Community Research and Development Information Service - CORDIS

  • European Commission
  • Projects and Results
  • Final Report Summary - AUTOFLOW (Automation of Flow Cytometric Analysis for Quality-Assured Follow-up Assessment to Guide Curative Therapy for Acute Lymphoblastic Leukaemia in Children)

Final Report Summary - AUTOFLOW (Automation of Flow Cytometric Analysis for Quality-Assured Follow-up Assessment to Guide Curative Therapy for Acute Lymphoblastic Leukaemia in Children)

Acute Lymphoblastic Leukaemia (ALL) is the most frequent leukaemia entity in children and adolescents. Flow Cytometry (FCM) to analyse Minimal Residual Disease (MRD) is one of the methodologies most useful in this respect, because it is widely available and applicable to most patients.

Throughout project, AutoFLOW achieved its goals on two paths:

1) The first and fundamental objective continued to be the career development of the participants:
All fellowships were implemented smoothly and according to the DoW (with minor deviations in agreement with REA). Three recruits have successfully fulfilled their allocated parts and a motivated and ambitious team of active and former fellows (altogether 17) took part in the conjoint efforts AutoFLOW.
This included the training and development of the project fellows and a close interaction and communication between the fellows, the HOMEs, the HOSTs, all involved AutoFLOW team, as well as for the involved institutions, their employees, students, etc. The training contents are related to project associated expertise as well as soft skills like presentation, language, interdisciplinary work processes, reporting, time and work management, intercultural training, teaching, etc. So far, altogether 30 single regularly ongoing workshops throughout the project.

To spread the word and make AutoFLOW truly interactive and visible, we covered a broad array of dissemination activities according to the DoW and beyond, outreach events, conference participation and hosting, media publications, workshops for the project collaborators and various external audiences, e.g. pupils, medical and ICT students, the interested public, expert and scientific consortia, etc. (for details cf. section on Dissemination below).

2) Secondly, the following scientific objectives of AutoFLOW could be implemented:
1. Develop and validate a pipeline-software package as a ready-to-use FCM-based MRD-assessment tool for the daily clinical practice for patients with ALL: FCM-MRD assessment is a cornerstone in risk stratification and treatment selection in several current paediatric ALL therapy regimens.
2. Augment correct assessment of FCM-MRD in samples which are difficult to analyse by conventional operator-based gating (e.g. ≥8-color applications); computer-aided analysis potentially has a superior resolution due to utilization of the whole multi-parametric FCM-data space at once instead of step-wise, two-dimensional plot-based visualization.
3. Reduce subjectivity caused by manual operator gating; increase result comparability and reproducibility through automation and standardization.
With the successful finalisation of WPs 2&3 a robust algorithm for the FCM-MRD classification was developed and integrated in an effective pipeline. The procedure works machine-independent and is able to detect also small MRD proportions. In a subsequent evaluation step, the pipeline was successfully validated (cf. Deliverables D2.1 & D3.1). The processing pipeline was then integrated into deployable and user-friendly telemedical software. AutoFLOW ensured maximal data security through its user management system (cf. Deliverables D4.1).
All stages were presented to the scientific community through publications and international conferences (cf. list of scientific publications below).

Description of the work performed since the beginning of the project and main results
In its four years duration, AutoFLOW could establish an intense and productive collaboration between all partner, fellows and beyond and set up a closely-knit net of fellow cooperation and transfer of knowledge. The development is progressing according to schedule and the scientific development of AutoFLOW generated the following highlights:

a) The AutoFLOW Database:
The setup of a quality controlled, annotated, and class labelled MRD-FCM measurements from 240 patients with acute lymphoblastic leukaemia (expected in Annex 1: 200 patients). The patient cohort represents the typical incidences of ALL immunophenotypic subtypes. Measurements at three different time points during treatment per patient, with an overall sample data base of 624 sample, allowing us to cover sample variations like phenotypic shifts and occurrence of normal regenerating bone marrow. Above that, the risk of insufficient separation of MRD from background could be minimised by switching to a new dual-tube staining panel. The database with highest data protection standard (and conformity to the new GDPR) is structurally designed for continued file input for further improvements (cf. Deliverable D1.1).

