Our Mission – Early and accurate diagnosis of Oral Cancer using AI (machine and deep learning)

The number of deaths from oral cancers in Canada is staggering— it will cause over 1,500 deaths, killing roughly 4 people per day. Of those 5,400 newly diagnosed individuals in Canada in 2021, only slightly more than half will be alive in 5 years. (Approximately 57%) . Our team at EOCD is driven and motivated to help fight these various forms of Oral cancers in Canada using advanced and innovative software tools for fast and effective early diagnosis and treatment of Oral Cancer patients. We hope to contribute to the fight for eradicating and preventing cancers by collaborating with medical professionals and researchers worldwide.

Using Deep Learning (DL) and Machine Learning (ML) platforms, we have developed specialized algorithmic solutions to analyze cytology images for faster and better assessment than human accuracy. Our solutions, which are both in API and Web Platforms, are designed to offer faster and better access for second and third diagnostic opinions by medical professionals to help develop a therapeutic approach quickly. This will help the patients and physicians by reducing anxiety caused by not knowing what they are dealing with and will result in selecting beat treatment method personalized by patient’s situation.

Our solutions do all the work—considers the type of cancer patient is dealing with, identify correlation between cancer and individual cell behavior, discovers patterns and trends, and analysis it using various algorithms. It’s a lot but not enough as we work to expand its reach and provide treatment advice for our patient using our platform.

You focus on savings lives, We focus on early diagnosis!

Optimized by TECHNOLOGY, uncovered by INNOVATION.

EOCD turns your practice into an accurate oral cancer pathology lab.

Technologies involved

EOCD’s technology is based upon the application of AI and Deep Learning approaches. Our company presents a clinical diagnostic tool to identify oral cancer using cytology images. We have developed and evaluated a deep learning-based algorithm that provides fully automated detection and screening of the oral cancer using cytology images. Our product has the potential to be used in dental clinics, hospitals and pathology centres for fast, accurate and reproducible assessment of cytology images of different types of oral cancer.

Our platform starts with histology images and their corresponding annotation mask that is obtained from experienced pathologist for training phase. We used a 2D Gaussian filter to generate fuzzy ground truth. Slide scanners do not provide sufficiently good focus for cytological samples and a focus selection step was applied to solve this. We developed a deep learning U-Net Convolutional Neural Network (CNN) followed by binary classifier to delineate the nucleus and detect the tumor malignancy. Next, we developed a Convolutional Neural Network model (U-Net Network Architecture) to analyse image dataset and detect nucleus. The model utilizes deep learning libraries of python such as Keras and Tensor flow. We trained a U-Net Convolutional Neural Network (CNN) using original images and them correspond fuzzy labels.

Although brush biopsy to obtain cytology slides is currently used as an assessment method in Oral Cancer, it goes ‎without saying that machine analysis of Cytology images through data mining and machine learning is ‎more accurate and reliable compared to human led analysis, as the software is able to ‎detect the slightest changes in individual cell patterns which are not generally discerned by the ‎unaided eye.

Based upon the result of the cytology image analysis, the practitioners will be able to diagnose ‎oral cancer and pre-cancer in the early stages of the disease and plan the anticancer treatment ‎accordingly.