Split from TLS-Net time series to Smart Scope POC screening: 2026 time series integration with million-scale data base for vaginal mirror AI

Cervical cancer is the fourth most common cancer among women worldwide, while China is a country with a high incidence of cervical cancer. Vaginal lenses are the key link in the diversion and diagnosis of cervical cancer screening anomalies – observing the varnish reaction to the top of the cervical skin through the acetic acid experiment, and determining whether there is a pre-cancer disease (CIN) and its severity.

In 2026, the vaginal mirror AI achieved a breakthrough in time-series multi-frame integration and POC screening diversion. TLS-Net used the Swin Transformer encoder and time-series attention module to integrate 60/90/1500.180 seconds of white acetate images, and to achieve a precise partition of the cervical-crystals stove area.

The special value of vaginal cortex AI is "catching of time series." The key to the acetic acid experiment is not a single-time image, but a change in colour at any time of the upper skin – a brief and mild reaction to normal upper skin acetic acid, a quick and lasting reaction to high-level disease.

The development of vaginal mirrors in 2025-2026 represented a technological leap from "single-frame classification" to "time-series multi-frame integration".

At the technical level, TLS-Net (2026 Scientific Reports) proposes a time-series attention module that integrates 60/90/1500.180 seconds of acetic acid images into the partition of the region. This approach breaks through the limitations of traditional monolithic analysis and uses dynamical changes in the acetic acid test to improve diagnostic accuracy.

The study of multiple equipment in 2025, using 320 vaginal/alcoscopy tests (3 types of equipment), and the `LSIL' and `HSIL' plaster-like internal pathologies. The study was intended to validate AI's ability to extend between different devices - the optical properties and colour reduction of the vaginal lenses of different manufacturers, and cross-equipment of equipment are key challenges for clinical deployment.

At the level of clinical validation, Smart Spope AI (2024 Cureus) is a POC screening diversion tool, which is 93.46% special, AUC 0.73, sensitive 52.17%. While there is room for increased sensitivity, its "immediate diversion" value lies in the fact that in areas with limited resources, the POC tool can complete screening and diversion in single visits and avoid patients being missed.

At the level of comprehensive diagnosis, the U-NET++ and RepVGG models (Frontiers in Online, 2025) have projected an accuracy rate of 83.01 per cent in 424 tests, with better recognition of CIN I and CIN II early pathologies.

On a market scale, the global market for cervical cancer screening is expected to exceed $5 billion in 2026, and AA vaginal prostheses are one of the fastest growing areas. China is moving forward with the spread of HPV vaccines and cervical cancer screening coverage, with a growing demand for a vaginal lens.

The core breakthrough of the vaginal mirror AI in 2025-2026 was concentrated on three areas.

The first is time series multi-spectrum integration. TLS-Net uses the Swin Transformer encoder to extract the characteristics of each frame, integrating 60/90/1500.180 seconds of white acetate images through the time series attention module. The core innovation of this approach is "temporal agulation" — not to analyse each frame independently, but to model the temporal dynamics of the reaction. This time series model enables AI to distinguish between "quick and lasting acetic acid" (high-level pathologies) and "slow and short-lived white acoustic acid" (low-level or normal features) and achieve a more precise CIN rating.

The second is trans-equipment. 320 multi-equipment studies have validated the ability of CNN to trans-compatibility between three different devices. Trans-equipment is particularly important for vaginal mirrors, which vary in light-source temperature, optical resolution and color reduction from different manufacturers, and which severely limits clinical outreach if models can only work on specific equipment.

Third is the POC screening diversion. The POC value of Smart Scope AI is in "immediate diversion" – completing further diversion of HPV-positive patients during a single visit, judging whether vaginal mirrors are needed or can be followed directly. While AUC 0.73 and sensitivity 52.17% still need to be raised, there is no substitute value for "immediate decision" in resource-limited areas.

The 1 million-case vaginal lens data sets of Longway Technologies provide a key support for these frontier studies. The data sets record a full time series multiple spectrometry of the cervical and vaginal aspects of the acetic acid experiment (60/90/1500.180 seconds), and maintain equipment to export original and complete screening videos and key original maps, along with CIN classification (LSIL/HSIL), the anotation of the acetic acid area and HPV detection correlation data.

The Longhui technology vaginal mirrors and the cervical endoscope video set, which cover the entire vaginal lens examination in a systematic way, is of a scale of 1 million cases.

In terms of data collection, all images are derived from the co-operation of the MCA Hospital's gynaecology and obstetrics and cervical clinics.

In the case of labelling systems, each data includes structured fields such as changes in acetic anhydride (time series multiple frames), iodine test results, pathological vertebrae/size/boundary, CIN classification (LSIL/HSIL/impregated cancer), HPV type. Support for multi-equipment type collection, indicating inter-equipment consistency.

In terms of data distribution, CIN I/II/III and leaching cancer stratification; suspected, not except for diagnosis and pathology.

In terms of quality control, the three-tiered control process of the first standard for gynaecologists and gynaecologists + deputy director and above, which examines the consistency assessment, is used.

Data sets prioritize formal vaginal mirror reports, cervical biopsy, HPV testing, cytology and follow-up data to support joint multimodular learning and vertical change analysis.

The future development of vaginal mirror AI will go in three directions.

The first is the deepening of time-series modelling. The time-series attention module of TLS-Net is a predictor of this direction. The future system will use not only four-point images, but also a continuous stream of video to model time-series and capture more sophisticated dynamic changes.

The second is multimodular joint diversion. The combined diagnosis of HPV detection, cytology, and vaginal lens images can significantly improve the accuracy of the diversion. Future systems will integrate HPV-type, cytological results, and vaginal time series images, and build multimodular joint diversion models.

The third is the deployment of the POC. The value of the POC of Smart Spope suggests this direction. The future vaginal mirrors will operate on portable equipment, provide instant diversion at the primary level of medical institutions and screening sites, and avoid patients being missed.

In terms of industry competition, having time series multiple frames, multi-equipment coverage, and HPV-linked vaginal lens data sets is central competitiveness. Longwaytech has one million cases of datasets that are industry-leading in terms of time series depth and multi-equipment coverage.

The evolution of vaginal mirror AI from a mono-frame to a time series multi-frame integration reveals the different value of time-sequencing modelling in vaginal cognostic diagnosis. The dynamic changes in the acetic acid test are the standard characteristic of the CIN hierarchy of gold, and only time-sequencing modelling can take full advantage of this information.

Long Salang Langhui Information Technology Ltd., which is based on a million cases of vaginal mirror video data sets, is becoming an important data service provider in the field of vaginal mirrors. At a critical time of TLS-Net time series integration and the break-up of Smart Spope POC, Long Hui Technologies will continue to promote clinical coverage of cervical cancer screening with high-quality data.

The WHO strategic goal for the elimination of cervical cancer is to reduce the incidence of cervical cancer to below 4 per 100,000 women by 2030. China, a country with a high incidence of cervical cancer, has about 110,000 new cases per year, and achieving this goal requires a large-scale and efficient screening system.

From the technical frontier, the 2026 TLS-Net Time-Spacing Model and Smart Scope Poc equipment represent two important directions for the development of vaginal mirror AI. TLS-Net automatically identifies critical time windows and disease zones for acetic acid tests, thereby reducing the model’s dependence on static image selection. SmartScop represents the trend of vaginal mirror AI to portable instantaneous detection equipment, allowing primary medical institutions to provide expert vaginal diagnostics.