b) The AutoFLOW algorithm for reliable, machine-independent detection of MRD:
Based on testing different state-of-the-art approaches, AutoFLOW developed a novel supervised, parametric method for identifying complex non-Gaussian clusters corresponding to cell populations. The methods have been benchmarked on our project dataset and experiments showed that the proposed holistic approach outperforms the other ones due to its adaption of the model (cf. Deliverable D2.1).
AutoFLOW created an adapted pipeline and enlarged the dataset by 83 new MRD samples from leukaemia patients diagnosed at Charité Berlin to evaluate our algorithm for automated classification of FCM read-outs (as achieved on Labdia data). With the quality assessment analysis related to non-biological variations, which serves as a feasibility study for a domain adaptation purpose, a milestone of evaluating an automated ALL FCM-MRD analysis tool could be reached. The results of the tests of this multi-device, -country, -lab scenario prove that AutoFLOW can establish device independence of up to 85%. Also, tests showed that erroneous classification events are consistent over the tested machines and can thus be attributed to errors in a) the data or b) the algorithm (cf. Deliverable D1.3).

c) AutoFLOW automated FCM-data viewer and analysis tool:
The fully automated FCM-data analysis prototype software was designed and implemented, validated and optimised (cf. Deliverable D1.4). These include a web-based (FlowONLINE) and a stand-alone desktop-based (FlowVIEW) clinician software, an administrative software for the user management (FlowADMIN), a number of databases and a secure telematics platform on which the entire system is hosted. The backend module safely but flexibly manages the AutoFLOW data and interacts with the front-ends, as well as with future new data inputs. With a hierarchical user management for test-users up to clinical professionals who can also influence the algorithm with their inputs and communication interfaces. The secure system enables dissemination of automated FCM-MRD analysis to medical centres with less expertise for the benefit of an even larger community of diseased children worldwide. FlowONLINE and FlowADMIN can be tested at, FlowVIEW at

AutoFLOW succeeded to fulfil its expected general objectives:
With its stand-alone tool (FlowVIEW) for in-depth, real-time automated gating and visualisation of FCM-MRD samples and the compact online version (FlowONLINE), a whole new horizon of ALL analysis could be established, which will also support less experienced medical centres to perform the diagnostic procedures. AutoFLOW managed to develop a highly developed prototype for objective and automated multi-parameter FCM data analysis with specifically reliable MRD quantification. With AutoFLOW, quality-assured MRD-assessment throughout the international-BFM FCM-network will be improved and standardised for the benefit of children with ALL in many countries. This means that the highly demanding task of gating and MRD assessment will in the future be able to be accessed by an exponentially wider range of labs and medical doctors than today. Much working time, effort, lab costs and training can be saved, subjectivity in diagnosis will be reduced and most importantly, adequate treatments can be issued even in more remote or less trained areas of the world (cf. the work of our Marie Curie Ambassador, fellow C. Takenga, with hospitals in the Republic of Congo (where not (only) the Flow machines, but foremost the training of qualified staff is missing).
With several follow-up initiatives, the AutoFLOW prototype will be optimised to become a medically approved decision making tool that will revolutionise the handling of Flow data in ALL and other leukaemias (and perhaps FLOW data in general.
Above that, AutoFLOW is creating awareness and interdisciplinary, inter-sectoral, international, network wide, and beyond exchange of ideas, approaches – and most of all collaborations and solutions in the field of medical computing, broadening both the horizons of academics and industry partners involved, as well as clinical, computing, medical, logistical, and ethical experts, having come together in this project.


Reported by



Life Sciences
Follow us on: RSS Facebook Twitter YouTube Managed by the EU Publications Office Top