A virtual dissection intelligent image processing platform

Abstract

The application relates to the technical field of medical image processing, and particularly discloses a virtual dissection intelligent image processing platform, which comprises a hardware unit and a software unit. The hardware unit is an industrial control computing host adopting an independent graphics card and is packaged as a virtual dissection image processing black box. The software unit is integrated in the hardware unit, and through AI technology, automatic recognition and segmentation of organs and injuries, high-fidelity false-color three-dimensional rendering and dynamic scene generation are realized, and finally, data packets suitable for devices are output after VR optimization; the application can support a forensic immersive dissection-free condition, and can also support simultaneous consultation of many people; the application can realize non-invasive intelligent virtual dissection analysis, and can provide a very reliable data reference for corpse cause-of-death judgment.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, specifically to a virtual anatomical intelligent image processing platform. Background Technology

[0002] In forensic practice, autopsy is a crucial method for determining the cause of death. Traditional autopsies require physical dissection, a process that is invasive, irreversible, and time-consuming. With the development of imaging technology, CT scans have become an auxiliary tool for autopsy examination. However, existing medical image processing software is mostly designed for live clinical diagnosis and lacks dedicated integrated systems tailored to the characteristics of cadaver tissue and the needs of forensic analysis. Furthermore, existing systems often cannot efficiently process multimodal CT data (such as spectral data and contrast data), and their 3D reconstruction and dynamic display functions have low levels of intelligence, making it difficult to directly support interactive analysis in a virtual reality (VR) environment.

[0003] Therefore, there is an urgent need for an integrated and intelligent dedicated image processing platform that can interface with cadaver CT scan data and automatically generate multimodal and multidimensional image results to assist forensic doctors in making efficient and accurate determinations of the cause of death. Summary of the Invention

[0004] The purpose of this invention is to provide a virtual autopsy intelligent image processing platform that can interface with cadaver CT scan data, automatically generate multimodal and multidimensional image results, and assist forensic doctors in making efficient and accurate determinations of the cause of death.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] This invention integrates multiple image processing functions into a single, unified black box, offering simple operation and highly targeted capabilities. It can fuse multimodal CT data to provide comprehensive cadaver information images; furthermore, it utilizes AI technology to automatically identify and segment organs and injuries, and intelligently generate dynamic images, reducing manual operation while improving analysis efficiency and objectivity; additionally, it supports VR-style output, aiding in the determination of complex causes of death; furthermore, this invention preserves the original state of the cadaver based on virtual dissection, allowing all images to be repeatedly accessed and analyzed from multiple angles.

[0007] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the main structure of the present invention;

[0009] Figure 2This is a schematic diagram of the base plate structure of the present invention;

[0010] Figure 3 This is a schematic diagram of the front baffle structure of the present invention;

[0011] Figure 4 This is a schematic diagram of the rear baffle structure of the present invention;

[0012] Figure 5 This is a schematic diagram of the graphics card bracket structure of the present invention. Detailed Implementation

[0013] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0015] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0016] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0017] like Figures 1-5 As shown, a virtual anatomical intelligent image processing platform includes hardware units and software units;

[0018] The hardware unit is an industrial control computing host with a dedicated graphics card. The dedicated graphics card is a mainstream high-performance gaming graphics card or a professional graphics card such as the NVIDIA Quadro series. It is packaged as a black box for virtual anatomical image processing and has high-performance graphics processing and parallel computing capabilities. It can be used to process large cadaver CT datasets at high speed and is connected to the cadaver CT scanner via gigabit Ethernet or fiber optic. The cadaver CT scanner is equipped with an energy dispersive spectrometer and angiography function and can receive raw tomographic scan data containing energy dispersive spectrometer and angiography data.

[0019] The software unit is integrated into the hardware unit and includes modules for generating whole-body black-and-white grayscale images from raw cadaver CT data, modules for generating whole-body angiography images from cadaver CT data after contrast agent injection, modules for generating whole-body spectral analysis images from CT spectral scan data, modules for generating color 3D images of the cadaver from the black-and-white grayscale images based on AI learning, modules for generating dynamic 3D images using the color 3D images and other images based on AI learning, and modules for processing the generated image data into a format suitable for VR devices.

