Skip to content

Overview of solutions in quantum information technologies, part 2

FieldValue
JournalInternational Journal of Electronics and Telecommunications
Published2026-07-31
AuthorsIgor Woźniak, Mateusz Jangas, Mateusz Piątek, Karol Franczuk, Piotr Liszewski
DOI10.24425/ijet.2026.1727

The review finds that present quantum hardware contributes through small modules inside classical workflows, not through general-purpose quantum systems. Its strongest reported figure comes from Quantum Tensor Hybrid Adaptation, which used roughly 76% fewer trainable parameters than rank-matched LoRA for two 7B language models. This result supports a bounded quantum task within a classical pipeline.

Maturity varies sharply. PCCT is entering routine clinical use, while X-ray SPDC, large QUBO reconstruction, and clinical positronium imaging remain at proof-of-principle or early translation stages.

Quantum modules occupy narrow workflow stages

Section titled “Quantum modules occupy narrow workflow stages”

The surveyed systems use quantum physics at one defined stage of a classical pipeline. The quantum stage supplies single-spin sensitivity, photon-resolved energy data, polarization correlations, or access to a high-dimensional parameter space through few qubits.

QE-PET adds polarization-resolved detection to coincidence hardware. QUBO reconstruction adds an optimization stage to CT. QTHA replaces one low-rank adaptation component while the language model remains classical.

Co-design is necessary. Hardware complexity, calibration, signal extraction, standardization, and clinical validation must be measured against a tuned classical alternative.

NV centers measure temperature, magnetic fields, and radicals

Section titled “NV centers measure temperature, magnetic fields, and radicals”

A nitrogen-vacancy center forms when nitrogen occupies a site beside a missing carbon atom in the diamond lattice. Its trapped electron spin can be read optically because fluorescence depends on spin configuration.

In intracellular thermometry, nanodiamond probes reported precision better than 0.1 °C. Laser-controlled gold nanoparticles produced temperature differences of several degrees over a few micrometers while the cell remained stable in the reported experiment. Raising laser power killed the cell, and the authors present this as a possible basis for monitored selective heating.

For magnetocardiography, nanodiamond sensors resolved Purkinje fiber signals in animal models that classical methods did not clearly show. Their fast response tracked electrical impulse movement within tens of milliseconds. Near mitochondria, harmful substances shortened T1 relaxation time, while antioxidants moved it toward its original value. The paper treats the latter as single-cell evidence for monitoring radical production and treatment effects.

X-ray imaging combines direct detection with quantum-inspired reconstruction

Section titled “X-ray imaging combines direct detection with quantum-inspired reconstruction”

Photon-counting CT is the most mature quantum-inspired X-ray method in the review. Each detector event corresponds to one absorbed X-ray photon, and the detector estimates its energy. Unlike energy-integrating CT, this preserves spectral information that supports material decomposition, soft-tissue contrast, higher spatial resolution, and fewer beam-hardening artifacts. The review reports that PCCT is entering routine clinical use.

X-ray spontaneous parametric down-conversion remains at proof-of-principle stage. Its correlated photon pairs can support coincidence detection and heralding, but low efficiency, photoionization loss, and strict coincidence requirements limit current experiments.

Ghost imaging uses a bucket detector and correlations with a known reference pattern. Computational ghost imaging removes the physical reference arm. Neither guarantees lower dose, because total dose also depends on measurement count, noise, and diagnostic quality.

QUBO expresses CT reconstruction as energy minimization. Binary qubit states and weights represent each pixel, and a discrete Radon transform links the image to the sinogram. The review describes QAOA on IBM gate-model processors and D-Wave Advantage annealing. Hybrid solvers divide large matrices into smaller quantum-solvable problems.

The stated D-Wave Advantage limit is about 180 fully connected logical variables. Direct solving is confined to 13 × 13 pixel images, although hybrid solvers can process larger matrices. The paper cites global optimization, artifact resistance, and reconstruction from fewer projections as advantages over local classical methods.

Cadmium telluride detectors operate around 1000 V. High flux causes pile-up, while photons near pixel boundaries cause charge sharing. Faster ASICs, anti-coincidence logic, and model-based correction address these errors.

The low-dose CT benefit remains conditional. These methods reduce dose only if they preserve equivalent diagnostic quality.

Polarization filtering adds information to PET

Section titled “Polarization filtering adds information to PET”

Conventional PET detects two 511 keV photons from positron-electron annihilation but records arrival time and direction without resolving polarization. Quantum-enhanced PET treats the polarization correlation as a filter for true pairs. The J-PET scanner uses a plastic scintillator and Compton scattering angles to estimate the relative polarization angle.

Entanglement is reduced in matter by pick-off annihilation. The review gives a theoretical vacuum value of R ≈ 2.85 and reports lower measurements in tissue-like conditions. Filtered events could improve SNR and permit lower radiotracer activity, but these remain stated benefits rather than a settled clinical dose result.

