This project aims to address the problem of COVID-19 detection using cues present in frontal chest X-ray images. Chest X-ray contains vital information such as the presence of opacities, mismatch in lung ratios, and cues from bronchi. While these details are prominent, we currently have limited samples to train the deep learning model. This project aims to build efficient machine learning models that not only predict whether the chest X-Ray has symptoms of COVID-19 or not, but it also presents the regions and characteristics based on which the model has given a particular prediction. Since a limited number of samples are available for training the models, the project will involve designing efficient machine learning algorithms that perform segmentation and classification with a smaller number of training samples.
With an increasing number of COVID-19 cases globally, all the countries are ramping up the testing numbers. While the RT-PCR kits are available in sufficient quantities in several countries, some are facing the challenges with limited availability of testing kits and processing centers in remote areas. This has motivated researchers to find alternate methods of testing which are reliable, easily accessible and faster. Chest X-Ray and CT images are some of the modalities that are gaining acceptance as an alternate screening modality. AI and ML based systems which automatically screen the suspected COVID-19 patients from X-Ray and CT images play a critical role in preventing the spread of the coronavirus due to no direct contact between the suspect and doctor during the assessment process. Towards this direction, our research has two primary contributions. We present a COVID-19 Multi-Task Network (CMTNet) which is an automated end-to-end network for COVID-19 screening. The proposed CMTNet performs both classification and segmentation of the lung and disease regions for X-Ray screening of the COVID-10 suspects. The CMTNet further predicts if lungs are affected with COVID-19 or Non-COVID-19 disorders and differentiate them from healthy lungs to further enhance the screening process. Inclusion of simultaneous disease segmentation in the CMTNet helps in making the decisions explainable. This enables the doctors to analyze the automatic screening decisions in more detail and take the appropriate decision quickly. It is observed that the proposed CMTNet achieves a sensitivity of 87.20% at 99% specificity, with an overall test classification accuracy of 98.79%. The proposed CMTNet achieves the highest TPR and lowest EER compared to existing algorithms.
We further address the problem by predicting the level of severity of infection in the lungs by using the longitudinal-multimodal medical data for the subsequent days, for monitoring the progression of the infection, optimized treatment, prediction of the ventilator support, and morbidity reduction. Therefore, our main goal is to predict the severity of the infection in lungs due to COVID-19 for subsequent days given the data for current and past days. To predict the level of severity of the infection, we train a deep auto encoder-decoder neural network for learning a lower dimensional representation of the multimodal data (X-Ray and CT images and Clinical Measurements) of a patient. The proposed network generates a compressed representation of the input data which is fed into another branch of the network (classifier) to predict the infection levels. Then, by using the compressed representations of the data of previous days, we train a Long Short Term Memory (LSTM) based sequence to sequence the prediction network to obtain the representations of subsequent days. This allows us to predict the severity of infection in the lungs for subsequent days given the data for current days. We achieve an accuracy of 79% for classifying the severity of lung infection into four categories: Normal, Ground Glass Opacities, Consolidation, and Pleural Effusion.
The design of a face shield to protect health workers from Covid-19 was conceived by the Automated Manufacturing research group at IIT Jodhpur using Computer Aided Design (CAD) tools. The designers printed a few prototypes in the initial stage to examine the effectiveness of usage, ergonomic comfort, feasibility of indexing, ease of manufacturing and assembly etc. The discussion level prototypes were printed using the FDM 3-D printing facility at the institute. Based on multiple brainstorming sessions, the designers arrived at the final version shown in Figure 1. Subsequently, 50 such prototypes were fabricated using the 3-D printing facility within a span of 3 days. The prototype design consists of three major parts; an indigenously designed fixture (Fig.1); transparent plastic sheet (typical thickness in the range of 0.3-0.5mm, Figure 2); an adjustable band to fix the face shield on the forehead ex. velcro, elastic band etc. These prototypes were supplied to Jodhpur district COVID-19 administration for user feedback and assessment. An important point to note is that the fixture is reusable with replaceable transparent plastic sheets easily available in the market.
