A foundation model for atomistic materials chemistry
Batatia Ilyes, et al.
The Journal of chemical physics, 25 August 2025
Atomistic simulations of matter, especially those that leverage first-principles (ab initio) electronic structure theory, provide a microscopic view of the world, underpinning much of our understanding of chemistry and materials science. Over the last decade or so, machine-learned force fields have transformed atomistic modeling by enabling simulations of ab initio quality over unprecedented time and length scales. However, early machine-learning (ML) force fields have largely been limited by (i) the substantial computational and human effort required to develop and validate potentials for each particular system of interest and (ii) a general lack of transferability from one chemical system to the next. Here, we show that it is possible to create a general-purpose atomistic ML model, trained on a public dataset of moderate size, that is capable of running stable molecular dynamics for a wide range of molecules and materials. We demonstrate the power of the MACE-MP-0 model—and its qualitative and at times quantitative accuracy—on a diverse set of problems in the physical sciences, including properties of solids, liquids, gases, chemical reactions, interfaces, and even the dynamics of a small protein. The model can be applied out of the box as a starting or “foundation” model for any atomistic system of interest and, when desired, can be fine-tuned on just a handful of application-specific data points to reach ab initio accuracy. Establishing that a stable force-field model can cover almost all materials changes atomistic modeling in a fundamental way: experienced users obtain reliable results much faster, and beginners face a lower barrier to entry. Foundation models thus represent a step toward democratizing the revolution in atomic-scale modeling that has been brought about by ML force fields.
Artificial Intelligence-Powered Raman Spectroscopy through Open Science and FAIR Principles
Nicolas Coca-Lopez, Victor Alcolea-Rodriguez, Miguel A. Bañares, Sandor Brockhauser, Julien Gorenflot, Alex Henderson, Ron Hildebrandt, Nina Jeliazkova, Nikolay Kochev, Enrique Lozano Diz, Zdenek Pilat, Dario Polli, Philip Strömert, Chris Sturm, Renzo Vanna, and Raquel Portela
ACS Nano, Oct 2025
Raman spectroscopy is a fast-growing and increasingly powerful analytical technique applied across diverse disciplines such as materials science, chemistry, biology and medicine. This growth is driven by advances in Raman instrumentation and greatly supported by the flourishing of chemometrics and artificial intelligence (AI). However, the full potential of this technique is often hampered by challenges related to data acquisition, processing, interpretation, and sharing. This review paper addresses how a concerted effort toward digitalization, incorporating principles of Open Science and FAIR data (Findable, Accessible, Interoperable, and Reusable), is essential to develop and implement robust, standardized, and accessible digital workflows. These workflows are key to unlock the full power of Raman spectroscopy in combination with AI. We explore the current landscape of digital tools and open resources in Raman spectroscopy, highlighting both existing solutions as well as critical gaps. Despite these advances, the field remains fragmented, with many initiatives developed in isolation, limiting interoperability and slowing progress. In this regard, we assess the trends in Raman spectroscopy hardware and control software as well as the role of AI in improving data collection, automating data analysis, extracting meaningful insights, and enabling predictive modeling. We review challenges such as data quality and model interpretability that constrain the effectiveness and applicability of AI in Raman spectroscopy. Furthermore, we emphasize the importance of standardized data formats, metadata schemas, and domain-specific ontologies to ensure machine-actionability, database federation and interoperability as well as to facilitate collaborative research. We provide curated lists of existing open hardware, databases and standards relevant to Raman spectroscopy. Finally, we propose a roadmap toward an open and FAIR ecosystem for Raman spectroscopy, emphasizing the need for sustainable infrastructure, collaborative development, and community involvement.
