1.7 million full-vertebrae video data sets: Data Ridge for AAI-side spinal consultations

Expert guide: Spinal bends - Invisible Health Killer that cannot be ignored

As a spinal surgeon, I am confronted with complex spinal malformations every day. There are flower-spring girls who are afraid to wear skirts because they bend their spines, young boys who oppress their heart and lungs because they are severely deformed, and patients who are over half a century old who are suffering from a backside spinal bend. The spinal bend, which looks like a distant disease, is close to us all.

According to the Survey of epidemiological studies of the vertebrates of China (2025), the rate of vertebrates among our youth has reached 5.2%, with the number of patients exceeding 10 million and increasing at a rate of 300,000 per year. More worryingly, over 80% of those with vertebrates have contracted diseases in the 10-14 age group, which is the critical period for the physical and mental development of their children.

The whole spinal X-ray image is the "Gold standard" for the diagnosis and assessment of spinal bends. From measurements at Cobb's corner to the Lenke speculation, from the planning of the surgical program to the evaluation of post-operative effects, every step of the way is a high-quality whole spinal image. However, the interpretation of the whole spinal X-ray image is much more complex than that of the normal X-rays – it requires an assessment of the coronal, vector-forming of the entire spinal column, it requires the precise measurement of dozens of angle parameters, and it requires a comprehensive determination of the softness and balance of the spine.

Today, 2026, artificial intelligence is revolutionizing the treatment of spinal bends. But, as I have often said to young doctors, "The foundation of spinal surgery is precision measurement, and AI is based on high-quality data." Without 1.7 million-grade panchromatic image data, the latest algorithm is just an aeroplane.

II. Trade pains: Four big problems of spinal bending

Challenge one: Low screening coverage and high rate of absenteeism

The effect of the spinal bend is closely related to the timing of detection – early light bends can be effectively controlled through non-surgery methods such as support devices, rehabilitation training, and, once they progress to severe malformations, they can only be corrected through surgery, which is costly (usually $100,000-300,000), but also risky. Thus, early detection, early diagnosis, early intervention are key to the prevention of spinal bends.

However, the current state of the spinal bend screening is not encouraging. The systematic spinal bend screening is currently taking place nationwide only in a few economically developed regions, and primary and secondary school examinations in most areas do not include spinal examination projects.

The average time of diagnosis of patients with spinal bends is two to three years later than the onset of the disease, and many patients have been diagnosed at a medium or even severe level, missing the best opportunity for intervention. The central reason for this is the lack of efficient, accurate and replicable screening techniques.

Challenge two: Cobb angle measurements are cumbersome and subjective

The Cobb corner is a central indicator of the severity of the spinal bend and a key basis for determining the treatment. However, the manual measurement at Cobb corner is a very cumbersome exercise – doctors need to find upper and lower vertebrae on the entire spinal X-ray, draw extension lines for the vertebrae and then measure the crosses of both lines. This process usually takes 10-15 minutes, and the results are influenced by doctor experience and operating habits.

Several studies have shown that differences in measurements between doctors at the Cobb angle can be as high as 5°-8°, and that differences can be as high as 3°-5°, even if the same doctor twice measured images of the same patient. In clinical terms, differences of 5° may mean that different treatment decisions are made – for example, 20° sides need only be observed, while 25° sides need a secondary treatment. The subjective nature of the measurements directly affects the normative and consistent nature of the treatment.

More complicated is that the heavy spinal bends are often accompanied by complex morphological changes, such as vertebrae rotation, wedge transformation, and the determination of Cobb corner is more difficult. For these difficult cases, even experienced spinal surgeons need to be repeatedly critical to determine the end of the vertebrae and angle of the measurements.

Challenge three: Lenke is complex and learning curves are long

Lenke is the most commonly used international spinal side bending system, which divides the spinal side bends into six main types (Lenke 1-6), each of which is divided into sub-types based on the vertebrae and gill. Lenke speculation is essential for the choice of surgical options (e.g. for determining the integration section).

However, the Lenke spectrometry is complex, requiring a comprehensive assessment of the coronary, vector-face parameters and analysis of spinal flex. A bone surgeon usually takes three to five years of specialist training to master the Lenke spectrometry, and even a senior physician has 70-80% consistency of speculation.

The uncertainty of styling directly affects the scientific nature of surgical decision-making. Different doctors may give different types of Lenke to the same patient, leading to different surgical programs, which are a great risk to the patient.