Positronium lifetime imaging could report tissue oxygenation and microstructure. Initial in vivo human studies show feasibility, but the authors require further validation and standardization before routine use.

NMR ensemble computing exposes the scaling cost

Section titled “NMR ensemble computing exposes the scaling cost”

NMR quantum computing uses ensembles of about 10^23 identical molecules. Its useful state is pseudo-pure because room-temperature nuclei remain in a mixed thermal state. A true pure state would require cooling near 10^-3 K. RF pulses, spatial averaging, and repeated experiments cancel much of the thermal background.

At 11.7 T, the review lists resonant frequencies of about 500 MHz for hydrogen-1, 126 MHz for carbon-13, and 470 MHz for fluorine-19. Liquid-sample T2 times range from hundreds of milliseconds to several seconds, allowing hundreds or thousands of gates.

The signal scales as n/2^n. A 12-qubit benchmark needed over 4,000 temporal repetitions to separate its output from thermal noise. Earlier work demonstrated teleportation and a 7-qubit Shor calculation that factored 15 into 3 and 5. The NMR platform therefore retains historical value but faces a signal loss that grows with register size.

Hybrid language-model modules cut parameter counts

Section titled “Hybrid language-model modules cut parameter counts”

Quantum Tensor Hybrid Adaptation combines a Matrix Product Operator with a parameterized quantum circuit. The quantum branch uses RY embeddings, controlled-RZ gates, and Pauli-Z expectation values. Both branches are combined linearly.

Against rank-matched LoRA on DeepSeek-R1-Distill-Qwen-7B and Qwen2-7B-Instruct, QTHA reduced trainable parameters by roughly 76%. It lowered training loss by up to 17%, improved standard text-generation metrics by a similar margin across three benchmarks, and reached its loss plateau about 20% earlier.

Inference ran on the superconducting Origin Wukong quantum computer, and the authors report that hardware shot noise did not remove the accuracy gain.

A frozen SetFit model with a quantum head used 353 to 914 trainable parameters across 10 to 18 qubits. It reached 92.70% accuracy on SST-2, 3.14% above a size-matched support vector classifier but below the 93.62% classical fine-tuning result. Its energy estimate placed the QPU and H100 GPU crossover at 46 qubits.

Hybrid Quantum Transformer replaced selected projections with about 80 gates on 10 qubits. It removed roughly 10% of classical parameters and retained coherent generation after pretraining on a 1.6 GB Chinese corpus. Replacing all four attention matrices or the full feed-forward network prevented convergence.

The remaining constraints are limited qubit counts, gradient-estimation overhead, and the lack of fault-tolerant hardware.

Engineering choices follow the physical resource

Section titled “Engineering choices follow the physical resource”

Engineers should identify the physical resource that classical hardware cannot supply directly, such as single-spin sensing or polarization-resolved events. PCCT has the clearest clinical path because photon counting already resolves each event. Entangled X-ray imaging and positronium imaging need further detector development and validation.

For LLM adaptation, the evidence supports compact modules at selected layers. Hybrid solvers extend beyond the direct QUBO size limit but still depend on the detector and calibration chain. Each proposal should be benchmarked against a tuned classical system for task accuracy, dose, energy, hardware complexity, and signal extraction.

The near-term engineering choice is a bounded quantum stage with its gain measured against a classical pipeline.

For NV-center quantum sensing, the paper requires nanodiamond with engineered nitrogen-vacancy defects and optical spin readout. The 6CCVD diamond materials catalog lists MPCVD-grown single-crystal and polycrystalline diamond, but its stated specifications do not establish nanodiamond formats or NV-center control. Compatibility therefore requires a specification review through engineering support before procurement.

View Original Abstract

The hardware available today does not support general-purpose quantum computation in medical settings, but specific subtasks such as sensing, photon-resolved detection, or image reconstruction can already be handled by quantum components embedded in classical systems. This article surveys recent developments across several branches of medical applications, with a complementary look at an analogous trend in machine learning. Nitrogen-vacancy centers in nanodiamonds enable nanoscale thermometry, high-resolution magnetocardiography, and real-time monitoring of free radicals inside living cells. In radiological imaging, photon-counting computed tomography is already entering routine clinical use, while QUBO-based image reconstruction on quantum annealers and gate-model processors points toward dose reduction in low-dose CT. Quantum-enhancedpositron emission tomography exploits polarization correlations of entangled annihilation photons to suppress background events and opens a path toward positronium-based biomarkers such as tissue hypoxia. Nuclear magnetic resonance ensemble computing is included to illustrate the scalability limits of fully quantum systems. The same hybrid pattern is then identified outside medicine, in efficient fine-tuning of large language models, where localized quantum modules reduce parameter counts without degrading task accuracy. The reviewed work differs in technical maturity, from systems already deployed clinically to proof-ofprinciple experiments such as X-ray spontaneous parametric down-conversion. Across this range, the quantum component handles one well-defined stage of an otherwise classical system.