MIT recently released an AI system that can detect melanoma. It uses a deep convolutional neural network (DCCN) to quickly analyze a wide-area image of the patient’s skin, which can help to detect early skin cancer efficiently.
The principle is to first use a smartphone camera to take photos for the patient’s skin. The system detects, extracts and analyzes all the pigmented skin lesions that can be observed in the picture. With the help of pre-trained deep convolutional neural network (DCCN) the system determines the suspiciousness of a single pigmented lesion and marks it. The extracted features are used for further evaluation of pigmented lesions, and the results are displayed in a heat map format.
The system was trained by 20,388 images from 133 patients at the Hospital Gregorio Marañón in Madrid and a number of publicly available images. It distinguished more than 90.3% of SPLs from nonsuspicious lesions, skin, and complex backgrounds, avoiding the cumbersome and time-consuming single lesion imaging.
“Our research suggests that systems leveraging computer vision and deep neural networks, quantifying such common signs, can achieve comparable accuracy to expert dermatologists. We hope our research revitalizes the desire to deliver more efficient dermatological screenings in primary care settings to drive adequate referrals.” The researchers explained.
The development of AI technology has overturned the old medical landscape and created a new medical decision-making system. At present, the application of AI in the medical field is mainly concentrated in medical imaging, auxiliary diagnosis and treatment, health management, hospital management, disease prediction, and drug research etc., covering multiple clinical aspects such as screening, diagnosis, treatment, and prognosis management.
Relying on its own technical advantages and rich data processing experience, Datatang provides professional data customization services for global medical enterprises and institutions.
Medical Data Annotation Service
● Cerebral Hemorrhage Image Annotation
This project is to mark the bleeding points in the image of the CT scan of the brain. The image format is a special format dedicated to the medical field. The professional medical marking tool ITK-snap is used for annotation.
● Throat Follicle Image Annotation
This project is to label the throat follicles. Each photo contains 4 kinds of labels.
● Brain Cell Image Annotation
This project is to annotate brain cell images. Each group of images contains brain cells and keratinocytes. The brain cells are labelled with red circles containing small green dots, and the keratinocytes are labelled with large green dots.
Off-the-Shelf Medical Datasets
Datatang has accumulated a number of medical datasets which can quickly optimize your AI models. Datatang strictly abides by relevant regulations, and all the data is collected with proper authorization agreement.
● 6,333 Chest Radiographies Data
70% of the data is chest radiographies of healthy people, the rest is chest radiographies of patients with pneumonia. The images were collected by Philips Medical Systems.
● 5,105 images Human Facial Skin Defects Data
The data includes five skin defects: acne, acne marks, pigmentation, wrinkles, and dark circles. The age distribution ranges from teenagers to old age, mainly young and middle-aged.
● 203,029 Groups — Medical Question Answering Data
The data contains 203,029 groups question and answer data between doctors and patients of different diseases, including the disease category and the process of conversation.
Recently, in a seminar hosted by NEJM paper, the participating experts pointed out that the development of AI medical systems cannot do without large-scale data. To develop the AI in healthcare, the processing and quality control of medical training data is the issue that requires urgent attention. Datatang is committed to providing professional and rigorous data services to better empower the transform of intelligent healthcare technology.
If you need data services, please don’t hesitate to contact us: firstname.lastname@example.org