Our proprietary Somnium Twin Engine™ represents a paradigm shift on how medicine is practiced. Through the convergence of disruptive technologies, the Twin Engine provides a platform for comprehensive and personalized digital twins, unlocking magnitudes of new research and therapies for patients. The timeline and exorbitant cost required to go from hypothesis generation to bedside clinical benefit is reduced to a fraction, enabling a new generation of healthier patients and more cost-efficient healthcare systems.
Unified Patient Twin
STE harmonizes clinical data from monitors, imaging, labs, and records into a real-time digital representation of each patient. It translates these signals into an evolving physiological model that reflects the patient's current and projected states.
Collaborative Intelligence
Dozens of specialized AI systems — for diagnostics, organ function, treatment planning, and prediction — collaborate through STE's innovative platform. Instead of dozens of isolated snapshots, clinicians see a single, evolving portrait: a whole-patient context that supports real-time decision-making.
Simulation and Clinical Reasoning
Clinicians can explore potential interventions virtually: altering therapy parameters and observing projected responses across organs and time. As real-world data accumulates, STE refines its internal models and continuously improves fidelity and personalization. Through STE, we can scale clinical reasoning from an individual patient level to the hospital level, ultimately leading to the best possible decisions.
Scaling Trust
STE solves one of the most important aspects of healthcare: trust. STE is designed from the ground up to meet emerging regulatory standards for responsible AI in medicine, and ensure we enhance the trust that exists between patients, healthcare workers, and the hospital systems.
The Road Ahead
Our initial deployments focus on critical care, where decisions are most time-sensitive and data-rich. From there, the Somnium Twin Engine will extend to cardiology, oncology, longevity, and aerospace medicine — building the first universal and wholistic framework for human digital twins.
Our results
To achieve viable preservation, we first diffusively loaded 300 micron thick acute rat cerebellar slices with a cryoprotectant solution (Fig1a and Fig1b) whose composition is identical to the established VMP cryoprotectant (Han et al. 2023, Fahy et al. 2004) save the removal of one component (X-1000 ice blocker).
In preparation for cryopreservation, acute slices of rat cerebellum were placed in a multifunctional sample holder with a custom well for diffusive loading of cryoprotectant (Fig4). The tissue was loaded with CPA, was rapidly cooled from 4ºC to -196ºC using a jet of liquid nitrogen (LN2), and then was rewarmed using an alternating magnetic field from an induction heater (Fig1d and Fig1e).
The sample holder's perfusion well was then placed back on top of the sample and cryoprotectant was unloaded using a linear ramp in perfused concentration into the well from the maximum CPA concentration to no CPA in 500 seconds. After ~45 minutes of post-cryopreservation incubation, the slices were transferred to a multielectrode array (3Brain Duplex) where activity was recorded at baseline before the addition of carbachol and the later addition of tetrodotoxin (Fig2).
Figure 1. Tissue Slice Cryopreservation Protocol.
a) System Overview. Acute neural tissue is loaded with VPMnoX cryoprotectant (53% w/v cryoprotectants) before vitrification and rewarming using a specialized chamber. Activity is then recorded on a multielectrode array.
b) CPA concentration as a function of time during cryoprotectant loading and unloading.
c) Vitrification and magnetic rewarming.
d) Cooling profile from the slice presented in Fig 2.
e) Rewarming profile from the slice presented in Fig 2.
For N=4 slices tested with this protocol, activity was present and the signal responded as expected to pharmacology with carbachol increasing activity from baseline and tetrodotoxin silencing action potentials. In Figure 2, we present data from one of these slices (additional data available upon request, email hunter@untillabs.com).
At baseline, a number of electrodes detected high frequency spiking (Fig2b and Fig2c). With application of carbachol to increase neural excitability, an increase in mean spike rate was observed. With tetrodotoxin application, spiking was blocked, confirming that the earlier signal represented physiological spiking.
While neural activity was present in multiple MEA channels, it is clear that there is a marked reduction in electrical activity following this cryopreservation protocol. Sample data from a healthy control slice at baseline (no pharmacological stimulation) is shown in Fig2e.
Of the 4 cryopreserved slices with maintained electrical activity, the slice depicted in Fig2 showed no signs of macro cracking. Large thermal gradients applied across a volume of tissue after a phase change are likely to cause cracking. Future experiments will involve tuning the vitrification and rewarming profile aiming to completely avoid cracking while also cooling and warming fast enough to avoid damaging ice formation.
Figure 2. MEA Electrical Activity from Vitrified Slice.
a) DAPI staining showing example morphology of a rat cerebellar slice after cryopreservation.
b) Raster plot from a multi-electrode array recording of a vitrified and rewarmed cerebellar slice. Color changes indicate onset of pharmacology, and black signifies a region of electrodes not in contact with tissue to track global noise. Zoomed panel shows excerpts from highly active channels (6-10) for visual clarity.
c) Mean firing rate for active areas in the same vitrified cerebellum slice as in (b), separated by pharmacology condition. Mean firing rate data from a separate healthy control slice is shown in gray.
d) Representative raw signal traces from the same vitrified cerebellar slice as in (b) and (c). Red lines denote detected spikes.
e) A representative raster plot showing high spiking activity from a healthy cerebellar slice being perfused with aCSF. Note time axis compared to (b).
f) Same as (d), but for a healthy cerebellar slice.
Have something to say?
Email us at contact@somniumbio.com