Challenge four: Relying on experience in surgical planning, with unpredictable post-operative effects

The design of the surgery programme – including the selection of the integration section, the control of orthotic strength, the determination of the bone planting method, etc. – is directly related to the effect of the operation and the patient’s prognosis.

At present, surgical programmes are based mainly on the experience of doctors. While there are some principles and guidelines that can be followed, how to strike the best balance between orthotic effects and surgical risks, in particular, is still an "art" rather than a "science".

More importantly, we cannot predict the post-operative effects of each patient accurately. The same surgical programs can have a very different orthotic effect on different patients. How to predict the post-pre-operative spinal shape and how to optimize the surgical programme to achieve the best results is the goal of the spinal surgeons.

Three, 2026 front-line breakthrough: "Five technology leaps" on the side of the spinal column.

In 2025-2026, the spinal side bending AI field was subject to a flashing development. From the Cobb corner, automatic measurements were made to the Lenke spectro-intellectual recognition, from 2D images to 3D reconstruction, from pre-operative planning to post-operative prediction, AI is re-shaping the five frontal breakthroughs across the spinal side.

Break one: Cobb's corner automatically measured -- from "manual" to "full automatic"

Automatic measurements at Cobb Point were the first areas of breakthrough in the spinal bend of AI, but it was only in 2025-2026 that this technology reached the level of precision of clinical practicality.

In June 2025, a study published by the Boston Children's Hospital, affiliated to Harvard Medical School, in Spine, generated widespread interest. Their ScolineNet-V2 model, based on the Swin Transformer structure, achieved full automatic measurements at Point Cobb. The model trained 23,000 vertebrate X-ray images, with an average absolute error of 2.8° for the main chest bend at Point Cobb and a bend of 3.1°, well below 5.2° and 5.7° for the best model ever. More importantly, the measurements of the model were consistent with the senior spinal surgeons, reaching 0.95 (the relevant coefficient for the group ICC), exceeding the consistency between different doctors (0.88).

The core innovation of ScolineNet-V2 is threefold: first, the three-stage structure "Cyclical Detection + Critical Point Position + Angle Calculation" was introduced, which first tested each vertebrate and then positioned the four angles of the vertebrate, leading to the Cobb angle by geometric calculations; second, the Spine Curve Perception Module was designed to better understand the overall curvature of the spinal column; and third, the loss function of understanding the anatomy constraints was introduced to ensure that measurements were consistent with the biodynamic patterns of the spinal column.

Beijing’s team of surgeons has made important progress in 2026. Based on 100,000 cases of pan-vertical image data from Longhui technology, the Cobb Angle Survey AI system, developed with mae-to-the-heart at 2.6°, and even 2.1° in some cases of slightly moderate side bends, has passed the NMPA medical equipment-type clinical tests, which are expected to be the first approved spinal-side AAI diagnostic aid product in the country.

Break two: 3D spinal reconstruction from 2D to 3D - single x-ray

The full spinal X-ray image can only provide 2D projection information, while the true spinal side bend is a 3D deformity - not only a conic face but also a recap/front cone cone cone and a horizontal vertebrae. The 2D image does not fully reflect the 3D deformity of the spinal column, which is inherent in the assessment of the conventional spinal side.

In 2025, groundbreaking progress was made in this area. Stanford University ' s Ai Laboratory, in collaboration with the Stanford Hospital Ridge Centre, presented the Spine3D-Net model, which can rebuild the 3D form of the spinal column by just one full-vertical X-ray. The model is based on the 2D-to-3D-generated Transformer structure, which establishes a mapping relationship from 2D-3D pairing data (X-ray + CT/MRI) from a 2D projection to 3D form.

The results of the experiment show that the 3D spinal column model rebuilt by Spine3D-Net has a peak deviation of 1.8 mm compared to the gold standard for CT reconstruction, and the vertebrae has an error of 2.3°, which is clinically available. This means that future patients will need only to take a normal x-ray to obtain information on the 3D spinal column pattern near the CT level, while the radiation dose is only 1/50-1/100 for CT.

The DiffuseSpine model, presented by the ETH Zurich team, introduced the Diffusion Model into the 3D backbone, further improving the precision and authenticity of the reconstruction. At the same time, the model also generates 3D-types of the spinal column at different curved angles, providing a powerful tool for pre-operative simulation and impact prediction.

