Researchers have found that smartphone-based motor assessments, when combined with clinical scores, can predict dopamine deficiency in Parkinson’s disease (PD) without the need for brain scans. The study, published in NPJ Digital Medicine, focused on using motion data collected from smartphones and established clinical scales to identify early signs of dopamine loss.
Currently, confirming dopamine deficiency in PD relies on advanced imaging techniques such as dopamine transporter (DaT) single-photon emission computed tomography (SPECT). These methods are costly, expose patients to radiation, and are not widely accessible. SPECT measures the striatal binding ratio (SBR), which reflects levels of DaT in areas like the caudate nucleus and putamen. Lower SBR values indicate greater neuron loss and motor impairment.
The researchers built on previous work involving the Oxford Parkinson’s Disease Centre (OPDC) smartphone application. This app has been shown to differentiate between healthy individuals, those with isolated REM sleep behavior disorder (iRBD), and PD patients. It also predicts scores on the Movement Disorder Society-Unified Parkinson’s Disease Rating Scale Part III (MDS-UPDRS-III), a standard measure of motor function in PD.
In this study, 93 participants—including those with iRBD, PD, or neither—had both a DaT scan and a smartphone-based assessment within a year. Machine learning models were trained to use smartphone data to predict whether a DaT scan would show evidence of dopamine deficiency.
The model based solely on smartphone data achieved an 80% discrimination value for predicting positive or negative DaT scans. When both smartphone data and MDS-UPDRS-III scores were used together, the accuracy improved further, reaching an area under the curve (AUC) value of 85%. Models based only on clinical scores or both sources performed slightly better than those using just smartphone data.
Regression models showed modest success in predicting SBR values directly from movement data, particularly for symptoms such as gait disturbances, manual dexterity issues, and tremor. The high-frequency sampling by smartphones allowed detection of subtle motor features that might be missed during routine examinations.
However, when focusing only on milder cases of PD, all models became less accurate. This suggests that while these tools can help identify established dopamine deficiency or more advanced disease stages, they may be less reliable for detecting very early changes.
The authors concluded that combining detailed movement analysis from smartphones with standardized clinical scoring could improve access to screening for people at risk of PD or related disorders such as dementia with Lewy bodies. They noted: "Despite the small sample size, the study findings confirm the feasibility of combining smartphone-based motor assessment with clinical MDS-UPDRS-III scores to predict DaT scan status in people with iRBD and PD."
This approach offers a potential low-cost alternative for initial triage before considering more expensive or invasive diagnostic procedures.
