Technology in Motion: FAIR data for Health Technology using computer imaging methods

Sil­via Pin­tea is a Post­doc researcher at the Fac­ul­ty of Elec­tri­cal Engi­neer­ing, Math­e­mat­ics and Com­put­er Sci­ence at Delft Uni­ver­si­ty of Tech­nol­o­gy (TU Delft). She spe­cial­izes in the fields of arti­fi­cial intel­li­gence and com­put­er vision. In a recent project ‘Tech­nol­o­gy in Motion’ Pin­tea inves­ti­gat­ed the use of com­put­er imag­ing as a means to mea­sure tremors in patients that would enable the devel­op­ment of an instru­ment for the diag­no­sis of Alzheimer patients.The goal of using com­put­er vision meth­ods is to bypass the use of sen­sors to diag­nose patients, which can be an uncom­fort­able pro­ce­dure. 

Sil­via recent­ly fin­ished a research project with col­leagues from Lei­den Uni­ver­si­ty Med­ical Cen­ter (LUMC) where all the data col­lec­tion (i.e. sen­sor data and videos) was con­duct­ed. LUMC pro­vid­ed sup­port regard­ing the required eth­i­cal research pro­ce­dures and ensur­ing secure man­age­ment and pub­li­ca­tion of the data. She was great­ly inter­est­ed to pub­lish the datasets behind the result­ing aca­d­e­m­ic pub­li­ca­tion, while of course safe­guard­ing patients’ iden­ti­ties. 

“Peo­ple pub­lish claims that they dis­cov­er a method to mea­sure this and that, but they do not pub­lish the data and if you con­tact them, often they can’t share the data because of IPR or pri­va­cy, etc. But, in our field it is very use­ful to have access to the data to test your meth­ods, soft­ware, etc. There­fore, I think is impor­tant to make avail­able as much data as pos­si­ble.”

Sil­via Pin­tea

To achieve this, Pin­tea invest­ed a con­sid­er­able amount of time and effort to anonymize the video record­ings in accor­dance to the agree­ments made with patients as well as in accor­dance with the require­ments of the LUMC eth­i­cal com­mit­tee. Despite the lev­el of effort this required, she was com­mit­ted to pub­lish­ing the dataset in accor­dance with the FAIR prin­ci­ples to ensure that the datasets would be find­able, acces­si­ble, inter­op­er­a­ble and repro­ducible.

“We wouldn’t be able to pub­lish the paper with­out mak­ing the data avail­able. In the field of com­put­er vision, the norm is the abil­i­ty to com­pare results with oth­ers. But, out­side the tech­ni­cal dis­ci­plines (e.g.medical research), there is no tra­di­tion of shar­ing data. Prob­a­bly a lot of data exist already, but is not pub­lished. If I had start­ed with an exist­ing dataset, I would have saved time.” 

Pin­tea is now focus­ing on her next research project — a col­lab­o­ra­tion with col­leagues from the petro­le­um explo­ration and extrac­tion indus­try to mea­sure seis­mic imag­ing. In keep­ing with her com­mit­ment to curat­ing FAIR research prac­tices and datasets, she hopes to con­vince her research part­ners to pub­lish the datasets gen­er­at­ed in the project.

“It is great to have a solu­tion like 4TU.ReseachData at hand for pub­lish­ing your datasets. It is very use­ful to pre-reserve a DOI for your data and include it in your paper before send­ing it for review. Peo­ple know that the data will be avail­able.”

Link to the dataset on 4TU.ResearchData.
Cov­er image by Raeng_ via Pix­abay