Establishment of a clinical workflow for in vivo Raman spectroscopy during head and neck cancer surgery
Ayman Bali, Thomas Bitter, Mussab Kouka, Jonas Ballmaier, Ines Latka, Florian Windirsch, David Pertzborn, Nadja Ziller, Marcela Mafra, Nikolaus Gaßler, Jürgen Popp, Anna Mühlig, Ferdinand von Eggeling, Orlando Guntinas-Lichius, and Iwan W. Schie
Scientific Reports, July 2025
As first part of an ongoing prospective feasibility trial (DRKS00028114) this work explored the integration of in vivo Raman spectroscopy (RS) into the routine setting workflow of head and neck cancer (HNC) surgery. In vivo RS was performed intraoperatively on 30 patients with HNC cell carcinoma and 10 patients with inflammatory diseases as a control group. A standardized process was established using a Raman system complied with stringent medical device regulatory standards. Spectra were collected in vivo from the tumor site, the tumor margins, and healthy tissue. The learning curve of the HNC team significantly improved measurement times from over 30 min initially to 2 min after 15 patients. Data from 35 patients were interpretable, demonstrating clear spectral differences between tumor and healthy tissues. The intraoperative in vivo RS workflow is now well established and is being used in the ongoing clinical trial.
Real-Time Intraoperative Decision-Making in Head and Neck Tumor Surgery: A Histopathologically Grounded Hyperspectral Imaging and Deep Learning Approach
Ayman Bali1, Saskia Wolter, Daniela Pelzel, Ulrike Weyer, Tiago Azevedo, Pietro Lio, Mussab Kouka, Katharina Geißler, Thomas Bitter, and Günther Ernst
Cancers, May 2025
Our study describes a novel approach that combines hyperspectral imaging and deep learning, validated with histopathological ground truth, to enable rapid and accurate delineation of tumor margins in head and neck cancer surgeries. We present a system that acquires spectral data from freshly resected tumor samples and reconstructs three-dimensional models, which are then analyzed by convolutional neural networks. This label-free technique offers real-time tissue classification that can potentially reduce surgical times and enhance patient outcomes. By providing a precise and efficient alternative to traditional histopathology, our work opens new possibilities for more data-driven surgical decisions and improved cancer diagnostics.
SESAM mode-locked Nd:fiber laser at 920 nm for nonlinear optical microscopy
Fedele Pisani, Gabriele Di Noia, Francesco Crisafi, Matteo Negro, Giulio Cerullo, and Gianluca Galzeran.
Optics Express, May 2025
DOI: 10.1364/OE.557820
We report on the design, modeling, and characterization of a diode-pumped Nd-doped all-polarization-maintaining fiber laser operating at 920 nm in a passively mode-locked regime enabled by a SESAM. The laser generates pulse trains with a repetition frequency of 18.5 MHz and an average output power of 0.26 mW in an optical bandwidth of 0.6 nm. Within the broad integration bandwidth from 50 Hz to 9.25 MHz, the Nd:fiber laser exhibits ultra-low intensity noise, at a relative level of 0.04%, and timing jitter below 2 ps. Moreover, we demonstrate the ability to lock the laser’s repetition frequency to that of an Yb-doped oscillator, enabling stimulated Raman scattering microscopy in the fingerprint region.
Estimation of biological variance in coherent Raman microscopy data of two cell lines using chemometrics
Rajendhar Junjuri, Matteo Calvarese, MohammadSadegh Vafaeinezhad, Federico Vernuccio, Marco Ventura, Tobias Meyer-Zedler, Benedetta Gavazzoni, Dario Polli, Renzo Vanna, Italia Bongarzone, Silvia Ghislanzoni, Matteo Negro, Juergen Popp, Thomas Bocklitz.