[0020] The dynamic three-dimensional images include the display of separation, transparency, rotation, or pathological process simulation of organs and blood vessels, including blood flow dynamics and damage formation processes.

[0021] Specifically, the AI ​​learning function is implemented based on a deep learning model, which is trained using a large dataset of labeled cadaver CT images. The deep learning model used by the module that generates color 3D images of cadavers from the grayscale images based on the AI ​​learning function is capable of specifically identifying common injuries, lesions, and postmortem changes in cadavers. The module that generates dynamic 3D images using the color 3D images and other images based on the AI ​​learning function includes a physical simulation engine and a time series prediction algorithm, which can generate dynamic images from specific perspectives or processes according to user instructions, such as displaying coronary artery blood flow or preset processes.

[0022] Specifically, all modules of the software unit are invoked and managed through a unified user interface, and it supports direct communication with standard DICOM protocol CT devices, enabling seamless data transmission and interaction.

[0023] Specifically, the software unit also includes a data access and quality control layer, which is used to receive raw processing result packets pushed by the hardware unit in real time through a dedicated high-speed network interface. The dedicated high-speed network interface supports 10GbE or fiber optic. The raw processing results include intermediate data such as the completed basic grayscale image, angiography image, energy spectrum image, and AI preliminary segmentation label. Data integrity verification adopts MD5 or SHA256 method, DICOM-SR structured report parsing, and automatic anomaly detection such as image missing, excessive artifacts, and abnormal dosage. When an anomaly occurs, an alarm is immediately triggered and the hardware unit is sent back to request retransmission or rescanning.

[0024] Specifically, the software unit also includes a multimodal image unified registration and fusion engine, which is used to uniformly register the basic grayscale three-plane images, namely transverse sagittal and coronal images, angiography two-dimensional MIP or VRT images and three-dimensional vascular tree images, energy spectrum imaging images, namely material separation images including iodine images, water images, uric acid images, calcium images, etc., and AI preliminary organ segmentation masks to the same cadaver coordinate system with an accuracy of ≤0.5mm. Finally, it outputs multimodal fusion volume data as the basis for all subsequent advanced processing.

[0025] Specifically, the software unit also includes a localized caching and collaboration platform for forensic workstations, which supports automatic encrypted storage of data on a local medical-grade NAS. The local medical-grade NAS supports two-way audit logs, multi-person collaborative VR consultations, and allows up to 16 forensic doctors from different locations to enter the same virtual autopsy table simultaneously on the same corpse. It also supports voice and spatial annotation, and the annotations can be saved as permanent evidence. It also supports real-time synchronization of pointers and measuring tools, collective voting on injury sites, and automatic generation of VR identification reports after virtual autopsy. The reports include all measurement data, ballistic diagrams, dynamic video links, and electronic signatures of participating forensic doctors.

[0026] Specifically, the software unit also includes an AI deep segmentation and intelligent damage recognition module, used to run a three-level cascaded deep learning network on multimodal fusion data. The deep learning network is based on nnU-Net++ with a Dice coefficient greater than 0.98. The first level achieves ultra-high precision segmentation of organs, bones, and soft tissues. The second level achieves fine-tuning of microstructures, including tracheal and bronchial trees, coronary arteries, cerebral gyri, and nerve bundles. The third level is a forensic pathological damage recognition model, which is the latest model trained in 2025. It includes automatic three-dimensional tracking and ballistic reconstruction of gunshot wounds, accurate measurement and blade feature inference of stab wounds, recognition of mechanical asphyxiation signs, quantification of pulmonary edema distribution in drowning, grading of burn depth and carbonization degree, automatic detection and trajectory reconstruction of multiple fragments in blast injuries, etc. The final output is an ultra-fine semantic segmentation body with forensic-specific labels.

[0027] Specifically, the software unit also includes a high-fidelity pseudo-color 3D rendering engine, which uses the segmentation results from the AI ​​deep precision segmentation and damage intelligent recognition module to drive the physically based rendering material system in real time, assigning medical-grade standard pseudo-color to each organ and tissue. Customizable schemes include classic forensic color schemes, transparency mode, children's mode, etc. It includes realistic lighting models including subsurface scattering, refraction, and metallic feel for shrapnel or prosthetics, generates 8K texture maps and normal maps in real time, and finally outputs a medical-grade film-quality color 3D corpse model with a single-frame rendering time of less than 60 milliseconds based on RTX5090 level hardware.

