CMOS Image Research Articles

Multicolor fluorescence microscopy for surgical guidance using a chip-scale imager with a low-NA fiber optic plate and a multi-bandpass interference filter.

In curative-intent cancer surgery, intraoperative fluorescence imaging of both diseased and healthy tissue can help to ensure the successful removal of all gross and microscopic diseases with minimal damage to neighboring critical structures, such as nerves. Current fluorescence-guided surgery (FGS) systems, however, rely on bulky and rigid optics that incur performance-limiting trade-offs between sensitivity and maneuverability. Moreover, many FGS systems are incapable of multiplexed imaging. As a result, clinical FGS is currently limited to millimeter-scale detection of a single fluorescent target. Here, we present a scalable, lens-less fluorescence imaging chip, VISION, capable of sensitive and multiplexed detection within a compact form factor. Central to VISION is a novel optical frontend design combining a low-numerical-aperture fiber optic plate (LNA-FOP) and a multi-bandpass interference filter, which is affixed to a custom CMOS image sensor. The LNA-FOP acts as a planar collimator to improve resolution and compensate for the angle-sensitivity of the interference filter, enabling high-resolution and multiplexed fluorescence imaging without lenses. We show VISION is capable of detecting tumor foci of less than 100 cells at near video framerates and, as proof of principle, can simultaneously visualize both tumors and nerves in ex vivo prostate tissue.

Characterization of Partial Discharges in Dielectric Oils Using High-Resolution CMOS Image Sensor and Convolutional Neural Networks.

In this work, an exhaustive analysis of the partial discharges that originate in the bubbles present in dielectric mineral oils is carried out. To achieve this, a low-cost, high-resolution CMOS image sensor is used. Partial discharge measurements using that image sensor are validated by a standard electrical detection system that uses a discharge capacitor. In order to accurately identify the images corresponding to partial discharges, a convolutional neural network is trained using a large set of images captured by the image sensor. An image classification model is also developed using deep learning with a convolutional network based on a TensorFlow and Keras model. The classification results of the experiments show that the accuracy achieved by our model is around 95% on the validation set and 82% on the test set. As a result of this work, a non-destructive diagnosis method has been developed that is based on the use of an image sensor and the design of a convolutional neural network. This approach allows us to obtain information about the state of mineral oils before breakdown occurs, providing a valuable tool for the evaluation and maintenance of these dielectric oils.

Comparison of modal and zonal wavefront measurements of refractive extended depth of focus intraocular lenses.

Extended depth-of-focus (EDoF) intraocular lenses (IOLs) are typically evaluated using commercially available aberrometers. Given the intricate optical design of these IOLs, employing an appropriate wavefront reconstruction method with a sufficient sampling resolution of the aberrometer is crucial. A high-resolution Shack-Hartmann wavefront sensor was developed by magnifying the pupil aperture by a factor of five onto a lenslet array (pitch: 133 µm) and utilizing a full-frame CMOS sensor (24 by 36 mm), resulting in a 26.6 µm sampling resolution. Zonal wavefront reconstruction was used and compared with Zernike-based modal wavefront reconstruction to retain detailed local slope irregularities. Four refractive EDoF IOLs with a power of 20D were examined, and the wavefront difference between the zonal and modal methods, expressed as the root mean squared error (RMSE), remained significant for two of the IOLs up to the 16th-order Zernike spherical aberrations (SAs). Conversely, a negligibly small RMSE was observed for the other two IOLs, as long as the Zernike SAs were higher than the 6th order. The raytracing simulation results from the zonal wavefronts exhibited a stronger correlation with the results of recent optical bench studies than those from the modal wavefronts. The study suggests that certain recent refractive EDoF IOLs possess a complex optical profile that cannot be adequately characterized by limited orders of SAs.