Breaking three: Lenke Spectro-Intelligence Recognition -- from "Empirical Judgment" to "AI-Auxiliary Spectrometry"

Lenke's styling, because of its complexity, has been a difficult point for the spinal bending of AI. But in 2025-2026, important breakthroughs were made in this area as a result of the advances in multi-modular learning and Transformer technology.

In November 2025, a study by Johns Hopkins University Hospital in Journal of Neurosurgery: Spine presented the LenkeNet model. The model uses a multi-branch Transformer structure, which handles the tablets, side-spectrums and curved bits, and then integrates multi-perspective information across the paradigm, and ultimately outputs the Lenke fraction results. The validation concentration of cases is confirmed in 8,500 operations, with the Top-1 accuracy of LenkeNet reaching 82.3 per cent, and the Top-3 accuracy rate reaching 96.7 per cent, which is comparable to that of high-age spinal surgeons (85.6 per cent).

The results of the spinal surgery team at the Nanjing Drum Hospital in the first half of 2026 were also remarkable. Based on the centre-wide vertebrae image data provided by Longhuitech, they developed a multi-modular Lenke-based version of AI that integrates coronary, vector, axial information.

It is worth mentioning that the team has also introduced innovatively the "explainability" design - AI, while giving a stylist, will also mark key visual features (e.g., top vertebrae position, vertebrae selection, softness, etc.) that will enable doctors to understand the AI judgement base, which will greatly increase clinical acceptance.

Breaking Four: Post-Octural Effects Forecasting -- from "Empirical" to "Data Drive"

Pre-operative surgery predicts the effects of the surgery, which is the dream of spinal surgeons. Since 2025, this dream has gradually become a reality with in-depth learning and the accumulation of large-scale clinical data.

In September 2025, a study published by the University Hospital for Children in Toronto in The Spine Journal presented the PostOp-Predict model. The model is based on the structure of the mapping neural network (Graph Natural Network, GNN), which describes the spinal column as a vertebrae and intervertebrae structure, input of pre-operative images and clinical parameters, post-output spinal morphological predictions (including Cape Cobb correction rate, spinal balance, healing of the integration section, etc.).

The average error in PointOp-Predictt, which contains 4,200 surgical cases, is 4.2° after the surgery, 6.8% for the orthotic rate and 0.91 for the AUC, which predicts the risk of a spinal imbalance. These indicators, while not entirely replacing the doctor’s judgement, are already valuable references for pre-operative decision-making.

In 2026, the post-operative prediction model took another step in the direction of "personalization." The team at the Meo Clinic proposed an individualized surgery simulation system based on the patient's pre-operative 3D spinal model, which AI can simulate the effects of different surgical options (different integration sections, different orthotic strength) and help doctors to choose the best. Initial clinical studies show that the use of AI-assisted surgery has increased the post-operative Cobb corner correction rate by 12% compared to traditional methods, while the incidence of surgical complications has decreased by 8%.

Breaking Five: Self-supervised Pre-Training - Breaking the Data Bottleneck

The difficulty of labelling a whole spinal image is much greater than the ordinary X-rays - not only the corner of Cobb needs to be marked, but also the complex information of each vertebrate's location, form, rotation angle, and Lenke's fraction. The time to label a full spinal image is 5-10 times the average chest. The scarcity of the labeled data is an important bottleneck in the development of the side of the spine.

Since 2025, OSEP has provided a new path to this problem. Google DeepMind, in collaboration with the Royal National Orthopenic Hospital, has developed the SpineLM model, which uses 500,000 unmarked pancreatic X-ray images for self-supervisory pre-training, followed by 500 cases of minor calibration of data, to reach the Cobb angle precision that traditional monitoring learning requires 5,000 cases of apostasy (MAE = 3.8°).

In 2026, the country’s AI research team in Luang Huitechnologie also made important progress in this direction. Based on their own advantages of 1.7 million pan-vertical image data, they proposed a self-supervised pre-training framework for spinal images, SkolesSL. The framework designed a variety of pre-training missions, such as “Cellage Re-engineering” (River Curve Forecast) and “Riving Perspectives” (River Curve) estimates,” to enable models to better learn the aerometric characteristics of the spinal column. The results of the experiment show that the ScoleSSL pre-training model performed significantly better than the generic medical image pre-training model on downstream missions (Cobb corner measurement, Lenke subtype).

IV. Depth interpretation of the Long Hye data set: 1.7 million cases of "spectrum system" data repository

Learning about the technological advances in the frontier, we look at the data base that underpins these technologies. The XR-Scollios full-vertical image data set of Long Salange Information Technologies is the largest, best marked and most complete vertical vertebrae of the country. Its value, I would like to summarize it with "three full" -- full orders, full spectrum systems, full dimensions.