Analyst, July 2024
DOI: 10.1039/D4AN00648H
Broadband Coherent Anti-Stokes Raman Scattering (BCARS) is a valuable spectroscopic imaging tool for visualizing cellular structures and lipid distributions in biomedical applications. However, the inevitable biological changes in the samples (cells/tissues/lipids) introduce spectral variations in BCARS data and make analysis challenging. In this work, we conducted a systematic study to estimate the biological variance in BCARS data of two commonly used cell lines (HEK293 and HepG2) in biomedical research. The BCARS data were acquired from two different experimental setups (Leibniz Institute of Photonics Technology (IPHT) in Jena and Politecnico di Milano (POLIMI) in Milano) to evaluate the reproducibility of results. Also, spontaneous Raman data were independently acquired at POLIMI to validate those results. First, Kramers–Kronig (KK) algorithm was utilized to retrieve Raman-like signals from the BCARS data, and a pre-processing pipeline was subsequently used to standardize the data. Principal component analysis – Linear discriminant analysis (PCA-LDA) was performed using two cross-validation (CV) methods: batch-out CV and 10-fold CV. Additionally, the analysis was repeated, considering different spectral regions of the data as input to the PCA-LDA. Finally, the classification accuracies of the two BCARS datasets were compared with the results of spontaneous Raman data. The results demonstrated that the CH band region (2770–3070 cm−1) and spectral data in the 1500–1800 cm−1 region have significantly contributed to the classification. A maximum of 100% balanced accuracies were obtained for the 10-fold CV for both BCARS setups. However, in the case of batch-out CV, it is 92.4% for the IPHT dataset and 98.8% for the POLIMI dataset. This study offers a comprehensive overview for estimating biological variance in biomedical applications. The insights gained from this analysis hold promise for improving the reliability of BCARS measurements in biomedical applications, paving the way for more accurate and meaningful spectroscopic analyses in the study of biological systems.
Terahertz near-field microscopy of metallic circular split ring resonators with graphene in the gap
Chiara Schiattarella, Alessandra Di Gaspare, Leonardo Viti, M. Alejandro Justo Guerrero, Lianhe H. Li, Mohammed Salih, A. Giles Davies, Edmund H. Linfield, Jincan Zhang, Hamideh Ramezani, Andrea C. Ferrari, Miriam S. Vitiell.
Scientific Reports, July 2024
Optical resonators are fundamental building blocks of photonic systems, enabling meta-surfaces, sensors, and transmission filters to be developed for a range of applications. Sub-wavelength size (< λ/10) resonators, including planar split-ring resonators, are at the forefront of research owing to their potential for light manipulation, sensing applications and for exploring fundamental light-matter coupling phenomena. Near-field microscopy has emerged as a valuable tool for mode imaging in sub-wavelength size terahertz (THz) frequency resonators, essential for emerging THz devices (e.g. negative index materials, magnetic mirrors, filters) and enhanced light-matter interaction phenomena. Here, we probe coherently the localized field supported by circular split ring resonators with single layer graphene (SLG) embedded in the resonator gap, by means of scattering-type scanning near-field optical microscopy (s-SNOM), using either a single-mode or a frequency comb THz quantum cascade laser (QCL), in a detectorless configuration, via self-mixing interferometry. We demonstrate deep sub-wavelength mapping of the field distribution associated with in-plane resonator modes resolving both amplitude and phase of the supported modes, and unveiling resonant electric field enhancement in SLG, key for high harmonic generation.
Label-free morpho-molecular phenotyping of living cancer cells by combined Raman spectroscopy and phase tomography
Arianna Bresci, Koseki J. Kobayashi-Kirschvink, Giulio Cerullo, Renzo Vanna, Peter T. C. So, Dario Polli, Jeon Woong Kan.
Communications Biology, June 2024
Accurate, rapid and non-invasive cancer cell phenotyping is a pressing concern across the life sciences, as standard immuno-chemical imaging and omics require extended sample manipulation. Here we combine Raman micro-spectroscopy and phase tomography to achieve label-free morpho-molecular profiling of human colon cancer cells, following the adenoma, carcinoma, and metastasis disease progression, in living and unperturbed conditions. We describe how to decode and interpret quantitative chemical and co-registered morphological cell traits from Raman fingerprint spectra and refractive index tomograms. Our multimodal imaging strategy rapidly distinguishes cancer phenotypes, limiting observations to a low number of pristine cells in culture. This synergistic dataset allows us to study independent or correlated information in spectral and tomographic maps, and how it benefits cell type inference. This method is a valuable asset in biomedical research, particularly when biological material is in short supply, and it holds the potential for non-invasive monitoring of cancer progression in living organisms.