Machine-Learning-Assisted Blood Parameter Sensing Platform for Rapid Next Generation Biomedical and Healthcare Applications

The pursuit of rapid diagnosis has resulted in considerable advances in blood parameter sensing technologies. As advances in technology, there may be challenges in equitable access for all individuals due to economic constraints, advanced expertise, limited accessibility in particular places, or insufficient infrastructure. Hence, simple, cost efficient, benchtop biochemical blood-sensing platform was developed for detecting crucial blood parameters for multiple disease diagnosis. Colorimetric and image processing techniques is used to evaluate color intensity. CMOS image sensor is utilized to capture images to calculate optical density for sensing. The platform is assessed with blood serum samples, including Albumin, Gamma Glutamyl Transferase, Alpha Amylase, Alkaline Phosphatase, Bilirubin, and Total Protein within clinically relevant limits. The platform had excellent Limits of Detection (LOD) for these parameters, which are critical for diagnosing liver and kidney-related diseases (0.27 g dl−1, 0.86 IU l−1, 1.24 IU l−1, 0.97 IU l−1, 0.24 mg dl−1, 0.35 g dl−1, respectively). Machine learning (ML) algorithms were used to estimate targeted blood parameter concentrations from optical density readings, with 98.48% accuracy and reduced incubation time by nearly 80%. The proposed platform is compared to commercial analyzers, which demonstrate excellent accuracy and reproducibility with remarkable precision (0.03 to 0.71%CV). The platform’s robust stability of 99.84% was shown via stability analysis, indicating its practical applicability.

The DMAPS upgrade of the Belle II vertex detector

Abstract The SuperKEKB collider will undergo a major upgrade to reach the target luminosity of 6 × 1035 cm-2 s-1 during the long shutdown (LS2) foreseen to start around year 2027. We are developing a new vertex detector (VTX) to replace the current one (VXD). This new pixel silicon tracker aims to be both more robust against the higher level of machine background and more performant in terms of precision on the decay vertices and standalone track finding efficiency. The baseline layout consists of two layers composing the inner part (iVTX) and three outer layers (oVTX), all arranged in a barrel-shaped geometry, with minimal material budget. All layers will be equipped with dedicated depleted monolithic active CMOS pixel sensors (DMAPS) named OBELIX, designed in the TowerJazz 180 nm technology. This paper will review all the aspects of the project: the detector specifications and the baseline design, the expected improved performance, the OBELIX features and its design status, including the tests of the forerunner chip TJ-Monopix2, and the fabrication and tests for the iVTX and oVTX ladder prototypes.

A monolithic active pixel sensor with node-based, data-driven, parallel readout for the high energy physics experiment vertex detector

We present the design of a prototype MAPS sensor MIC6_V1 based on a 55 nm Quad-well CMOS Image Sensor process for the CEPC vertex detector. A new node-based, data-driven, parallel readout architecture is implemented to achieve high spatial resolution, fast readout, and low power consumption. The size of MIC6_V1 is 2.8 mm × 2.8 mm, which contains a pixel matrix of 64 rows by 64 columns, and the pixel size is 23.6 μ m × 20 μ m. The integration time is 5 μs, and the hit arrival time measurement accuracy is 10 ns.

Simulation and first characterization of MAPS test structures with gain for timing applications

Thanks to their advantages in terms of easiness of manufacturing and reduced production costs, Monolithic Active Pixel Sensors (MAPS) represent an appealing solution for radiation imaging applications, which require to cover large areas with pixelated detectors. In the next upgrade of the ALICE detector, that will have to deal with the higher event rate resulting from the planned increase in the LHC luminosity, it is foreseen to include two additional sensor layers to perform Time of Flight (ToF) measurements. Trying to reach the challenging timing resolution required by the ALICE ToF layers, an internal gain layer has been included in the test structures of the third engineering run of the ARCADIA project to improve the timing performance of this MAPS technology. In the paper we will present an overview of the main results obtained from the electrical and the dynamic characterization of the fabricated devices, which have been compared with the behavior expected from the preliminary TCAD simulations carried out in the design phase. The experimental results confirmed the feasibility of embedding a gain layer in the ARCADIA 110 nm CMOS technology to develop monolithic LGADs.

Prototype design of a Monolithic Active Pixel Sensor for the Electron-ion collider in China

The Nupix-R1 is a prototype Monolithic Active Pixel Sensor designed for the Electron-ion collider in China. It consists of 128 rows × 128 columns of square pixels with a pitch of 26.3 µm. Since it is implemented in a 130 nm commercial triple-well process, the in-pixel circuit contains only NMOS transistors. The Nupix-R1 can simultaneously measure the energy, arrival time and position of particle hits. In addition, it can also record only the hit positions with a novel zero-compression scheme to increase the readout speed and reduce the data volume. This paper will discuss the design of the Nupix-R1 sensor.