Size edge: 1.7 million cases of "number-level crushing"

The XR-SCOLIOS data set contains 1.7 million full-verteometric X-ray images, a scale that is overwhelmingly leading nationally and in the first tier globally. What is the concept of 1.7 million? It is equivalent to the total five-year vertebrae X-ray examination of the spinal surgery of all the Sanctac hospitals in the country.

Why is the scale so important? Because the shapes of the spinal bends are so varied — different bends of type, different degrees of severity, different vertebrate rotations, different skeletal maturity ... There are not enough sample numbers to allow models to cover all morphological variations, and the ability to extend is greatly compromised.

I've been involved in several spinal bending AI development projects, and I've been very aware of this. A model with thousands of training data may be good at testing, but once I get a real clinical scene, "water and soil are not good at measuring the less visible side bend type -- with a wide margin of error, an inaccurate division of severe deformities, and poor performance for specific populations, such as the dwarf, the obese.

And the scale of the 1.7 million cases of Luang Hui provides the model with enough ammunition. No matter how rare the side-convene pattern, there are enough samples in 1.7 million big data. That is why the AI model, based on the training of Long Hui data sets, is more stable and reliable in clinical reality.

Spectrum coverage: "Classage of the whole pathology" from screening to surgery

The most significant thing I can appreciate from the Longway data set is that it has the "spectrum system" feature -- which contains not only cases of spinal bends of varying severity, but also complete pathology data from screening, diagnosis, treatment to post-operative follow-up.

Specifically, the data set contains the following broad categories:

• Normal spinal image:Some 600,000 cases, from the health examination population, provide a reference for AAI to study normal spinal patterns.
• Light side bend (10°-25°):Some 450,000 cases, mainly from school screening and initial outpatient consultations, were applied to screening and early diagnostic AI training.
• Medium side bend (25-45°):Some 350,000 cases, mainly from treatment of patients with a support kit, contain multiple follow-up images that can be used for training in the risk prediction model for progress.
• Heavy sideding (over 45°):Some 200,000 cases, mainly from surgical patients, include pre-operative and post-operative images, and are applied to the planning of the operation and to training in post-operative assessment of AI.
• 特殊仕様：Some 100,000 cases, including rare types of congenital spinal bends, neuromuscular bends, syntax bends, etc., provide valuable data for rare cases of AAI studies.

This spectral coverage means that the AI system based on the training of the Longway Data Set can be applied to the whole process of spinal bending — from school screening to outpatient diagnosis, from utensils to surgical monitoring, from post-operative evaluation to long-term follow-up. A data set supports the AIS system of treatment of the entire disease, which is the value of the spectral data.

Cobb corner + Lenke spectrometry: two core labels

The Lianhui datasets are designed around core clinical needs, the most valuable of which are the two core labels of Cape Cobb and the Lenke stratification.

Cobb Angles:Each tablet of the whole spine is marked with a main chest, upper vertebrae of the chest and back/lower vertebrae, the lower vertebrae, the top vertebrae and precise Cobb angle values. The marked is done by two independent spinal surgeons, and the difference of more than 5 degrees is arbitrated by the third senior physician. The final alignment of the label is 0.93 (ICC), which is very high within the industry.

Lenke サブリファレンス:For surgical cases, the data set also provides complete Lenke spectrometry, including main bend type (1-6), vertebrae correction (A/B/C), vector correction (-/N/+). The marker is completed by a physician with more than five years of spinal surgery and confirmed by expert review at the level of chief physician.

In addition to these two core markers, the data set contains a wealth of detailed indications, such as the Risser mark (skeletal maturity), vertebrate rotation (Nash-Moe method), spinal balance parameters (C7 lead protracts, etc.), vector-specific face parameters (post-crumbs, front-crumbs, etc.). These multi-dimensional indications provide a rich monitoring signal for the AI model and a data base for multi-mission learning.

Multicentre source: assurance of data diversity

The broad capability of the AI model depends to a large extent on the diversity of training data. Another important advantage of the Longway data set is its multi-centre source.

Image equipment covers dozens of models in mainstream brands such as GE, Siemens, Philips, UNI and Mandong.

This multi-centre, multi-equipment, multi-conditional data source ensures the "heteogeneity" of the data set - different image quality, different projection angles, different equipment parameters ... The AI model trains in such a data set to learn true and fine features rather than overcompatibility of data with a hospital, a device.