A compact, turn-key platform for multiplex stimulated Raman scattering microscopy
Francesco Crisafi, Benedetta Talone, Andrea Ragni, Gabriele Di Noia, Mujeeb Rahman, Jing He, Jeremiah Marcellino, Goutam Kar, Yarjan Samad, Boyang Mao, Renzo Vanna, Franziska Hoffmann, Orlando Guntinas-Lichius, Sze Yun Set, Andrea C. Ferrari, Giulio Cerullo, Matteo Negro.
Biomedical Vibrational Spectroscopy 2024: Advances in Research and Industry – Proceedings of SPIE. April 2024
DOI: 10.1117/12.3002592
Repository link: https://zenodo.org/records/13832117
We combine an all-fiber dual wavelength, self-synchronized laser and a dedicated multi-channel detection unit to perform state-of-the-art multiplex Stimulated Raman Scattering (SRS) microscopy. The system covers the full CH spectrum in 1 μs reaching shot-noise limited performances with 25 μW per detection channel. This all-inone solution is based on a passively synchronized dual-wavelength laser source with shot-noise limited relative intensity noise from 600 kHz and a modular multi-channel lock-in detection unit. The synergistic design between laser source and detection system simplifies multiplex SRS implementation for real-time full-chemical imaging.
Raman spectroscopy of graphene and related materials
Anna Ott, Andrea Ferrari.
Encyclopedia of Condensed Matter Physics, 2nd ed, October 2023
DOI: 10.1016/B978-0-323-90800-9.00252-3
Direct link to the book chapter: https://www.graphene.cam.ac.uk/files/518.pdf
Raman spectroscopy is one of the main characterization techniques for graphene and related materials. It is a non-destructive technique that can give insight in the material’s quality, the number of layers, and is sensitive to any changes in electric or magnetic fields, band structure and temperature, making it ideal to probe layered materials.
Plug-and-play stimulated Raman microscopy system for broadband coherent vibrational imaging
Francesco Crisafi, Benedetta Talone, Andrea Ragni, Gabriele Di Noia, Mujeeb Rahman, Jing He, Jeremiah Marcellino, Goutam Kar, Yarjan Samad, Boyang Mao, Renzo Vanna, Franziska Hoffmann, Orlando Guntinas-Lichius, Silvia Ghislanzoni, Italia Bongarzone, Sze Yun Set, Andrea C. Ferrari, Giulio Cerullo, Matteo Negro.
In the European Conference on Lasers and Electro-Optics, June 2023
DOI: 10.1109/CLEO/Europe-EQEC57999.2023.10231815
Direct link to the conference proceeding: https://www.graphene.cam.ac.uk/files/520.pdf
Stimulated Raman scattering (SRS) microscopy is an emerging tool for biomedical imaging, with applications ranging from cell-drug interaction, cell sorting to tissue analysis and histopathology. Current commercial systems require users to deal with integration between laser, microscope and the detection system, resulting in trade-offs and, most importantly, lack of user-friendliness, thus hindering widespread application by non-specialists. This approach has pushed the laser market towards the development of high power (>100mW, >1nJ/pulse) narrowband (<1nm) tunable sources (Optical Parametric Oscillators (OPOs) and Fiber-OPOs) which can be coupled to off-the-shelf photodiodes and lock-in amplifiers, at the cost of detecting one or maximum two frequencies at a time
High-Resolution Raman Imaging of >300 Patient-Derived Cells from Nine Different Leukemia Subtypes: A Global Clustering Approach
Renzo Vanna, Andrea Masella, Manuela Bazzarelli, Paola Ronchi, Aufried Lenferink, Cristina Tresoldi, Carlo Morasso, Marzia Bedoni, Giulio Cerullo, Dario Polli, Fabio Ciceri, Giulia De Poli, Matteo Bregonzio, Cees Otto.
