Periodic Reporting for period 1 - iToBoS (Intelligent Total Body Scanner for Early Detection of Melanoma)
Période du rapport: 2021-04-01 au 2022-09-30
Melanoma is one of the most aggressive cancers that can be discovered at an early stage, and it is responsible for 60% of lethal skin neoplasia. Its incidence has been increasing in white population and could become a public health challenge because of an increase in life expectancy of the elderly population. Total body skin examination, the primary screening mechanism for melanoma, checks each pigmented skin lesion individually in search of typical melanoma signs. This can be a very time-consuming technique for patients with atypical mole syndrome or a large number of naevi.
iToBoS aims at developing an AI diagnostic platform for early detection of melanoma. The platform includes a novel total body scanner and a Computer Aided Diagnostics (CAD) tool to integrate various data sources such as medical records, genomics data and in vivo imaging. This approach will lead to a highly patient-tailored, early diagnosis of melanoma. The project will develop and validate an AI cognitive assistant tool to empower healthcare practitioners, offering a risk assessment for every mole. Beyond integrating all available information about the patient to personalise the diagnostic, it will provide methods for visualising, explaining and interpreting AI models, thus overcoming the “black box” nature of current AI-enabled CAD systems, and providing dermatologists with valuable information for their clinical practice.
iToBoS aims at developing an AI diagnostic platform for early detection of melanoma. The platform includes a novel total body scanner and a Computer Aided Diagnostics (CAD) tool to integrate various data sources such as medical records, genomics data and in vivo imaging. This approach will lead to a highly patient-tailored, early diagnosis of melanoma. The project will develop and validate an AI cognitive assistant tool to empower healthcare practitioners, offering a risk assessment for every mole. Beyond integrating all available information about the patient to personalise the diagnostic, it will provide methods for visualising, explaining and interpreting AI models, thus overcoming the “black box” nature of current AI-enabled CAD systems, and providing dermatologists with valuable information for their clinical practice.
On the first 18 months, the project has advanced in several fronts. First, the overall architecture behind the iToBoS concept has been defined, specifying the system requirements and ensuring interoperability between each of the main modules of the project. This definition included not only the algorithms used for image processing, 3D reconstruction, lesion detection and classification, data integration and anonymization, but also the tools for applying machine learning to the gathered data and the full body scanner hardware to collect the skin images. Once of the main advances of this first period is the definition and validation of the high-resolution imaging module (HRIM), including the design of the liquid lenses for image acquisition. We have also advanced in the design and integration of this HRIM in the total body scanner. In this sense, a first prototype of the total body scanner was designed and the development of a prototype arch was completed (with some parts still pending to arrive by the end of the first reporting period, due to lack of components in the market). Given the materials shortage in the market, and the delays that this imposed in the design of the scanner prototype, a contingency plan has been implemented to keep advancing in the development of the AI cognitive assistant for early detection of melanoma.
This contingency plan is based on the acquisition of skin images using the best total body scanner that could be found in the market (VECTRA360 by Canfield). The team started to develop the tools for image ingestion, as well as the anonymization and masking tools both for the images and the patient data. In this sense, we have defined the specific demographic and clinical data from the patients to be used by the AI, together with the lesion images. Secondly, a number of tasks such as lesion detection, lesion classification and development of eXplainable AI (XAI) have started by using public image datasets, allowing the advance in these domains. Moreover, both ethical risks as well as ethical opportunities, such as being able to detect melanoma earlier or personalizing the diagnostics, have been assessed during this first period. All the medical research ethics aspects have been addressed and the clinical protocols for the use of the VECTRA images and the patient’s data have been submitted to the corresponding ethical committees. Finally, a patient engagement plan with touch points and educational needs has been developed, specifying how iToBoS will engage with the broader patient community. And regarding outreach, the consortium has defined the dissemination and communication strategy of iToBoS to set the guidelines, actions and tools to channel the efforts aimed at achieving a wide impact of the project among the target audience to extend the results and benefits of the project to the society in general.
This contingency plan is based on the acquisition of skin images using the best total body scanner that could be found in the market (VECTRA360 by Canfield). The team started to develop the tools for image ingestion, as well as the anonymization and masking tools both for the images and the patient data. In this sense, we have defined the specific demographic and clinical data from the patients to be used by the AI, together with the lesion images. Secondly, a number of tasks such as lesion detection, lesion classification and development of eXplainable AI (XAI) have started by using public image datasets, allowing the advance in these domains. Moreover, both ethical risks as well as ethical opportunities, such as being able to detect melanoma earlier or personalizing the diagnostics, have been assessed during this first period. All the medical research ethics aspects have been addressed and the clinical protocols for the use of the VECTRA images and the patient’s data have been submitted to the corresponding ethical committees. Finally, a patient engagement plan with touch points and educational needs has been developed, specifying how iToBoS will engage with the broader patient community. And regarding outreach, the consortium has defined the dissemination and communication strategy of iToBoS to set the guidelines, actions and tools to channel the efforts aimed at achieving a wide impact of the project among the target audience to extend the results and benefits of the project to the society in general.