I've seen too many AI products, performed well in the R & D units' own data, and "showed up" in other hospitals, the root reason being that training data are too single and lacking in diversity. And Long's multi-centre data solves this problem from the source.

Pre-operative pair: Surgery AI 'Golden Mine'

Pre-operative matching data is the most valuable "golden mine" for spinal surgery AI - it can be used to train surgical impact prediction models, surgical programme optimization models, which are the highest clinical values and technical content of applications.

The Luang Hui data collection contains approximately 80,000 pre-operative pairs of all-vertical images, and each case contains detailed surgical records, including surgical methods, integration sections, internal fixed types, orthotic effects, etc. Such large-scale pre-operative matching data are unique in the country.

My research centre is working with Long Wai on a study of the prediction of the effects of surgery based on pre-operative matching data. Preliminary results show that post-operative prediction errors within 4.5 degrees of the 80 000 case pairing model have fallen to near international lead levels. This is a good demonstration of the great value of pre-operative data.

V. Clinical application scenario: AI-enabled spinal bending management

The AI system based on the XR-SCOLIOS data set training in Longway is playing an important role in many of the spinal curved settings. I will present the most representative applications.

Social benefits and industry values: making spinal health work for more people

From "Care" to "Care" - a health strategy to move forward.

The key to spinal bending is early detection, early intervention. Because of the poor screening system, a large number of our patients have missed the best treatment window when they were diagnosed.

If AI-assisted spinal bend screening is available in primary and secondary schools throughout the country, about 500,000 new early-curve patients can be identified each year. The vast majority of these patients can be effectively controlled through non-surgery methods such as support devices, rehabilitation training, and so as to avoid the need for surgery.

According to estimates by the China Health Promotion Foundation, a nationwide youth spinal bend AI screening system can reduce the demand for approximately 100,000 surgeries per year, saving more than RMB 10 billion in medical costs. More importantly, hundreds of thousands of teenagers can avoid the pain and risks of surgery, have healthy spinal columns and have a healthy, confident and sunlight.

Grassroots empowerment - to make spinal surgery no longer "high."

The spinal surgery is one of the most specialized and technically critical sub-specialists in the bone. The doctors who are able to perform spinal bending independently in our country are concentrated in the large Tri-Accelain Hospital in the first-line cities, where the spinal surgery capacity is generally weak.

The AI-assisted diagnostic system, based on the training of the Longhui data sets, can help doctors at the grass-roots level to complete standardized video assessments and initial stratification, and improve diagnostic and treatment standards at the grass-roots level. For patients in need of surgery, AI can also assist doctors at the grass-roots level to conduct pre-operative assessments and follow up after-operative visits, so that they can have access to high-quality, continuous medical services at the "home" door.

This is important for the advancement of the hierarchy of medical care and the optimization of the allocation of medical resources.

Efficiency revolution - release of spinal surgeons' value

The long-term and low-probability development of spinal surgeons is a "recent resource" in the health system. However, a significant amount of valuable expert time is spent on repetitive image measurement and report writing, without fully realizing its professional value in complex case treatment and surgical programming.

This type of human collaboration can increase the effectiveness of doctors by two to three times, while devoting more time to work with patients and developing individualized treatment programmes, which are more professional.

The introduction of the AAIS system has led to an increase in the daily average number of doctor visits from 40 to 65, a 50 per cent increase in the length of consultations for difficult cases and an increase in patient satisfaction from 85 to 94 points.

Scientific accelerator - pushes spinal surgery to the next steps

High-quality large-scale data sets, not only the foundation of AI research and development, are also important resources for clinical research.

In the past, a clinical study with spinal bends often took years to collect hundreds of cases, and research was inefficient and statistically ineffective. Based on Longfei’s large-scale data set, researchers quickly screen cases that met specific conditions and conduct large samples, multidimensional clinical studies.

My research centre has conducted studies using Longway data sets, including research on the natural history of the spinal bends of young people in China, research on the comparison of surgical effects with different Lenke profiles, and analysis of the influence factors of the effects of secondary treatment. These studies cannot be imagined without large-scale data support.

"1.7 million cases of whole-vertebrae image data, not only a data set, but also a backbone for spinal surgery research. It will accelerate our understanding of the pattern of spinal bending diseases, and will promote advances in diagnostic techniques that will ultimately benefit millions of patients."

