Overview of Solutions in Quantum Information Technologies, part 2
Paper details
Section titled “Paper details”| Field | Value |
|---|---|
| Journal | International Journal of Electronics and Telecommunications |
| Published | 2026-07-17 |
| Authors | Igor Woźniak, Mateusz Jangas, Mateusz Piątek, Karol Franczuk, Piotr Liszewski |
Hybrid quantum-classical modules produced the review’s clearest numerical result in language model adaptation. QTHA cut trainable parameters by roughly 76% against a rank-matched LoRA baseline and lowered training loss by up to 17%. It also improved standard text-generation metrics across three benchmarks.
Across the review, quantum components handle one defined stage inside a classical system. The maturity range is wide. Photon-counting CT has entered routine clinical use, while X-ray SPDC, large-scale QUBO reconstruction, and clinical positronium imaging remain proof-of-principle or early-translation technologies.
Where quantum modules fit
Section titled “Where quantum modules fit”This review covers medical sensing, radiological imaging, PET, NMR ensemble computing, and LLM adaptation rather than presenting a new device experiment. Its X-ray taxonomy separates quantum-enabled methods, which use entanglement, heralding, or nonclassical photon correlations, from quantum-inspired methods, which use photon counting, spectral selection, and computational reconstruction.
A defect spin, photon energy, polarization correlation, or quantum state space supplies the new resource. Classical electronics, acquisition geometry, and workflows still carry most of the system. The authors treat calibration, hardware complexity, and signal extraction as costs that each proposed module must justify.
NV-center nanodiamonds measure local signals
Section titled “NV-center nanodiamonds measure local signals”An NV center forms when nitrogen occupies a site next to a missing carbon atom in the diamond lattice. Its trapped electron provides a local spin probe. Laser-induced fluorescence reports the spin configuration, while diamond’s chemical inertness and mechanical stability support operation near living tissue.
Researchers used gold nanoparticles as laser-controlled heat sources inside cells. NV thermometry reached precision better than 0.1°C and measured several-degree temperature differences across only a few micrometers. The cell remained stable at the reported heating level. Increasing laser power killed the cell, and the review presents monitoring during selective heating as a possible cancer-treatment application.
NV magnetometers reconstructed heart electrical activity by scanning the magnetic field. Animal models resolved Purkinje fibers that classical methods often missed. Their fast response captured impulse movement in real time, typically within tens of milliseconds.
Radical monitoring placed sensors near mitochondria and tracked T1 relaxation. Harmful substances shortened T1, while antioxidants returned it toward its original value.
NMR reveals why fully quantum scaling stalls
Section titled “NMR reveals why fully quantum scaling stalls”NMR uses nuclear spin states as qubits in a liquid sample under an external field. At 11.7 Tesla, the review lists resonant frequencies of approximately 500 MHz for hydrogen, 126 MHz for carbon-13, and 470 MHz for fluorine-19. Liquid-sample T2 values range from hundreds of milliseconds to several seconds, allowing hundreds or thousands of gate operations before decoherence.
An ensemble computer measures the averaged signal from approximately 10^23 identical molecules rather than one molecule. A true pure ground state would require cooling near 10^-3 K, so pseudo-pure states use calibrated RF pulses plus spatial and temporal averaging.
NMR enabled quantum teleportation in 1998 and a seven-qubit implementation of Shor’s factoring algorithm in 2001. That experiment factored 15 as 3 times 5.
The scaling penalty is explicit. Signal strength scales as n/2^n and drops by half with every added qubit. A 12-qubit system in 2006 required sensitive coils and more than 4,000 temporal repetitions to separate the output from thermal background.
Photon-counting CT leads clinical adoption
Section titled “Photon-counting CT leads clinical adoption”Photon-counting CT is the review’s most mature quantum-inspired medical method and is already in routine clinical practice. Energy-integrating detectors sum deposited energy over time, while photon-counting detectors register each photon and estimate its energy. This preserves spectral information for material decomposition and soft-tissue contrast. It also supports higher spatial resolution and reduced beam-hardening artifacts.
CdTe detectors operate around 1000 V and use pixelated anodes to convert X-ray photons into electron-hole pairs. Two practical errors remain. Pile-up merges photons within detector dead time, while charge sharing splits charge between neighboring pixels. Faster ASICs, adjacent-pixel anti-coincidence logic, and model-based corrections mitigate both effects.
