SDG 3: Good Health and Well-being
2. Project Details
Company or Institution
General description of the AI solution
SoftMining has developed semantic analysis methods of wide applicability. One of these methods, co-funded by the University of Salerno and EIT Health, allows the real-time and full-data analysis of electrocardiograms. Unlike traditional feature extraction methods, SoftMining’s (SM) ECHealth method uses all acquired data to compute a distance between the current state and a reference state. The reference can be the resting state or pathological state of the patient. The analysis can be conducted in real-time and processed from a smartphone. The particular ECHealth application addresses the evaluation of clinical trials. The market for medical devices is steadily growing, as are the costs associated with clinical trials of new drugs or vaccines. As a result, we have had numerous contacts with pharmaceutical companies interested in the approach and have been invited to numerous conferences to present the approach’s advantages. Patients undergoing clinical trials are often unable to express the effects of a substance on their body unambiguously: sometimes because of expression difficulties, other times because the effect is not manifest. ECHealth allows detecting these effects before the patient can recognize them.
Our technology has already been tested with the contribution of EIT Health and the University Hospital San Giovanni di Dio e Ruggi d’Aragona (Salerno). A certification path as medical device has been undertaken in collaboration with Sitem Insel of Bern (CH). We own the technologies and algorithms, and some of them are in the patenting process.
Excellence and Scientific Quality: Please detail the improvements made by the nominee or the nominees’ team or yourself if your applying for the award, and why they have been a success.
The SM team started to work in the semantic analysis field in 2014 (Piotto, Stefano, et al. Journal of computer-aided molecular design 30.9 (2016): 753-759, by exploiting these approaches in the biological field, particularly addressing the docking problem. Later, proteome and genome analysis were exploited thanks to the definition of a MEF-based system . At first, commons AI tools and techniques were used, particularly neural networks (NN) related to data mining, pattern recognition and signal processing. To face the intrinsic computational difficulties behind artificial intelligibility and to make the approach as too general-purpose as possible, the SM team start to define and design a new type of NN, reconsidering from scratch the concept of artificial neuron and the neural network itself. The research group started to design an Associative Knowledge-Based Multi-Layer Neural Network built on the hypothesis that each observable object can be expressed by the expressed features that we consider a high-level abstraction of the internal object structure. Thanks to this new NN architecture, SM developed a general-purpose framework to analyze heterogeneous kinds of objects.
These tools allow researchers and private companies to look for relations, correlations, and hidden data relationships. For Proteome and Genome analysis, SM developed a tools named Protcomp (TRL 6). For text-based objects, the SySa tools are available at TRL 8. The ECG analysis tool was cofounded by EIT Health and available for researchers and clinicians (TRL 6). The analysis was general and innovative and finally led to the design of two novel classes of antibiotics, currently patent pending.
These tools are deployed into an integrated platform in the SoftMining cloud space, supported and contributed by the AWS program, that allows researchers and private companies to use SM technologies without the need to design and build their infrastructures. SM Unified platform is currently at stage TRL 8..
Scaling of impact to SDGs: Please detail how many citizens/communities and/or researchers/businesses this has had or can have a positive impact on, including particular groups where applicable and to what extent.
The method we have developed, the training of convolutional neural networks with data from semantic analysis, is innovative and of general applicability.
The application of ECHealth to clinical trials will have important economic benefits, due to a better evaluation of drug candidates, and social benefits due to increased safety of the solutions adopted. The use of this AI can be extended to individuals who want to make, themselves, a personalized analysis of their medical data, without other people or private companies ever coming into contact with sensitive data.
The effect of this technology will perhaps be even greater in developing countries. We have been invited to participate in two NIH (US) projects to apply these approaches in Ghana for infectious disease containment.
The SM Platform (SMP) aims to provide all the SM knowledge and tools in a SaaS way that means that each SDG stakeholder can combine SM tools to address their scientific problems without the need to build an infrastructure or hire AI specialists into organics. In particular, health-related startup and national health system can speed up their migration to AI systems relying on a well-tested platform instead of starting from scratch. Furthermore, the SMP is designed to work with real-time and offline data, particularly social simulator techniques and AI-powered Agent-base simulation. In the UNESCO context, this might potentially help the government (in particular health agencies) be driven by data and not only by offline observation (survey, exit poll, others offline tecniques).
Scaling of AI solution: Please detail what proof of concept or implementations can you show now in terms of its efficacy and how the solution can be scaled to provide a global impact ad how realistic that scaling is.
Softmining was accepted into Amazon Accelerator Program. Then, it was able to access the full Amazon AWS infrastructure by using credit from Amazon itself. The SM platform relies on Amazon AWS cloud infrastructure that provides QoS and scalability on-demand and on-premise. In SMP, each stakeholder request is represented by a JOB. Moreover, the SMP handles two Queue systems to provide different QoS. In case of high system load, the SM orchestrator and SM provisioning agent instance new slave to allow the queues to be free as fast as possible.
If chosen from governments, SMP can be used as an underlying system (back-end) to build health and research systems. Also, the SaaS capabilities can allow thirdy party applications to exploit the SMP system, allowing vendors to build new AI-based applications, especially in Digital Health and Telemedicine. The Medical branch of IoT can be the primary stakeholder of SMP thanks to the simple cloud, fog, edge networks. For example, a wearable system can record ECG that can be sent to SMP to detect if a problem in a cardiac signal can be found. More advanced, multiple ECG recordings of the same patient can be processed by SMP, looking for the correlation between signals acquired and known disease.
The rise of AI seems to be unstoppable, and providing a simple way to utilize the power of these tools might be the next killer application that can move humanity into a real AI era. SMP can act as a catalyzation to these reactions, allowing these companies to go on the market quickly.
SMP is currently in a dealing process with two pharma companies.
Ethical aspect: Please detail the way the solution addresses any of the main ethical aspects, including trustworthiness, bias, gender issues, etc.
AI technologies are ethical and sustainable. In almost all of our applications we strictly avoid using sensitive data. When, as in the case of the ECHealth project for monitoring patients engaged in clinical trials, personal data is used, we take great care to reduce risk. For example, SoftMining uses an outside company to certify methods and relies on a regulatory body to get the green light.
I metodi sviluppati sono anche offerti alla comunità internazionale mediante pubblicazioni su riviste internazionali (Piotto S, Di Biasi L, Marrafino F, Concilio S. Evaluation of epidemiological risk using contact tracing open data. Journal of Medical Internet Research. 2021; Piotto, Stefano, et al. “GRIMD: distributed computing for chemists and biologists.” Bioinformation 10.1 (2014): 43)
Ethical aspects are crucial for us and for this reason we participate in two international projects on ethical aspects of AI.
SoftMining does not discriminate on the basis of gender, religion, ethnicity or any aspect other than skills and motivation. To ensure maximum transparency all staff interviews are video-recorded and made available for possible evaluation. During the last year we hired a Machine learning expert (from India, male), and a chemist (female), and we hired 3 ML designers (from Italy, 2 female, 1 male). SoftMining is an Affirmative Action and Equal Opportunity Employer. All qualified applicants will receive consideration for employment without regard to race, color, religion, sex, sexual orientation, gender identity, age, national origin, or protected veteran status and will not be discriminated against on the basis of disability.