Current measures of neurodegenerative diseases are highly subjective and based on episodic visits. Consequently, drug development decisions rely on sparse, subjective data, which have led to the conduct of large-scale phase 3 trials of drugs that are likely not effective. Such failures are costly, deter future investment, and hinder the development of treatments. Given the lack of reliable physiological biomarkers, digital biomarkers may help to address current shortcomings. Objective, high-frequency data can guide critical decision-making in therapeutic development and allow for a more efficient evaluation of therapies of increasingly common disorders.

Artificial intelligence is becoming increasingly important in dermatology, with studies reporting accuracy matching or exceeding dermatologists for the diagnosis of skin lesions from clinical and dermoscopic images. However, real-world clinical validation is currently lacking. We review dermatological applications of deep learning, the leading artificial intelligence technology for image analysis, and discuss its current capabilities, potential failure modes, and challenges surrounding performance assessment and interpretability. We address the following three primary applications: (i) teledermatology, including triage for referral to dermatologists; (ii) augmenting clinical assessment during face-to-face visits; and (iii) dermatopathology. We discuss equity and ethical issues related to future clinical adoption and recommend specific standardization of metrics for reporting model performance.

Background: Clinician rating scales and patient-reported outcomes are the principal means of assessing motor symptoms in Parkinson disease and Huntington disease. However, these assessments are subjective and generally limited to episodic in-person visits. Wearable sensors can objectively and continuously measure motor features and could be valuable in clinical research and care. Methods: We recruited participants with Parkinson disease, Huntington disease, and prodromal Huntington disease (individuals who carry the genetic marker but do not yet exhibit symptoms of the disease), and controls to wear 5 accelerometer-based sensors on their chest and limbs for standardized in-clinic assessments and for 2 days at home. The study’s aims were to assess the feasibility of use of wearable sensors, to determine the activity (lying, sitting, standing, walking) of participants, and to survey participants on their experience. Results: Fifty-six individuals (16 with Parkinson disease, 15 with Huntington disease, 5 with prodromal Huntington disease, and 20 controls) were enrolled in the study. Data were successfully obtained from 99.3% (278/280) of sensors dispatched. On average, individuals with Huntington disease spent over 50% of the total time lying down, substantially more than individuals with prodromal Huntington disease (33%, p = 0.003), Parkinson disease (38%, p = 0.01), and controls (34%; p < 0.001). Most (86%) participants were “willing” or “very willing” to wear the sensors again. Conclusions: Among individuals with movement disorders, the use of wearable sensors in clinic and at home was feasible and well-received. These sensors can identify statistically significant differences in activity profiles between individuals with movement disorders and those without. In addition, continuous, objective monitoring can reveal disease characteristics not observed in clinic.

Mammalian skin is innervated by diverse, unmyelinated C fibers that are associated with senses of pain, itch, temperature, or touch. A key developmental question is how this neuronal cell diversity is generated during development. We reported previously that the runt domain transcription factor Runx1 is required to coordinate the development of these unmyelinated cutaneous sensory neurons, including VGLUT3ϩ low-threshold c-mechanoreceptors (CLTMs), MrgprD ϩ polymodal nociceptors, MrgprA3 ϩ pruriceptors, MrgprB4 ϩ c-mechanoreceptors, and others. However, how these Runx1-dependent cutaneous sensory neurons are further segregated is poorly illustrated. Here, we find that the Runx1-dependent transcription factor gene Zfp521 is expressed in, and required for establishing molecular features that define, VGLUT3 ϩ CLTMs. Furthermore, Runx1 and Zfp521 form a classic incoherent feedforward loop (I-FFL) in controlling molecular identities that normally belong to MrgprD ϩ neurons, with Runx1 and Zfp51 playing activator and repressor roles, respectively (in genetic terms). A knock-out of Zfp521 allows prospective VGLUT3 lineage neurons to acquire MrgprD ϩ neuron identities. Furthermore, Runx1 might form other I-FFLs to regulate the expression of MrgprA3 and MrgprB4, a mechanism preventing these genes from being expressed in Runx1-persistent VGLUT3 ϩ and MrgprD ϩ neurons. The evolvement of these I-FFLs provides an explanation for how modality-selective sensory subtypes are formed during development and may also have intriguing implications for sensory neuron evolution and sensory coding.

Most wearable sensor studies in Parkinson’s disease have been conducted in the clinic and thus may not be a true representation of everyday symptoms and symptom variation. Our goal was to measure activity, gait, and tremor using wearable sensors inside and outside the clinic. In this observational study, we assessed motor features using wearable sensors developed by MC10, Inc. Participants wore five sensors, one on each limb and on the trunk, during an in-person clinic visit and for two days thereafter. Using the accelerometer data from the sensors, activity states (lying, sitting, standing, walking) were determined and steps per day were also computed by aggregating over 2 s walking intervals. For non-walking periods, tremor durations were identified that had a characteristic frequency between 3 and 10 Hz. We analyzed data from 17 individuals with Parkinson’s disease and 17 age-matched controls over an average 45.4 h of sensor wear. Individuals with Parkinson’s walked significantly less (median [inter-quartile range]: 4980 [2835–7163] steps/day) than controls (7367 [5106–8928] steps/day; P = 0.04). Tremor was present for 1.6 [0.4–5.9] hours (median [range]) per day in most-affected hands (MDS-UPDRS 3.17a or 3.17b = 1–4) of individuals with Parkinson’s, which was significantly higher than the 0.5 [0.3–2.3] hours per day in less-affected hands (MDS-UPDRS 3.17a or 3.17b = 0). These results, which require replication in larger cohorts, advance our understanding of the manifestations of Parkinson’s in real-world settings.