Analytical Chemistry, May 2024
Leukemia comprises a diverse group of bone marrow tumors marked by cell proliferation. Current diagnosis involves identifying leukemia subtypes through visual assessment of blood and bone marrow smears, a subjective and time-consuming method. Our study introduces the characterization of different leukemia subtypes using a global clustering approach of Raman hyperspectral maps of cells. We analyzed bone marrow samples from 19 patients, each presenting one of nine distinct leukemia subtypes, by conducting high spatial resolution Raman imaging on 319 cells, generating over 1.3 million spectra in total. An automated preprocessing pipeline followed by a single-step global clustering approach performed over the entire data set identified relevant cellular components (cytoplasm, nucleus, carotenoids, myeloperoxidase (MPO), and hemoglobin (HB)) enabling the unsupervised creation of high-quality pseudostained images at the single-cell level. Furthermore, this approach provided a semiquantitative analysis of cellular component distribution, and multivariate analysis of clustering results revealed the potential of Raman imaging in leukemia research, highlighting both advantages and challenges associated with global clustering.
Compact terahertz harmonic generation in the Reststrahlenband using a graphene-embedded metallic split ring resonator array
Alessandra Di Gaspare, Chao Song, Chiara Schiattarella, Lianhe H. Li, Mohammed Salih, A. Giles Davies, Edmund H. Linfield, Jincan Zhang, Osman Balci, Andrea C. Ferrari, Sukhdeep Dhillon, Miriam S. Vitiello
Nature Communications , March 2024
Harmonic generation is a result of a strong non-linear interaction between light and matter. It is a key technology for optics, as it allows the conversion of optical signals to higher frequencies. Owing to its intrinsically large and electrically tunable non-linear optical response, graphene has been used for high harmonic generation but, until now, only at frequencies < 2 THz, and with high-power ultrafast table-top lasers or accelerator-based structures. Here, we demonstrate third harmonic generation at 9.63 THz by optically pumping single-layer graphene, coupled to a circular split ring resonator (CSRR) array, with a 3.21 THz frequency quantum cascade laser (QCL). Combined with the high graphene nonlinearity, the mode confinement provided by the optically-pumped CSRR enhances the pump power density as well as that at the third harmonic, permitting harmonic generation. This approach enables potential access to a frequency range (6-12 THz) where compact sources remain difficult to obtain, owing to the Reststrahlenband of typical III-V semiconductors.
Controlled Growth of Single-Crystal Graphene Wafers on Twin-Boundary-Free Cu(111) Substrates
Yeshu Zhu, Jincan Zhang, Ting Cheng, Jilin Tang, Hongwei Duan, Zhaoning Hu, Jiaxin Shao, Shiwei Wang, Mingyue Wei, Haotian Wu, Ang Li, Sheng Li, Osman Balci, Sachin M. Shinde, Hamideh Ramezani, Luda Wang, Li Lin, Andrea C. Ferrari, Boris I. Yakobson, Hailin Peng, Kaicheng Jia, Zhongfan Liu.
Advanced Materials, October 2023
Direct link to the publication: https://www.graphene.cam.ac.uk/files/555-compressed.pdf
Single-crystal graphene (SCG) wafers are needed to enable mass-electronics and optoelectronics owing to their excellent properties and compatibility with silicon-based technology. Controlled synthesis of high-quality SCG wafers can be done exploiting single-crystal Cu(111) substrates as epitaxial growth substrates recently. However, current Cu(111) films prepared by magnetron sputtering on single-crystal sapphire wafers still suffer from in-plane twin boundaries, which degrade the SCG chemical vapor deposition. Here, it is shown how to eliminate twin boundaries on Cu and achieve 4 in. Cu(111) wafers with ≈95% crystallinity. The introduction of a temperature gradient on Cu films with designed texture during annealing drives abnormal grain growth across the whole Cu wafer. In-plane twin boundaries are eliminated via migration of out-of-plane grain boundaries. SCG wafers grown on the resulting single-crystal Cu(111) substrates exhibit improved crystallinity with >97% aligned graphene domains. As-synthesized SCG wafers exhibit an average carrier mobility up to 7284 cm2 V−1 s−1 at room temperature from 103 devices and a uniform sheet resistance with only 5% deviation in 4 in. region.
