In recent years, there has been a massive advancement in the available treatments for colon cancer. To ensure these therapies, such as immunotherapies, are effective, it is important to accurately diagnose the individual patient for providing specifically tailored treatment. Now, researchers have paired artificial intelligence (AI) with infrared (IR) imaging to develop an automated and precise method for diagnosing colon cancer and tailoring treatments to the patient. This label-free and automated technique complements existing methods for analyzing tissue samples. Over the course of the past several years, a research team at the Centre for Protein Diagnostics (PRODI) at Ruhr University Bochum (Bochum, Germany) has been working on creating a new digital imaging method known as label-free IR imaging. This method measures the genomic and proteomic composition of the examined tissue, providing molecular information based on the infrared spectra. The information is then decoded using AI and displayed as false-color images utilizing image analysis methods from the field of deep learning. The PRODI team successfully demonstrated that using deep neural networks, it was possible to effectively determine the microsatellite status, a prognostically and therapeutically relevant parameter, in colon cancer. In this process, the tissue sample passes through a standardized, user-independent, automated process and allows for spatially resolved differential classification of the tumor within an hour. On the other hand, classical diagnostics is used to determine the microsatellite status either through complex immunostaining of various proteins or via DNA analysis. The ever-improving therapy options have made fast and uncomplicated determination of such biomarkers extremely important. Based on IR microscopic data, the researchers modified, optimized, and trained neuronal networks to establish label-free diagnostics. In contrast to immunostaining, the new approach does not need dyes and is much faster than DNA analysis. “We were able to show that the accuracy of IR imaging for determining microsatellite status comes close to the most common method used in the clinic, immunostaining,” said PhD student Stephanie Schörner. Related Links:Ruhr University Bochum 

Precise classification of histopathological images is crucial to computer-aided diagnosis in clinical practice. Magnification-based learning networks have attracted considerable attention for their ability to improve performance in histopathological classification. However, the fusion of pyramids of histopathological images at different magnifications is an under-explored area. In this paper, we proposed a novel deep multi-magnification similarity learning (DSML) approach that can be useful for the interpretation of multi-magnification learning framework and easy to visualize feature representation from low-dimension (e.g., cell-level) to high-dimension (e.g., tissue-level), which has overcome the difficulty of understanding cross-magnification information propagation. It uses a similarity cross entropy loss function designation to simultaneously learn the similarity of the information among cross-magnifications. In order to verify the effectiveness of DMSL, experiments with different network backbones and different magnification combinations were designed, and its ability to interpret was also investigated through visualization. Our experiments were performed on two different histopathological datasets: a clinical nasopharyngeal carcinoma and a public breast cancer BCSS2021 dataset. The results show that our method achieved outstanding performance in classification with a higher value of area under curve, accuracy, and F-score than other comparable methods. Moreover, the reasons behind multi-magnification effectiveness were discussed.