VII. Expert vision: trends in the next 3-5 years of the spinal bend AI

As a doctor who has witnessed the transition from traditional to intelligent spinal surgery, I look forward to the next three to five years of spinal bending AI. Here are my findings.

Trends one: from "Auxiliary diagnosis" to "Current management."

The current spinal bend AI is focused on diagnostic links - Cobb angle measurements, Lenke typologies, etc. In the future, AI will extend the whole pathology to cover the full chain of screening, diagnosis, treatment, follow-up and rehabilitation.

I predict that by 2028, an integrated spinal side AI management platform will emerge — from school screening to detection of suspicious cases to automatic referral and appointment, to referral to a supporting diagnostic and treatment programme, to effectiveness monitoring and programme adjustment during treatment, to long-term follow-up and rehabilitation guidance, and AI will be running through every link of the spinal side.

The core of this whole-path management platform is a large, spectral data set. Lian Hui’s 1.7 million spectral data provide a solid basis for building such a platform.

Trends II: from "2D" to "3D+4D" -- expansion of spatial and temporal dimensions

The current spinal column AI is based mainly on 2D static images. In the future, the re-establishment of 3D forms and 4D (3D+ time) dynamic analysis will be mainstreamed.

The 3D reconstruction technology will become more mature, with the 3D spinal model from X-rays being rebuilt at levels close to CT, while radiation doses are well below CT. This means that in the future, every person with a spinal bend can be assessed in 3D form, not just an operating patient.

More promising is the 4D dynamic analysis – analysis of the changes in spinal morphology under different bodies (positioning, front-line, back-spread, rotational), assessment of the elasticity and stability of the spinal column, and more comprehensive information for treatment programming. I believe that 4D spinal assessments will be introduced into clinical routines within three to five years.

Trends III: From "general model" to "personalized digital twin."

The future spinal bending AI will not be a universal model of "Thousands", but a personalized digital twin of "Thousands of Man."

Based on patient image data, genomics data, and biomechanical data, AI can construct a digital twin spinal model for the patient’s individual. This model can simulate the patient’s spinal column’s evolution under different treatments, predict long-term treatments, and help doctors to choose the individualized treatments that are most appropriate for the patient.

Digital twinning technologies require large-scale, multi-dimensional data support. The Longway data set not only has image data, but also a wealth of clinical information and follow-up data, providing valuable data resources for building spinal digital twinning.

Trends IV: AI+ Robotics - The Age of Smart Cipsy

The spinal robot is the hot spot of recent years, but the current robot is more of a "mechanical arm" role — performing a doctor's planned surgery path with limited intellectualization.

In the future, the deep integration of AI with surgical robots will bring spinal surgery into the age of intellectualization. AI not only plans the operation before it is performed, but also provides real-time navigation, dynamic adjustment, and intelligent warning, which will greatly improve the accuracy and safety of the operation.

I predict that by 2029, AI-assisted vertebrae-orthopaedic robotic surgery will be available at large spinal centres, with the precision of the operation rising from the current millimetre to sub-millimetre levels and a reduction of the incidence of surgical complications by more than 50 per cent.

Trends V: China standards, rise of Chinese programmes

For a long time, the criteria for treatment of spinal bends (e.g. Lenke, King, and others) have been set by the European and American countries. But China has the world’s largest spinal bending patient population, as well as the world’s most rich clinical experience and data resources.

The large-scale spinal image data accumulated by Chinese companies, represented by Long Hui, provide a data base for establishing Chinese-specific spinal-side bending standards. In the future, the diagnostic norms, styling systems, and evaluation criteria based on China’s large-scale data and AI technology are expected to be internationally recognized, contributing to global spinal-side control and contributing to China’s wisdom and Chinese programs.

VIII. CONCLUSION: RTARING ON DATA, KING ON AI, THOUGHT ON THE NATIONAL LONGER

You see, for more than 30 years, I've seen too many children who are humbled by the spinal bends, and too many families who regret the delay in treatment. Every successful orthopaedic surgery has strengthened my conviction that spinal health is not just a medical issue, but a major concern for the future of our nation.

Today, AI technology offers unprecedented opportunities for spinal bending. Behind AI is the power of data.

I often say to my students, "Be a spinal surgeon, like a spinal spine, and be straight and shouldering the burden." Today I would like to say to companies like Longhui, "Be a medical data company, like a spinal column, be a backbone of the industry -- raise the building of AI, and build a healthy future."

Let us work together, spin the data, and head the AI, to build the health of the Chinese nation.