Periodic Reporting for period 1 - CHARM (Chemometric histopathology via coherent Raman imaging for precision medicine)
Período documentado: 2022-05-01 hasta 2023-04-30
CHARM aims to develop a medical device based on high-speed, low-cost Raman digital imaging technology and artificial intelligence to transform cancer diagnosis and treatment.
CHARM addresses two limitations of digital pathology: first, the need for tissue staining, which is time-consuming and introduces variability between samples, and the inability to measure the tissue molecular composition, which can be crucial to determine tumor subtypes and grades. Our approach reduces the pathologist’s workload by avoiding staining and by including a prescreening performed by the AI agent. The pathologist is provided with a decision tree (allowing interpretability) and a diagnostic report that supports the differentiation between normal and tumor tissue and includes suggestions for cancer type and grade classification. The instrument will offer histopathologists a reliable, fast, and low-cost Clinical Decision Support System for cancer diagnosis and personalized cancer therapy.
Our aim is to develop a Class C (In-Vitro Diagnostic Regulation) medical device consisting of 1) a turnkey low-cost broadband Coherent Raman Scattering microscope (enabled by our patented graphene-based fiber laser technology); 2) an Artificial Intelligence model based on deep learning, statistics, and machine learning. The instrument will offer histopathologists a reliable, fast, and low-cost Clinical Decision Support System for cancer diagnosis and personalized cancer therapy.
CHARM addresses two limitations of digital pathology: first, the need for tissue staining, which is time-consuming and introduces variability between samples, and the inability to measure the tissue molecular composition, which can be crucial to determine tumor subtypes and grades. Our approach reduces the pathologist’s workload by avoiding staining and by including a prescreening performed by the AI agent. The pathologist is provided with a decision tree (allowing interpretability) and a diagnostic report that supports the differentiation between normal and tumor tissue and includes suggestions for cancer type and grade classification. The instrument will offer histopathologists a reliable, fast, and low-cost Clinical Decision Support System for cancer diagnosis and personalized cancer therapy.
Our aim is to develop a Class C (In-Vitro Diagnostic Regulation) medical device consisting of 1) a turnkey low-cost broadband Coherent Raman Scattering microscope (enabled by our patented graphene-based fiber laser technology); 2) an Artificial Intelligence model based on deep learning, statistics, and machine learning. The instrument will offer histopathologists a reliable, fast, and low-cost Clinical Decision Support System for cancer diagnosis and personalized cancer therapy.
The laser source has been patented (currently under a confidentiality period) and the related product, called STRALE, is expected to be commercialized in the second half of 2023.
A working prototype of the Coherent Raman Scattering microscope has been built and used to collect the first images from biopsies. The prototype is being optimized to integrate the recently developed multichannel lock-in amplifier and reduce acquisition time.
Furthermore, we collected a reference panel of samples from patients with head and neck cancer, including the associated clinical data and pathohistological annotations. Thanks to this data, we have been able to develop an AI-based model that can virtually stain images to replicate the current appearance of samples used for diagnosis and we are currently working on a model to distinguish and locate the cancerous tissue from the healthy one.
Finally, the saturable absorber has been designed and the final optimization should be ready soon.
A working prototype of the Coherent Raman Scattering microscope has been built and used to collect the first images from biopsies. The prototype is being optimized to integrate the recently developed multichannel lock-in amplifier and reduce acquisition time.
Furthermore, we collected a reference panel of samples from patients with head and neck cancer, including the associated clinical data and pathohistological annotations. Thanks to this data, we have been able to develop an AI-based model that can virtually stain images to replicate the current appearance of samples used for diagnosis and we are currently working on a model to distinguish and locate the cancerous tissue from the healthy one.
Finally, the saturable absorber has been designed and the final optimization should be ready soon.
The CHARM technology will be capable of automatically analyzing unstained tissues, providing fast and accurate tumor identification (differentiating normal vs neoplastic tissues) with high accuracy and final tumor diagnosis prediction (differentiating and grading histologic subtypes), thus offering to the histopathologist a decision tree compatible with existing clinical protocols but with biomolecular-based objectivity and reduced time to result.
During CHARM, we will prove the feasibility of Coherent Raman Scattering microscopy for the diagnosis of a representative case: Head and Neck Cancer. When diagnosed early, head and neck cancers can be treated more easily and the chances of survival increase tremendously. However, currently, 2 in 3 of all head and neck cancer patients are diagnosed at an advanced stage. Head and Neck cancer can be used as a model for a larger variety of cancers characterized by field cancerization: esophageal, gastrointestinal, non-melanoma skin, non-small cell lung, and most of the breast cancers.
During CHARM, we will prove the feasibility of Coherent Raman Scattering microscopy for the diagnosis of a representative case: Head and Neck Cancer. When diagnosed early, head and neck cancers can be treated more easily and the chances of survival increase tremendously. However, currently, 2 in 3 of all head and neck cancer patients are diagnosed at an advanced stage. Head and Neck cancer can be used as a model for a larger variety of cancers characterized by field cancerization: esophageal, gastrointestinal, non-melanoma skin, non-small cell lung, and most of the breast cancers.