[0028] Specifically, the software unit also includes an AI dynamic 3D scene generator, which receives static color 3D models and dynamic demand commands input by forensic experts, and generates heartbeat and realistic hemodynamic simulations (i.e., 4DFlow) and dynamic simulations of the entire process of bullet wound formation. The dynamic simulations of the entire process of bullet wound formation include tissue tearing, cavity effect, temporary cavity expansion and collapse, dynamic evolution of organ damage from stab wounds including bleeding rate and volume estimation, dynamic filling process of pulmonary edema in drowning, shock wave propagation and simultaneous damage sequence of multiple organs in blast injuries, progressive transparency animation of any organ or structure, explosive separation display of a single organ or injury, and dynamic sequences of 360-degree automatic flight roaming paths, i.e., classic forensic observation routes. All dynamic sequences can be exported as 60fps 8K video or interactive sequence frames.

[0029] Specifically, the module used to process the generated image data into a format suitable for VR devices can perform VR extreme optimization on all assets generated by the high-fidelity pseudo-color 3D rendering engine and the AI ​​dynamic 3D scene generator. The optimization includes automatic LOD generation (Level 6 and above), GPU Instance massive instance optimization, Foveated Rendering and Eye-tracking adaptation, and skeleton binding. It outputs three data packages compatible with mainstream virtual reality head-mounted display devices.

[0030] The first option is an exclusive package for MetaQuest3 or Quest Pro, in APK format, less than 4GB in size, and supports wireless PCVR.

[0031] The second option is the VarjoAero or XR-4 professional-grade package, with a resolution of 8K×2, supporting eye tracking and fingertip tracking;

[0032] The third type is the PCVR premium package, which supports devices such as Valve Index, Pimax Crystal, Bigscreen Beyond, and Web XR versions. The Web XR version requires zero installation, and tablets or mobile phones can also enter VR mode, supporting forensic experts to conduct immersive observation, interactive operations, and measurements in a VR environment.

[0033] When using

[0034] First, a full-body scan is performed by a cadaver CT scanner, generating raw tomographic data containing energy spectrum and contrast data. This data is then transmitted to the hardware unit via gigabit Ethernet or fiber optic cable. After initial processing, the hardware unit pushes the raw processing result package to the data access and quality control layer of the software unit. Once verified and found to be safe, the entire intelligent processing process is initiated.

[0035] Then, the software unit sequentially completes the multimodal image unified registration and fusion engine, the AI ​​deep fine segmentation and intelligent damage recognition module, the high-fidelity pseudo-color 3D rendering engine, the AI ​​dynamic 3D scene generator, and the VR immersive interaction optimization and output module, ultimately generating multimodal images, 3D models, and VR data packages that can be used for analysis. Within 15 minutes, forensic doctors can put on VR headsets and enter the fully restored corpse to conduct immersive virtual dissections, ultimately achieving knife-free, zero-pollution, infinite backtracking, and simultaneous operation by multiple people in different locations.

[0036] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0037] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A virtual anatomical intelligent image processing platform, characterized in that, Includes hardware units and software units; The hardware unit is an industrial control computing host with a dedicated graphics card, packaged as a virtual anatomical image processing black box. It has high-performance graphics processing and parallel computing capabilities, used for high-speed processing of large cadaver CT datasets, and is connected to the cadaver CT scanner via gigabit Ethernet or fiber optic to receive raw tomographic scan data containing energy spectrum and contrast data. The software unit is integrated into the hardware unit and includes a module for generating whole-body black-and-white grayscale images from raw cadaver CT data, a module for generating whole-body angiography images from cadaver CT data after contrast agent injection, a module for generating whole-body spectral analysis images from CT spectral scan data, a module for generating color 3D images of the cadaver from the black-and-white grayscale images based on AI learning, a module for generating dynamic 3D images using the color 3D images and other images based on AI learning, and a module for processing the generated image data into a format suitable for VR devices. The dynamic three-dimensional images include the display of separation, transparency, rotation, or pathological process simulation of organs and blood vessels, including blood flow dynamics and damage formation processes.