First measurements with monolithic active pixel test structures produced in a 65 nm CMOS process

The Inner Tracking System (ITS) of the ALICE experiment at CERN will undergo an upgrade during the LHC long shutdown 3, in which the three innermost tracking layers will be replaced. This upgrade, named the Inner Tracking System 3 (ITS3), employs stitched wafer-scale Monolithic Active Pixel Sensors fabricated in a 65 nm CMOS process. The sensors are 260 mm in length and thinned to less than 50 μm then bent to form truly half-cylindrical half-barrels. The feasibility of this process for the ITS3 was explored with the first test production run (MLR1) in 2021, whose goal was to evaluate the charged particle detection efficiency and the sensor performance under non-ionising and ionising radiation up to the expected levels for ALICE ITS3 of 1013 1 MeV neq cm-2 (NIEL) and 10 kGy (TID). Three sensor flavours were produced to investigate this process: Analog Pixel Test Structure (APTS), Circuit Exploratoire 65 (CE65) and Digital Pixel Test Structure (DPTS).This contribution gives an overview of the MLR1 submission and test results, describing the different sensor flavours and presenting the results of the performance measurements done with particle beams for various chip variants and irradiation levels.

First results of the upgraded ALICE Inner Tracking System in LHC Run 3

Major upgrades of the ALICE experiment at CERN were completed during the LHC Long Shutdown 2 (2019–2021). The ALICE detector is currently taking data and has been doing so from the start of the third period of operation of the LHC (Run 3) on July 5th, 2022. One key part of these upgrades is the new Inner Tracking System (ITS2), a full silicon-pixel vertexing and tracking detector constructed entirely with CMOS monolithic active pixel sensors (ALPIDE). The ITS2 consists of three inner layers (50 μm thick sensors) and four outer layers (100 μm thick sensors) covering 10 m2 and containing 12.5 billion pixels with a pixel pitch of 27 μ m × 29 μm. It offers a significant improvement in impact-parameter resolution and tracking efficiency, thanks to the increased granularity, the very low material budget (0.35% X0 /layer in the inner barrel) as well as a smaller beam pipe radius.The ITS2 was successfully installed in the ALICE experiment in May 2021, followed by a period of comprehensive on-site commissioning, before starting data taking in July 2022. In this paper, the detector construction and commissioning will be introduced briefly. The performance results from the first phase of proton-proton collisions recorded to date in LHC Run 3 will be discussed in detail, which include detector calibration, long-term evolution of the ALPIDE sensor threshold and noise, and the first measurement of the impact parameter.

ALICE ITS3: a bent stitched MAPS-based vertex detector

The ALICE ITS3 is a novel vertex detector replacing the innermost layers of ITS2 during LS3. Composed of three truly cylindrical layers of wafer-sized 65 nm stitched Monolithic Active Pixel Sensors, ITS3 provides high-resolution tracking of charged particles generated in heavy-ion collisions. This contribution presents an overview of the ITS3 detector, highlighting its design features, integration and cooling, and the ongoing development towards the final sensor. Furthermore, the paper introduces the off-detector service electronics, which play an essential role in the readout, control, and power supply of the detector.

Limiting Factors of Simultaneous Measurement Method for Turbidity and Total Suspended Solids Based on Image Processing Approaches

Image processing is one of the computational approaches that can be applied to measure fluctuations in surface water pollutant concentrations. Identifying factors that can affect and become a limitation in the development of image processing-based simultaneous measurement methods is the focus of the discussion in this article. Nineteen variables have been identified from the three-factor categories of hardware configuration, software settings, and the measured suspension characteristics. Measurements were made on thirty images extracted from video captured with a USB Mechanic-DX-230 camera with a 23-megapixel resolution Panasonic CMOS sensor equipped with a macro lens in 130X magnification on the S-EYE-1.6.0.11 interface. Image frame extraction was performed with VirtualDub2 build 4428/release, followed by digital image processing and analysis with ImageJ 1.46r. The lens’s focal length to the sample cell is 5 cm, setting the brightness parameter minimum of 90 and maximum of 255 and minimum threshold settings of 209 and maximum of 255, which is a controlled factor to achieve the best repeatability rate. It is indicated by a relative standard deviation of up to 6% in the measurement chamber with a dark background. Complying with the stated factors is essential to ensure measurement results’ reliability and validity.

