PALERMO (ITALPRESS) – Finally a “positive” study: AI does not replace reality, but can help to read it better when designed, controlled and validated with rigor. The study, carried out by a research group of the IRCCCS Ospedale Galeazzi-Sant’Ambrogio in Milan with the methodological and computational contribution of Marco Giacalone and Davide Lamartina, was published on European Spine Journal and is indexed on PubMed.
The research involved the use of a computational method assisted by artificial intelligence to expand and analyze a small clinical sample, always keeping reference to real data. Starting from 123 real asymptomatic subjects, a biologically plausible synthetic dataset of 10,000 cases was built, used to make anaemic correlations of the spine more stable.
The most important part was not to generate synthetic data, but to demonstrate that the correlations identified in the expanded sample remained verifiable on real data through independent statistical procedures.
The research stems from the need to study spinopelvic alignment in asymptomatic subjects, an essential element to understand physiological sagittal balance and define useful reference values in the evaluation of spinal deformities.
The availability of data on healthy subjects is, however, limited by ethical, logistic and related constraints to radiological exposure: it is not in fact simple, nor always appropriate, to subject healthy people to radiographic examinations for the sole purpose of expanding a sample of research. In this context, controlled generation of synthetic data can be a useful strategy for exploring anatomical relationships without losing reference to real data.
Starting from 123 real asymptomatic subjects, a synthetic dataset of 10,000 biologically plausible anatomical configurations was created. The correlations identified in the expanded sample were then verified on the 123 real cases through statistical bootstrap, a procedure that allows to evaluate the robustness of the results repeatedly returning to the original data.
For the study, X-rays were analyzed in an upright position of the entire spine, recording demographic characteristics and multiple spinopelvical parameters, including pelvic incidence (PI), pelvic inclination (PT), sacral slope (SS), lumbarous gross (LL), thoracic ciphers (TK) and cervical alignment measures. Methodologically, a probable Gaussian resamping approach was used, driven by anatomical and biological constraints.
The correlations identified in the synthetic dataset were subsequently validated with respect to the data measured on the real subjects. The result is a feasible and reproducible approach to study spinopelviche relationships in numerically limited clinical samples. This combination allows you to detect biologically plausible correlations and can help reduce the need for additional imaging studies on healthy subjects.
The applied methodology can also serve as an exploratory tool in research on the spine and in other fields characterized by limited datasets. The bidirectional workflow has allowed the actual data to be checked for corrections in the synthetic dataset, reducing the risk that the results were due to the case or limited number of the sample.
In summary, the study does not propose to replace clinical data with artificial data, but demonstrates how a computational method assisted by artificial intelligence can be used to generate hypotheses, verify them on real data and preserve biological plausibility. The authors of this research are: IRCCS Ospedale Galeazzi-Sant’Ambrogio, Milano Domenico Compagnone, Riccardo Cecchinato, Pedro Berjano and Claudio Lamartina. Marco Giacalone, LUMSA Santa Silvia, Palermo.
-Photo press@mediacreative.it
(ITALPRESS).