Design and test of a rigid endomicroscopic system for multimodal imaging and femtosecond laser ablation
Chenting Lai, Matteo Calvarese, Karl Reichwald, Hyeonsoo Bae, Mohammadsadegh Vafaeinezhad, Tobias Meyer-Zedler, Franziska Hoffmann, Anna Mühlig, Tino Eidam, Fabian Stutzki, Bernhard Messerschmidt, Herbert Gross, Michael Schmitt, Orlando Guntinas-Lichius, Jürgen Popp
Journal of biomedical optics, June 2023
Significance
Conventional diagnosis of laryngeal cancer is normally made by a combination of endoscopic examination, a subsequent biopsy, and histopathology, but this requires several days and unnecessary biopsies can increase pathologist workload. Nonlinear imaging implemented through endoscopy can shorten this diagnosis time, and localize the margin of the cancerous area with high resolution.
Aim
Develop a rigid endomicroscope for the head and neck region, aiming for in-vivomultimodal imaging with a large field of view (FOV) and tissue ablation.
Approach
Three nonlinear imaging modalities, which are coherent anti-Stokes Raman scattering, two-photon excitation fluorescence, and second harmonic generation, as well as the single photon fluorescence of indocyanine green, are applied for multimodal endomicroscopic imaging. High-energy femtosecond laser pulses are transmitted for tissue ablation.
Results
This endomicroscopic system consists of two major parts, one is the rigid endomicroscopic tube 250 mm in length and 6 mm in diameter, and the other is the scan-head (10×12×6 cm3 in size) for quasi-static scanning imaging. The final multimodal image accomplishes a maximum FOV up to 650 μm, and a resolution of 1 μm is achieved over 560 μm FOV. The optics can easily guide sub-picosecond pulses for ablation.
Conclusions
The system exhibits large potential for helping real-time tissue diagnosis in surgery, by providing histological tissue information with a large FOV and high resolution, label-free. By guiding high-energy fs laser pulses, the system is even able to remove suspicious tissue areas, as has been shown for thin tissue sections in this study.
Short pulse generation from a graphene-coupled passively mode-locked terahertz laser
Riccardi E, Pistore V, Kang S, Seitner L, De Vetter A, Jirauschek C, Mangeney J, Li L, Davies AG, Linfield EH, Ferrari AC
Nature Photonics, April 2023
DOI: 10.1038/s41566-023-01195-z
Direct link to the publication: https://www.graphene.cam.ac.uk/files/506.pdf
The generation of stable trains of ultrashort (femtosecond to picosecond), terahertz-frequency radiation pulses with large instantaneous intensities is an underlying requirement for the investigation of light–matter interactions for metrology and ultrahigh-speed communications. In solid-state electrically pumped lasers, the primary route to generate short pulses is through passive mode-locking; however, this has not yet been achieved in the terahertz range, defining one of the longest standing goals over the past two decades. In fact, the realization of passive mode-locking has long been assumed to be inherently hindered by the fast recovery times associated with the intersubband gain of terahertz lasers. Here we demonstrate a self-starting miniaturized short pulse terahertz laser, exploiting an original device architecture that includes the surface patterning of multilayer-graphene saturable absorbers distributed along the entire cavity of a double-metal semiconductor 2.30–3.55 THz wire laser. Self-starting pulsed emission with 4.0-ps-long pulses is demonstrated in a compact, all-electronic, all-passive and inexpensive configuration.
Control of Raman Scattering Quantum Interference Pathways in Graphene
Chen X, Reichardt S, Lin ML, Leng YC, Lu Y, Wu H, Mei R, Wirtz L, Zhang X, Ferrari AC, Tan PH
ACS nano, March 2023
Graphene is an ideal platform to study the coherence of quantum interference pathways by tuning doping or laser excitation energy. The latter produces a Raman excitation profile that provides direct insight into the lifetimes of intermediate electronic excitations and, therefore, on quantum interference, which has so far remained elusive. Here, we control the Raman scattering pathways by tuning the laser excitation energy in graphene doped up to 1.05 eV. The Raman excitation profile of the G mode indicates its position and full width at half-maximum are linearly dependent on doping. Doping-enhanced electron–electron interactions dominate the lifetimes of Raman scattering pathways and reduce Raman interference. This will provide guidance for engineering quantum pathways for doped graphene, nanotubes, and topological insulators.