Quantum Efficiency Measurement and Modeling of Silicon Sensors Optimized for Soft X-ray Detection

Hybrid pixel detectors have become indispensable at synchrotron and X-ray free-electron laser facilities thanks to their large dynamic range, high frame rate, low noise, and large area. However, at energies below 3 keV, the detector performance is often limited because of the poor quantum efficiency of the sensor and the difficulty in achieving single-photon resolution due to the low signal-to-noise ratio. In this paper, we address the quantum efficiency of silicon sensors by refining the design of the entrance window, mainly by passivating the silicon surface and optimizing the dopant profile of the n+ region. We present the measurement of the quantum efficiency in the soft X-ray energy range for silicon sensors with several process variations in the fabrication of planar sensors with thin entrance windows. The quantum efficiency for 250 eV photons is increased from almost 0.5% for a standard sensor to up to 62% as a consequence of these developments, comparable to the quantum efficiency of backside-illuminated scientific CMOS sensors. Finally, we discuss the influence of the various process parameters on quantum efficiency and present a strategy for further improvement.

Hot-Carrier Damage in N-Channel EDMOS Used in Single Photon Avalanche Diode Cell through Quasi-Static Modeling.

A single photon avalanche diode (SPAD) cell using N-channel extended-drain metal oxide semiconductor (N-EDMOS) is tested for its hot-carrier damage (HCD) resistance. The stressing gate-voltage (VGS) dependence is compared to hot-hole (HH) injection, positive bias temperature (PBT) instability and off-mode (VGS = 0). The goal was to check an accurate device lifetime extraction using accelerated DC to AC stressing by applying the quasi-static (QS) lifetime technique. N-EDMOS device is devoted to 3D bonding with CMOS imagers obtained by an optimized process with an effective gate-length Leff = 0.25 µm and a SiO2 gate-oxide thickness Tox = 5 nm. The operating frequency is 10 MHz at maximum supply voltage VDDmax = 5.5 V. TCAD simulations are used to determine the real voltage and timing configurations for the device in a mixed structure of the SPAD cell. AC device lifetime is obtained using worst-case DC accelerating degradation, which is transferred by QS technique to the AC waveforms applied to N-EDMOS device. This allows us to accurately obtain the AC device lifetime as a function of the delay and load for a fixed pulse shape. It shows the predominance of the high energy hot-carriers involved in the first substrate current peak during transients.

Failure mechanisms of a silicon-based CMOS image sensor irradiated by a 1550 nm nanosecond laser.

Cameras, LiDAR, and radars are indispensable for accurate perception of the surrounding environment and autonomous driving. Failure mechanisms of silicon-based CMOS image sensor (CIS) irradiated by 1550 nm nanosecond laser were investigated systematically in this paper. The damages of CIS were divided into point damage, line damage, and cross damage according to different damage performances. The damage thresholds under different irradiation conditions (different repetition rates, pulse widths, and irradiation times) were explored. Large repetition rates and long irradiation times would induce more heat accumulation, more temperature increase, and a low point damage threshold. The damage threshold for a pulse with a narrow pulse width is lower than that for a pulse with a long pulse width. The damaged CIS was analyzed further by focused ion beam (FIB) and scanning electron microscope (SEM). The damage location in the internal CIS structure was analyzed and the overall failure process was summarized. The results we get could enrich the database of laser damage mechanisms and laser damage thresholds of CIS, which will provide meaningful guidance for the camera design technology and anti-laser reinforcement technology of optoelectronic devices.

Segmentation of Low-Light Optical Coherence Tomography Angiography Images under the Constraints of Vascular Network Topology.

Optical coherence tomography angiography (OCTA) offers critical insights into the retinal vascular system, yet its full potential is hindered by challenges in precise image segmentation. Current methodologies struggle with imaging artifacts and clarity issues, particularly under low-light conditions and when using various high-speed CMOS sensors. These challenges are particularly pronounced when diagnosing and classifying diseases such as branch vein occlusion (BVO). To address these issues, we have developed a novel network based on topological structure generation, which transitions from superficial to deep retinal layers to enhance OCTA segmentation accuracy. Our approach not only demonstrates improved performance through qualitative visual comparisons and quantitative metric analyses but also effectively mitigates artifacts caused by low-light OCTA, resulting in reduced noise and enhanced clarity of the images. Furthermore, our system introduces a structured methodology for classifying BVO diseases, bridging a critical gap in this field. The primary aim of these advancements is to elevate the quality of OCTA images and bolster the reliability of their segmentation. Initial evaluations suggest that our method holds promise for establishing robust, fine-grained standards in OCTA vascular segmentation and analysis.