The overall iToBoS system, including the full-body scanner, and its associated AI Cognitive Assistant, is in itself an advance beyond the current state of the art. This unprecedented system, which integrates dermoscopic images with information related to the patient’s phenotype, genotype and family history, will be able to automatically generate a detailed, structured and informed health status assessment of the patient, based on the proposed personalised holistic melanoma risk score model. The iToBoS system will enable an increased precision of melanoma diagnosis, contributing to improved clinical decisions that will facilitate its daily use in clinical practise thanks to its intelligent human-computer interface.
The unprecedented system proposed by iToBoS will lead to an increase in the EU’s capacity to innovate in the field of melanoma diagnosis, achieving quicker, more encompassing and highly personalised diagnostics.
The reduction in the time required to perform a full body exploration of the skin of the patient will be substantial, with expected acquisition time of a scan at 6 minutes, as compared to the 40-50 minutes required for explorations using common contact dermoscopy. The new diagnosis tool to be developed is expected to increase the melanoma detection rate by, at least, 20%. The personalized diagnostic and associated treatment generated with the tools and algorithms developed in iToBoS will reduce the costs associated with long term treatments. At the same time, it is expected to increase the 5 year survival rate from less than 23% to 98%.
The international positioning of iToBoS is validated by the different nationalities not only from Europe, but also from Australia and Israel, and with an external Advisory Board also covering the USA market. The outcomes of iToBoS will be disseminated with the publication of 40 scientific publications in peer-reviewed journals and 30 conference-related publications. iToBoS will participate in at least 50 scientific, medical or commercial events and organize international challenges. We expect at least 4 other skin-related diseases to benefit from the total body skin scanner, opening the door to other markets within the medical area.
iToBoS will contribute to fulfilling the social right to explanation and the transparency requirements of the EU’s Ethics guidelines for trustworthy AI (April 2019). A huge effort will be devoted to pushing forward explainable AI. Explainability in the AI to be developed within the project is crucial to presenting the decisions taken by the system in a way that can be both understood and followed by humans. Due to the AI systems to be developed in iToBoS having a significant impact on the diagnosis of a patient, the project will develop suitable explanations for the AI decision-making process adapted to the expertise of the dermatologist.
There are 10 companies partnering the project, 6 of which are SMEs, that will greatly benefit from its outcomes, extending their expertise in the healthcare sector and reaching new potential market niches. A great deal of dissemination will be devoted to reaching SMEs outside the project.
The unprecedented system proposed by iToBoS will lead to an increase in the EU’s capacity to innovate in the field of melanoma diagnosis, achieving quicker, more encompassing and highly personalised diagnostics.
The reduction in the time required to perform a full body exploration of the skin of the patient will be substantial, with expected acquisition time of a scan at 6 minutes, as compared to the 40-50 minutes required for explorations using common contact dermoscopy. The new diagnosis tool to be developed is expected to increase the melanoma detection rate by, at least, 20%. The personalized diagnostic and associated treatment generated with the tools and algorithms developed in iToBoS will reduce the costs associated with long term treatments. At the same time, it is expected to increase the 5 year survival rate from less than 23% to 98%.
The international positioning of iToBoS is validated by the different nationalities not only from Europe, but also from Australia and Israel, and with an external Advisory Board also covering the USA market. The outcomes of iToBoS will be disseminated with the publication of 40 scientific publications in peer-reviewed journals and 30 conference-related publications. iToBoS will participate in at least 50 scientific, medical or commercial events and organize international challenges. We expect at least 4 other skin-related diseases to benefit from the total body skin scanner, opening the door to other markets within the medical area.
iToBoS will contribute to fulfilling the social right to explanation and the transparency requirements of the EU’s Ethics guidelines for trustworthy AI (April 2019). A huge effort will be devoted to pushing forward explainable AI. Explainability in the AI to be developed within the project is crucial to presenting the decisions taken by the system in a way that can be both understood and followed by humans. Due to the AI systems to be developed in iToBoS having a significant impact on the diagnosis of a patient, the project will develop suitable explanations for the AI decision-making process adapted to the expertise of the dermatologist.
There are 10 companies partnering the project, 6 of which are SMEs, that will greatly benefit from its outcomes, extending their expertise in the healthcare sector and reaching new potential market niches. A great deal of dissemination will be devoted to reaching SMEs outside the project.