Raman spectroscopy shows great potential as a diagnostic tool for thyroid cancer due to its ability to detect biochemical changes during cancer development. This technique is particularly valuable because it is non-invasive and label/dye-free. Compared to molecular tests, Raman spectroscopy analyses can more effectively discriminate malignant features, thus reducing unnecessary surgeries. However, one major hurdle to using Raman spectroscopy as a diagnostic tool is the identification of significant patterns and peaks. In this study, we propose a Machine Learning procedure to discriminate healthy/benign versus malignant nodules that produces interpretable results. We collect Raman spectra obtained from histological samples, select a set of peaks with a data-driven and label independent approach and train the algorithms with the relative prominence of the peaks in the selected set. The performance of the considered models, quantified by area under the Receiver Operating Characteristic curve, exceeds 0.9. To enhance the interpretability of the results, we employ eXplainable Artificial Intelligence and compute the contribution of each feature to the prediction of each sample.

Over the last 50 years, the incidence of human thyroid cancer disease has seen a significative increment. This comes along with an even higher increment of surgery, since, according to the international guidelines, patients are sometimes addressed to surgery also when the fine needle aspiration gives undetermined cytological diagnosis. As a matter of fact, only 30% of the thyroid glands removed for diagnostic purpose have a post surgical histological report of malignancy: this implies that about 70% of the patients have suffered an unnecessary thyroid removal. Here we show that Raman spectroscopy investigation of thyroid tissues provides reliable cancer diagnosis. Healthy tissues are consistently distinguished from cancerous ones with an accuracy of $$\sim $$∼ 90%, and the three cancer typology with highest incidence are clearly identified. More importantly, Raman investigation has evidenced alterations suggesting an early stage of transition of adenoma tissues into cancerous ones. These results suggest that Raman spectroscopy may overcome the limits of current diagnostic tools.

Ultrastructural fluctuations at nanoscale are fundamental to assess properties and functionalities of advanced out-of-equilibrium materials. We have taken myelin as a model of supramolecular assembly in out-of-equilibrium living matter. Myelin sheath is a simple stable multi-lamellar structure of high relevance and impact in biomedicine. Although it is known that myelin has a quasi-crystalline ultrastructure there is no information on its fluctuations at nanoscale in different states due to limitations of the available standard techniques. To overcome these limitations, we have used Scanning micro X-ray Diffraction, which is a non-invasive probe of both reciprocal and real space to visualize statistical fluctuations of myelin order of the sciatic nerve of Xenopus Laevis. The results show that the ultrastructure period of the myelin is stabilized by large anti-correlated fluctuations at nanoscale, between hydrophobic and hydrophilic layers. The ratio between the total thickness of hydrophilic and hydrophobic layers defines the conformational parameter, which describes the different states of myelin. Our key result is that myelin in its out-of-equilibrium functional state fluctuates point-to-point between different conformations showing a correlated disorder described by a Levy distribution. As the system approaches the thermodynamic equilibrium in an aged state the disorder loses its correlation degree and the structural fluctuation distribution changes to Gaussian. In a denatured state at low pH, it changes to a completely disordered stage. Our results clarify also the degradation mechanism in biological systems by associating these states with variation of the ultrastructural dynamic fluctuations at nanoscale