As face shields were required in huge quantities by the district administration in a short span of time, 3-D printing was not a time efficient option for mass production. The group explored the injection molding process (available with most plastic manufacturers across the country) as a commercially viable option. The group worked jointly with M/s. Iscon Surgicals Ltd., Jodhpur, one of the leading surgical equipment manufacturing industry under Jodhpur City Knowledge and Innovation Cluster initiative of Government of India to transform this prototype to a commercially viable solution. The group actively participated in the development of the injection molding dies and necessary tooling for the product. The dies were machined using high speed CNC machining and CAM solution. Upon transfer of technology to the collaborating industry during the first week of April 2020, the necessary tooling was developed. The final product with necessary accessories and packaging was developed by end of April, 2020 and it was launched in the market.
Taking cognizance of paucity of alcohol-based hand sanitizers in the market due to the sudden outbreak of COVID-19 pandemic, our team consisting of faculty, Ph.D. Students, technical staff members and the deputy librarian, in consultation with the Institute’s Medical Services Committee, took up the responsibility of in-house preparation of alcohol-based hand sanitizer for the campus community. The components used for preparing this hand-sanitizer are Isopropanol, Aloe Vera extract, Glycerine, Hydrogen Peroxide, Distilled water and Fragrance.
This work was started in the month of March 2020 and the first phase of distribution took place on 21 March 2020. The preparation is being continued and is supplied to all Offices and Departments on a regular basis. Besides this, Covid-19 related hygiene awareness posters were also prepared by this team and posted at various places in the campus. The team has also prepared Standard Operating Procedures (SOP), in consultation with various stakeholder groups for planning and executing the operations of various entrances, Offices and Student arrival on campus.
There is a growing need for antiviral and antibacterial painting in hospitals and public areas for effective prevention of the spread of infection. To meet this growing demand, nanotechnology has been explored by different research groups and industries. Silver nanoparticles embedded paint was demonstrated as biocidal coatings for decontamination of surfaces in hospitals. HeiQ (A Swiss Technology firm) has developed vesicle and silver based coating technology for reducing virus infections including human coronavirus in year 2020 during the outbreak of COVID-19 (SARS-CoV2). Another promising approach in this regard could be the use of reactive oxygen species (ROS) producing nanoparticles. It has been demonstrated that carbon quantum dots (CQDs) exhibit antiviral activity to human norovirus virus-like-particles by generating ROS. This project aims to develop metal ion-decorated carbon quantum dots (M-CQDs) based antiviral paint. Upon light activation (preferably room light), M-CQD will produce highly reactive oxygen species (ROS) on the surface of the walls. That ROS will damage the viral proteins leading to the effective disinfection/decontamination of walls/surfaces. Presence of selective metal ions in M-CQDs will enhance the trapping of the virus particle on the wall and will facilitate the virus damage/ degradation by bringing the virus to the close proximity of the ROS source. The advantages of using M-CQDs are (i) its superior light absorbing capabilities; and (ii) highly resistant to photodegradation/photobleaching, so that it can produce ROS over a significant period of time.
The increased number of Covid-19 patients in the various parts of the country poses challenges for medical professionals related to focussed health monitoring and care for critical cases. IIT Jodhpur, AIIMS Nagpur, and IIIT Nagpur have indigenously designed and developed a useful prototype for active tracking and monitoring of COVID positive and suspect patients. The device is in the form of a smart wristband designed to overcome the limitations of existing mobile-based applications used for tracking and monitoring. The existing solutions depend on the continuous use of mobile phones by COVID-suspects/patients and the availability of stable internet connection. The mobile applications can track the movement of an individual but cannot monitor specific symptoms as it is performed based on subjective self-assessment of the user. The smart wristband, named as COVID-19 tracker by the team, can provide objective and reliable data of vital parameters such as temperature, pulse rate, respiratory rate, and oxygen saturation so that the quarantined person will get a health alarm seeking early medical help. The device can also provide mobile-free operation using a geofencing technology, which will be able to provide a real-time alert on any breach in the quarantine zone. In addition to concept development, indigenous structure and body parts of the device were designed by the Automated Manufacturing research group at IIT Jodhpur, considering ergonomic factors, the effectiveness of application, compactness, tamper-proof design, etc. AIIMS Nagpur validated the device for its specificity and effectiveness. The team is currently augmenting a full-scale prototype with additional features and in discussion with industrial collaborators for the development of the device for mass-scale deployment. The RAKSHAK scheme launched DST, Govt. of India is supporting this joint research work of three institutes.
Overall, IIT Jodhpur has worked as a team in these unprecedented times using cutting edge research and technology to offer various solutions to deal with COVID 19.