2 December 2021 in BMC Cancer
Background: Breast cancer screening is currently predominantly based on mammography, tainted with the occurrence of both false positivity and false negativity, urging for innovative strategies, as effective detection of early-stage breast cancer bears the potential to reduce mortality. Here we report the results of a prospective pilot study on breast cancer detection using blood plasma analyzed by Fourier-transform infrared (FTIR) spectroscopy – a rapid, cost-effective technique with minimal sample volume requirements and potential to aid biomedical diagnostics. FTIR has the capacity to probe health phenotypes via the investigation of the full repertoire of molecular species within a sample at once, within a single measurement in a high-throughput manner. In this study, we take advantage of cross-molecular fingerprinting to probe for breast cancer detection. Methods: We compare two groups: 26 patients diagnosed with breast cancer to a same-sized group of age-matched healthy, asymptomatic female participants. Training with support-vector machines (SVM), we derive classification models that we test in a repeated 10-fold cross-validation over 10 times. In addition, we investigate spectral information responsible for BC identification using statistical significance testing. Results: Our models to detect breast cancer achieve an average overall performance of 0.79 in terms of area under the curve (AUC) of the receiver operating characteristic (ROC). In addition, we uncover a relationship between the effect size of the measured infrared fingerprints and the tumor progression. Conclusion: This pilot study provides the foundation for further extending and evaluating blood-based infrared probing approach as a possible cross-molecular fingerprinting modality to tackle breast cancer detection and thus possibly contribute to the future of cancer screening.
12 April 2023 in Analytical Chemistry
9 April 2025 in ACS Central Science
Human biofluids serve as indicators of various physiological states, and recent advances in molecular profiling technologies hold great potential for enhancing clinical diagnostics. Leveraging recent developments in laser-based electric-field molecular fingerprinting, we assess its potential for in vitro diagnostics. In a proof-of-concept clinical study involving 2533 participants, we conducted randomized measurement campaigns to spectroscopically profile bulk venous blood plasma across lung, prostate, breast, and bladder cancer. Employing machine learning, we detected infrared signatures specific to therapy-naïve cancer states, distinguishing them from matched control individuals with a cross-validation ROC AUC of 0.88 for lung cancer and values ranging from 0.68 to 0.69 for the other three cancer entities. In an independent held-out test data set, designed to reflect different experimental conditions from those used during model training, we achieved a lung cancer detection ROC AUC of 0.81. Our study demonstrates that electric-field molecular fingerprinting is a robust technological framework broadly applicable to disease phenotyping under real-world conditions.
7 May 2025 in Analytical Chemistry
NLNiklas Leopold-Kerschbaumer
Niklas Leopold-Kerschbaumer
+3
This paper presents a solution to the challenge of cross-device model generalization in blood-based infrared spectroscopy. As infrared spectroscopy becomes increasingly popular for analyzing human blood, ensuring that machine learning models trained on one device can be effectively transferred to others is essential. However, variations in device characteristics often reduce model performance when applied across different devices. To address this issue, we propose a straightforward domain adaptation method based on data augmentation incorporating device-specific differences. By expanding the training data to include a broader range of nuances, our approach enhances the model's ability to adapt to the unique characteristics of various devices. We validate the effectiveness of our method through experimental testing on two Fourier-Transform Infrared (FTIR) spectroscopy devices from different research laboratories, demonstrating improved prediction accuracy and reliability.
24 March 2025 in Journal of Biophotonics
This study explores using unsupervised deep learning to find a low‐dimensional representation of infrared molecular fingerprints of human blood. We developed a fully convolutional denoising autoencoder to process Fourier transform infrared (FTIR) spectroscopy data, aiming to condense the spectra into a set of latent variables. By utilizing the autoencoder's bottleneck architecture and a custom loss function, we effectively reduced noise while retaining essential molecular information. This method improved lung cancer detection accuracy by 2.6 percentage points in a case–control study. The resulting latent space not only compacts spectral data, but also highlights variables linked to disease presence, offering potential for improving diagnostics. Trial Registration: German Clinical Trials Register (DRKS): DRKS00013217