2. The virtual anatomical intelligent image processing platform as described in claim 1, characterized in that, The AI ​​learning function is implemented based on a deep learning model, which is trained using a cadaver CT image dataset. The deep learning model used by the module that generates color 3D images of cadavers from the black and white grayscale images based on the AI ​​learning function can specifically identify common injuries, lesions, and post-mortem changes in cadavers. The module that generates dynamic 3D images using the color 3D images and other images based on the AI ​​learning function includes a physical simulation engine and a time series prediction algorithm, which can generate dynamic images from a specific perspective or process according to user instructions or preset procedures.

3. The virtual anatomical intelligent image processing platform as described in claim 1, characterized in that, All modules of the software unit are invoked and managed through a unified user interface, and it supports direct communication with standard DICOM protocol CT devices.

4. The virtual anatomical intelligent image processing platform as described in claim 1, characterized in that, The software unit also includes a data access and quality control layer, which is used to receive raw processing result packets pushed by the hardware unit in real time through a dedicated high-speed network interface, perform data integrity verification, DICOM-SR structured report parsing and automatic anomaly detection, and alarm when an anomaly occurs and send a request back to the hardware unit to retransmit or rescan.

5. The virtual anatomical intelligent image processing platform as described in claim 1, characterized in that, The software unit also includes a multimodal image unified registration and fusion engine, which is used to uniformly register the basic grayscale three-plane image, angiography image, energy spectrum imaging image and AI preliminary organ segmentation mask to the same cadaver coordinate system, and output multimodal fused data as the basis for all subsequent advanced processing.

6. The virtual anatomical intelligent image processing platform as described in claim 1, characterized in that, The software unit also includes a localized caching and collaboration platform for forensic workstations, supporting encrypted data storage, multi-person collaborative VR consultations, voice and spatial annotation, real-time synchronized pointers and measuring tools, collective voting on injury sites, and automatic generation of VR identification reports after virtual autopsy.

7. The virtual anatomical intelligent image processing platform as described in claim 1, characterized in that, The software unit also includes an AI deep fine segmentation and intelligent damage recognition module, which is used to run a three-level cascaded deep learning network on multimodal fusion data to achieve ultra-high precision organ, bone and soft tissue segmentation, fine-tuning of microstructures and identification of pathological damage for forensic purposes.

8. The virtual anatomical intelligent image processing platform as described in claim 1, characterized in that, The software unit also includes a high-fidelity pseudo-color 3D rendering engine, which, based on the segmentation results of the AI ​​deep precision segmentation and damage intelligent recognition module, drives the PBR material system in real time to assign medical-grade standard pseudo-color to each organ and tissue, including a realistic lighting model, and generates 8K texture maps and normal maps in real time, outputting a medical-grade film-quality color 3D cadaver model.

9. The virtual anatomical intelligent image processing platform as described in claim 8, characterized in that, The software unit also includes an AI dynamic 3D scene generator, which receives static color 3D models and dynamic demand commands input by forensic experts. It generates dynamic sequences of heartbeat and hemodynamics simulation, the entire process of bullet wound formation, dynamic evolution of organ damage from stab wounds, dynamic filling process of pulmonary edema in drowning, shock wave propagation and simultaneous damage sequence of multiple organs in blast wounds, progressive transparency animation of any organ or structure, explosive separation display of single organ or injury, and 360° automatic flight routing path. All dynamic sequences can be exported as 60fps 8K video or interactive sequence frames.

10. The virtual anatomical intelligent image processing platform as described in claim 9, characterized in that, The module that processes the generated image data into a format suitable for VR devices can perform VR-extreme optimization on all assets generated by the high-fidelity pseudo-color 3D rendering engine and AI dynamic 3D scene generator. The optimization includes: automatic LOD generation, GPU Instance massive instance optimization, Foveated Rendering and Eye-tracking adaptation, and skeleton rigging. It outputs three data packages compatible with mainstream virtual reality head-mounted display devices and a Web XR version, supporting forensic experts to conduct immersive observation, interactive operations, and measurements in VR environments.

Images