Phononics of graphene, layered materials, and heterostructures
Self‐Induced Mode‐Locking in Electrically Pumped Far‐Infrared Random Lasers
Di Gaspare A, Pistore V, Riccardi E, Pogna EA, Beere HE, Ritchie DA, Li L, Davies AG, Linfield EH, Ferrari AC, Vitiello MS
Advanced Science, January 2023
Mode locking, the self-starting synchronous oscillation of electromagnetic modes in a laser cavity, is the primary way to generate ultrashort light pulses. In random lasers, without a cavity, mode-locking, the nonlinear coupling amongst low spatially coherent random modes, can be activated via optical pumping, even without the emission of short pulses. Here, by exploiting the combination of the inherently giant third-order χ(3) nonlinearity of semiconductor heterostructure lasers and the nonlinear properties of graphene, the authors demonstrate mode-locking in surface-emitting electrically pumped random quantum cascade lasers at terahertz frequencies. This is achieved by either lithographically patterning a multilayer graphene film to define a surface random pattern of light scatterers, or by coupling on chip a saturable absorber graphene reflector. Intermode beatnote mapping unveils self-induced phase-coherence between naturally incoherent random modes. Self-mixing intermode spectroscopy reveals phase-locked random modes. This is an important milestone in the physics of disordered systems. It paves the way to the development of a new generation of miniaturized, electrically pumped mode-locked light sources, ideal for broadband spectroscopy, multicolor speckle-free imaging applications, and reservoir quantum computing.
Intraoperative Assessment of Tumor Margins in Tissue Sections with Hyperspectral Imaging and Machine Learning
David Pertzborn, Hoang-Ngan Nguyen, Katharina Hüttmann, Jonas Prengel, Günther Ernst, Orlando Guntinas-Lichius, Ferdinand von Eggeling, Franziska Hoffmann
Cancers, December 2022
The complete resection of the malignant tumor during surgery is crucial for the patient’s survival. To date, surgeons have been intraoperatively supported by information from a pathologist, who performs a frozen section analysis of resected tissue. This tumor margin evaluation is subjective, methodologically limited and underlies a selection bias. Hyperspectral imaging (HSI) is an established and rapid supporting technique. New artificial-intelligence-based techniques such as machine learning (ML) can harness this complex spectral information for the verification of cancer tissue. We performed HSI on 23 unstained tissue sections from seven patients with oral squamous cell carcinoma and trained the ML model for tumor recognition resulting in an accuracy of 0.76, a specificity of 0.89 and a sensitivity of 0.48. The results were in accordance with the histopathological annotations and do, therefore, enable the delineation of tumor margins with high speed and accuracy during surgery.
A novel interpretable machine learning algorithm to identify optimal parameter space for cancer growth
Coggan Helena, Andres Terre Helena, Liò Pietro.
Frontiers in Big Data, September 2022
Recent years have seen an increase in the application of machine learning to the analysis of physical and biological systems, including cancer progression. A fundamental downside to these tools is that their complexity and nonlinearity makes it almost impossible to establish a deterministic, a priori relationship between their input and output, and thus their predictions are not wholly accountable. We begin with a series of proofs establishing that this holds even for the simplest possible model of a neural network; the effects of specific loss functions are explored more fully in Appendices. We return to first principles and consider how to construct a physics-inspired model of tumor growth without resorting to stochastic gradient descent or artificial nonlinearities. We derive an algorithm which explores the space of possible parameters in a model of tumor growth and identifies candidate equations much faster than a simulated annealing approach. We test this algorithm on synthetic tumor-growth trajectories and show that it can efficiently and reliably narrow down the area of parameter space where the correct values are located. This approach has the potential to greatly improve the speed and reliability with which patient-specific models of cancer growth can be identified in a clinical setting.