High precision orientation mapping from 4D-STEM precession electron diffraction data through quantitative analysis of diffracted intensities

The association of scanning transmission electron microscopy (STEM) and detection of a diffraction pattern at each probe position (so-called 4D-STEM) represents one of the most promising approaches to analyze structural properties of materials with nanometric resolution and low irradiation levels. This is widely used for texture analysis of materials using automated crystal orientation mapping (ACOM). Herein, we perform orientation mapping in InP nanowires exploiting precession electron diffraction (PED) patterns acquired by an axial CMOS camera. Crystal orientation is determined at each probe position by the quantitative analysis of diffracted intensities minimizing a residue comparing experiments and simulations in analogy to x-ray structural refinement. Our simulations are based on the two-beam dynamical diffraction approximation and yield a high angular precision (∼0.03°), much lower than the traditional ACOM based on pattern matching algorithms (∼1°). We anticipate that simultaneous exploration of both spot positions and high precision crystal misorientation will allow the exploration of the whole potentiality provided by PED-based 4D-STEM for the characterization of deformation fields in nanomaterials.

Single Slope Single Ramp Column ADC with Digital CDS for CMOS Image Sensors

Single Slope Single Ramp Column ADC with Digital CDS for CMOS Image Sensors

Motion-resistant three-wavelength spatial frequency domain imaging system with ambient light suppression using an 8-tap CMOS image sensor.

We present a motion-resistant three-wavelength spatial frequency domain imaging (SFDI) system with ambient light suppression using an 8-tap complementary metal-oxide semiconductor (CMOS) image sensor (CIS) developed at Shizuoka University. The system addresses limitations in conventional SFDI systems, enabling reliable measurements in challenging imaging scenarios that are closer to real-world conditions. Our study demonstrates a three-wavelength SFDI system based on an 8-tap CIS. We demonstrate and evaluate the system's capability of mitigating motion artifacts and ambient light bias through tissue phantom reflectance experiments and in vivo volar forearm experiments. We incorporated the Hilbert transform to reduce the required number of projected patterns per wavelength from three to two per spatial frequency. The 8-tap image sensor has eight charge storage diodes per pixel; therefore, simultaneous image acquisition of eight images based on multi-exposure is possible. Taking advantage of this feature, the sensor simultaneously acquires images for planar illumination, sinusoidal pattern projection at three wavelengths, and ambient light. The ambient light bias is eliminated by subtracting the ambient light image from the others. Motion artifacts are suppressed by reducing the exposure and projection time for each pattern while maintaining sufficient signal levels by repeating the exposure. The system is compared to a conventional SFDI system in tissue phantom experiments and then in vivo measurements of human volar forearms. The 8-tap image sensor-based SFDI system achieved an acquisition rate of 9.4 frame sets per second, with three repeated exposures during each accumulation period. The diffuse reflectance maps of three different tissue phantoms using the conventional SFDI system and the 8-tap image sensor-based SFDI system showed good agreement except for high scattering phantoms. For the in vivo volar forearm measurements, our system successfully measured total hemoglobin concentration, tissue oxygen saturation, and reduced scattering coefficient maps of the subject during motion (16.5cm/s) and under ambient light (28.9lx), exhibiting fewer motion artifacts compared with the conventional SFDI. We demonstrated the potential for motion-resistant three-wavelength SFDI system with ambient light suppression using an 8-tap CIS.

Detector development for particle physics

In high-energy physics experiments, tracking and vertexing is nowadays mostly done using semiconductor detectors. Among the employed detectors are hybrid pixel sensors, passive sensors and recently also depleted monolithic active pixel sensors (DMAPS), which integrate the particle sensor with front-end electronics. Currently, the dominant material for the production of such sensors is silicon. However, the use of silicon carbide is currently being investigated. In this work we report on our progress on the development of silicon-based DMAPS. Further, we present two approaches for reading out passive silicon carbide detectors at particle rates from the kHz to the GHz range.