“Doctor, what is the condition of my disease? Is it benign or malignant?” “I have been to several hospitals for this disease, and the results of each diagnosis are inconsistent. Why is this? What’s going on?” “Will I relapse after treatment, and how long can I live?” These thorny questions are often asked by patients to clinicians, and usually occur in the radiologist’s consultation and review.
To answer the above questions, in addition to “pathology” which is the gold standard for diagnosis, CT examination also plays an important role. The basic principle of CT imaging is that the X-ray beam irradiates the tissues and organs of different parts of the human body to be examined, and different attenuation occurs due to differences in density and thickness, resulting in the detector receiving the attenuated X-rays, which are converted into different grayscales after signal conversion. Corresponding CT images of human tissues and organs. When a patient’s body has lesions, the pathological tissue of different densities can also be detected by CT equipment. This is the basic principle that CT can detect lesions.
Generally speaking, CT is a diagnostic tool, and its image quality is the core index to help doctors make a diagnosis. Therefore, the development of CT images generally follows how to improve image quality while reducing radiation. To this end, leading players in the industry have been committed to introducing advanced technologies such as deep learning and energy spectrum imaging into CT to improve its efficiency.
According to Sun Xuguang, Chief Technology Officer of GE Healthcare China, CT images have generally gone through three stages of development: from the earliest filtered back-projection stage, the images are similar to the early TV analog signals. A lot of snowflakes and noise; in the second stage iteratively, the analog signal slowly turned into a high-definition signal; the third stage after 2019 introduced a deep learning algorithm, and the intuitive effect is “as if TVs are now entering the 4K and 8K era”, can See more detail, better tissue boundaries, and less noise.
TrueFidelity™ GSI, independently developed by GE Healthcare, is an important embodiment of this idea. It advocates to “restore the true image of the image”, combining the truth spectrum image technology and CT tomography Data image restoration technology is combined. The technology was showcased intensively at the World Artificial Intelligence Conference held in September this year.
On-site shooting by reporters
It is reported that TrueFidelity™ is GE Healthcare’s deep learning image reconstruction system. Compared with traditional algorithms, TrueFidelity™ can continuously learn high-dose high-definition images and cooperate with GE Healthcare’s related hardware technology , which can reconstruct the original data obtained under low radiation dose into high-quality images and restore the truth of the images.
GSI is an energy spectrum imaging technology. It breaks through the boundary of conventional CT, which has only a single CT value, and can improve the display of small structures. Lesion, identify the material components, make more accurate qualitative diagnosis of benign and malignant, and can carry out quantitative analysis and tracking of material components. “To take a famous example, using CT images to scan a glass of water, what we see is only water; but using CT energy spectrum imaging, we can see whether the glass of water is sugar water or saline, and get richer information.”
TrueFidelity™ GSI, which combines energy spectral imaging (GSI) and deep learning architecture, is to help Apex CT equipment, regardless of the raw data of conventional scans or energy spectral scans, can learn from Results The noise can be judged autonomously, especially for early tumor lesions. TrueFidelity™ GSI makes the distinction between mildly enhanced lesions and surrounding tissue structures clearer, thereby helping the clinic move forward the time window for early diagnosis and early treatment of serious diseases for patients. It is reported that the technology also adopts the technical method of instantaneous switching of high and low pressure to accurately control the exposure of the tube, help doctors reduce visual fatigue, and thus facilitate accurate diagnosis of diseases.
In layman’s terms, AI algorithm (deep learning) helps to improve image quality, energy spectrum technology helps to find more details of lesions, and the two “buff superposition” help CT to exert the maximum function, Solve more complex clinical problems; in addition, patients who were originally unable to undergo energy spectrum CT examination due to their special body shape can also obtain high-quality energy spectrum examination images.
“For example, if a tumor biopsy is to be performed, but the internal components of the tumor are actually heterogeneous, if a biochemical tissue is taken from an area inside the tumor for biopsy, it may not be able to obtain the most accurate results. the result of”. Therefore, biopsy is bound to rely on other auxiliary diagnostic methods. TrueFidelity™ GSI provides a corresponding solution, that is, “based on more accurate image information, it helps pathological biopsy to collect the really needed tissue samples”, in order to obtain more accurate tumor diagnosis results.
At present, the TrueFidelity™ GSI technology has been certified by China’s NMPA, and the related equipment equipped with this technology has also been used in Shanghai Ruijin Hospital and is being gradually promoted. Sun Xuguang also said that this is just the tip of the iceberg in the application of AI technology in the field of medical imaging. He believes that in the future, AI technology can also better help the development of the entire medical technology, not only diagnosis, but even treatment. For invasive surgery applications, the idea is to allow patients to receive treatment with minimal harm, all of which depend on the development of intelligent imaging technology.
It is certain that AI technology not only brings us convenience at our fingertips, but also accelerates the transformation and implementation of traditional medical scenarios, maximizing the effect of medical resources, It also improves people’s health and improves the prognosis of patients. With the advent of more AI technologies such as TrueFidelity™ GSI technology, AI will “encounter” tumor diagnosis and treatment and precision medicine, which will also usher in a new round of medical revolution.