X-ray SPDC, coincidence imaging, and heralded detection remain proof-of-principle because of low efficiency, photoionization losses, and strict coincidence requirements. Ghost and computational ghost imaging move complexity from detectors to reconstruction, but reduced instantaneous signal does not guarantee a lower total dose.
QUBO shifts CT reconstruction into optimization
Section titled “QUBO shifts CT reconstruction into optimization”A QUBO formulation maps each CT pixel to qubit states and binary weights. The discrete Radon transform relates this image to the sinogram, and the binary identity q^2 = q produces linear and quadratic error terms.
The calculations run through QAOA on gate-model processors such as those from IBM, or annealers such as D-Wave Advantage. Hybrid solvers decompose large matrices into smaller quantum-solvable subproblems.
Compared with local iterative methods, QUBO targets a global minimum and the entire sinogram at once. The review also reports improved artifact resistance and the possibility of reconstruction with fewer projections.
Segmentation can occur directly from raw data by restricting the search to discrete tissue classes. Those classes require far fewer variables than the 1024 to 4096 grayscale levels used in conventional images.
D-Wave Advantage has about 180 fully connected logical variables, limiting direct solvers to 13 by 13 pixel images. Hybrid solvers handle larger matrices. Lower-dose LDCT reconstruction is presented as a potential benefit, not a settled clinical result.
Polarization data extend PET information content
Section titled “Polarization data extend PET information content”Conventional PET detects two 511 keV photons from positron-electron annihilation but records only their arrival time and direction. It therefore ignores polarization and registers false lines of response from random coincidences and Compton scattering.
QE-PET uses the entangled polarization correlation of photons from para-positronium as a filter. The theoretical vacuum entanglement coefficient is approximately 2.85, but tissue-like measurements are lower because pick-off annihilation disrupts the pair.
J-PET’s plastic scintillators estimate polarization from Compton scattering angles, allowing true and background events to be separated. The review describes higher SNR and lower radiotracer activity as potential benefits rather than established clinical outcomes.
Positronium lifetime imaging is feasible in vivo and sensitive to oxygen concentration and tissue microstructure. Initial human studies show feasibility, but validation and standardization remain necessary before routine use.
Quantum adapters reduce LLM adaptation cost
Section titled “Quantum adapters reduce LLM adaptation cost”QTHA replaces a classical low-rank adapter with a Matrix Product Operator tensor network and a Quantum Neural Network. The tensor branch reduces parameter count through factorization. The quantum branch uses RY angle embeddings, controlled RZ gates, and Pauli Z expectation values.
Kong et al. report roughly 76% fewer trainable parameters than rank-matched LoRA and up to 17% lower training loss. The loss also reached its plateau about 20% earlier than with LoRA. These results were tested on the Origin Wukong superconducting quantum computer.
Kim et al. kept a Sentence Transformer frozen and attached a quantum classification head for SST-2 sentiment analysis. With 353 to 914 trainable parameters on 10 to 18 qubits, the best model reached 92.70% accuracy. It gained 3.14% over a size-matched support vector classifier and remained below the 93.62% fully fine-tuned classical result.
The energy analysis placed the crossover at 46 qubits, beyond which quantum processor inference used less energy than equivalent H100 GPU simulation.
HyQuT used 10 qubits and about 80 gates to replace either an 8M parameter feed-forward gate projection or a 150M parameter query projection. It removed roughly 10% of classical parameters and retained stable convergence after training on a 1.6 GB Chinese corpus. Replacing all four attention matrices or the full feed-forward network prevented convergence.
Remaining barriers include limited qubit counts, gradient estimation overhead, and the absence of fault-tolerant hardware.
Engineering choices and evidence gaps
Section titled “Engineering choices and evidence gaps”The engineering choice is which measurement or computational resource the classical pipeline lacks. A narrow module can use a single spin, a photon spectrum, or polarization data without replacing the surrounding instrument.
Comparisons must use well-tuned classical baselines at equal diagnostic quality. A quantum method does not automatically reduce dose or improve accuracy. Clinical systems also need validation and standardization.
Co-design between the quantum stage and classical readout remains a shared requirement across sensing, imaging, PET, and machine learning. PCCT remains the clearest deployed example because it records individual X-ray photons and already operates in routine clinical practice.
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-enhanced positron 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-of-principle 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.
Advanced Thermal Simulation Service Diamond Heat Spreader Performance Analysis
View Specs →
High Concentration Quantum Grade Single Crystal Diamond
View Specs →
Single Crystal Cvd Diamond Plate
View Specs →Original source
Section titled “Original source”- OpenAlex: https://openalex.org/W7169820700