Building Trustworthy AI for Clinical Use: A Start-Up Success Story

Aisthesis Medical – AIMED, a pioneering MedTech startup, is at a reasonable level of digital maturity compared to other European SMEs, especially if you take into account, they are relatively new. However, during the assessment of its digital maturity we identified several areas of improvement. The following needs were identified: 

• Need to receive funding

• Need to increase the level of technological readiness

• Harmonisation with current copyright regulations

• Lack of universal data collection protocol

• Lack of continuous training of staff

• Reduction of the company’s environmental impact

The results from the first DMA highlighted the importance of focusing on the following areas:

• Copyright protection: it would be important to safeguard the intellectual property of its software, as well as its financial and moral rights. In addition, the company will have to register, if it has not already done so, its trademark with the Trademark Directorate of the General Secretariat of Trade and Consumer Protection of the Ministry of Development & Investment.

• Information/Training of Staff: AIMED staff should keep abreast of recent relevant technological developments through regular training programs, collaborations with academic and research institutions, and through participation in scientific conferences.

• Universal Data Collection Protocol: it is recommended that the company establishes a Universal Data Collection Protocol to collect data from various institutions and open databases.

• Maturation of Services/Products: the company should do so through seamless monitoring of market trends in the sector in which the company operates, and continuous adjustment of the company's growth plan on an ongoing basis.

However, in addition to the assistance required to increase AIMENT’s digital maturity indicators, the company was in need of recommendations and suggestions regarding their technical necessities, in order words they required help improving the technologies they provided.

The aim of the collaboration between AIMEN and the Technical University of Crete, was to on one side assess the company’s digital maturity level, and on the other, inform them about the latest market trends and make suggestions for their further development. Additionally, the TUC established particular guidelines to develop a methodology for prediction in clinical. 

Thus, an action plan was developed to help incorporate algorithmic innovations to achieve accurate diagnosis. The services provided during this collaboration are divided into two categories: a) Data Preprocessing and b) Innovative Techniques in Machine/Deep Learning. 

Data Processing – this service involved following current regulations to manage the data and pre-process in order to reduce or even eliminate the noise introduced by abnormal activities, all while delivering the useful information. Moreover, it dealt with issues of data anonymization and de-noising, missing values and generation of synthetic data, and data fusion. It is important to notice that, the data preprocessing stage is a key step prior to further analyses, ensuring the production of reliable and interpretable results in the field of medicine. 

Innovative Techniques – this service provided an extensive analysis of recent widely used methodologies. These methodologies revolved around innovative machine learning and deep learning algorithms. In this context, the following results were expected: prediction over inputs of a multimodal nature, automated machine learning techniques, parameter and hyperparameter optimization, specific interpretable AI approaches, and data auto-labelling techniques. 

Additionally, the rationale behind the required investment to finance the proposed solution is based on the AIMED’s financial benefits and returns already achieved, as well as those expected from it from its implementation. Furthermore, the development and application of the proposed methodology means a reduction in the use of hardware resources. 

The main benefits that the company has already seen from this collaboration relate to the information provided by TUC about recent market trends and the completion of the questionnaire regarding their digital maturity level. Benefits that are to be expected soon include:

• Copyright protection

• The company’s personnel will be provided with up-to-date information and receive the essential trainings in order to keep them informed about new technological developments related to the services and products provided by the company. The staff will have the opportunity to expand their knowledge by attending scientific conferences, cooperating with academic and research institutions and completing additional training programs. 

• The development of a universal data collection protocol will contribute to the secure and rapid collection of data from various institutions and open databases. The objective is to achieve the collection of a wealth of data for analysis in a shorter period of time, thus pushing the company’s momentum.

• AIMED will achieve market consolidation through the creation of competitive products and services, and by maintaining constant dedication to this goal and keeping abreast of current trends.

In addition, the Technical University of Crete created specific guidelines to achieve AIMED’s objectives and to ultimately give them the tools to effectively and accurately predict sepsis. In this context, the company received consulting services in the form of two deliverables, with detailed descriptions of modern data preprocessing techniques and methodologies and innovative machine and deep learning techniques. The overall funding provided was close to twenty thousand euros which allowed the company to train their personnel in very advanced AI topics, protect intellectual property and use advances for medical monitoring and diagnosis.