Financial Data and Key Metrics Changes - Total revenue for Q2 2025 was $6.7 million, a decrease of 13% compared to Q2 2024, but a 5% decrease when adjusted for discontinued services [11] - Non-GAAP gross margin for Q2 2025 was 52%, significantly higher than the 35% reported a year ago [12] - Non-GAAP operating expenses for Q2 2025 were $8.8 million, down 53% from $18.8 million in Q2 2024 [12][16] - Cash and cash equivalents at the end of the quarter were $27.4 million, with $11 million subject to restrictions [13] Business Line Data and Key Metrics Changes - The company sold 7,233 flow cells in Q2 2025, reflecting a 17% increase year-over-year [12] - Revenue from consumables and software combined grew 16% year-over-year in Q2 2025 [15] - The percentage of consumables and software revenue increased from 55% in 2024 to 73% in Q2 2025 [15] Market Data and Key Metrics Changes - The optical genome mapping (OGM) community surpassed 10,000 published clinical research genomes, marking a significant milestone [19] - The 2025 quarter had the highest number of publications in the history of optical genome mapping [19] Company Strategy and Development Direction - The company is focusing on digital pathology, particularly in cytogenetics and molecular pathology, with plans to expand into clinical and anatomic pathology [6] - The strategy includes supporting routine users, driving utilization of VIA software, building reimbursement support for OGM, and improving profitability [10] - The company aims to consolidate workflows in pathology through AI-driven software and optical genome mapping [7][8] Management's Comments on Operating Environment and Future Outlook - Management reiterated full-year revenue guidance of $26 to $30 million, with Q3 revenues expected to be in the range of $6.7 to $7.2 million [20] - The company raised expectations for new OGM installations in 2025 to 20-25 systems, up from the previous range of 15-20 [20] - Management expressed confidence in the ongoing strategy and the potential for significant business upside [51] Other Important Information - The company has made substantial progress in reducing operating expenses, cutting over $100 million in annual non-GAAP operating expenses since 2023 [16] - The release of new software updates is expected to enhance the usability of OGM workflows [18] Q&A Session Summary Question: How universal is the use of VIA software among OGM users? - Management indicated that VIA is installed in about a third of the OGM systems, with higher adoption among routine users, suggesting significant opportunities for expansion [24][25] Question: Are there efforts to market VIA to non-OGM users? - Management confirmed that a substantial amount of software is sold for non-OGM applications, and there is potential for cross-selling to convert these users into OGM customers [27][28] Question: What role does AI play in the VIA software? - AI enhances the analysis of variants by leveraging databases to improve accuracy and speed in reporting [29] Question: Can you clarify the return rate for new systems placed? - Management stated it is premature to discuss return rates, but they expect low returns due to the strategic shift in customer focus [36][37] Question: How does the new CPT code differ from the first one? - The new CPT code is for constitutional genetic disorders, which may lead to differences in pricing and facilitate reimbursement [40][41] Question: Is the instrument placement guidance conservative? - Management acknowledged that the guidance is conservative, considering potential delays in installations [45] Question: What is the expected time for new customers to reach maturity? - It typically takes a minimum of three months, with six to nine months being a healthy timeframe for labs to reach routine use [47][48]

Core Viewpoint - The research team from Harbin Institute of Technology has proposed a novel Hierarchical Distillation Multi-Instance Learning (HDMIL) framework aimed at rapidly identifying irrelevant patches in whole slide images (WSI) for efficient and accurate classification [1][3][17]. Group 1: Research Background - Pathological images are considered the "gold standard" for cancer diagnosis, with whole slide images (WSI) being a mainstream method due to their high resolution and large data volume [1][2]. - Multi-Instance Learning (MIL) is a primary method for analyzing WSI, but it faces challenges due to the vast amount of information contained in WSI, leading to high costs in data preprocessing and redundancy issues [2][3]. Group 2: Methodology - The HDMIL framework includes two key components: a Dynamic Multi-Instance Network (DMIN) for classifying high-resolution WSI and a Lightweight Instance Prescreening Network (LIPN) tailored for low-resolution WSI [5][9]. - The DMIN utilizes a self-distillation training strategy to identify irrelevant areas in WSI, while the LIPN is designed to quickly identify irrelevant regions in low-resolution WSI, indirectly indicating irrelevant patches in high-resolution WSI [9][12]. Group 3: Experimental Results - HDMIL demonstrated a 28.6% reduction in inference time compared to previous advanced methods across three public datasets [3][15]. - The framework was validated on three datasets: Camelyon16 for breast cancer lymph node metastasis detection, TCGA-NSCLC for lung cancer subtype classification, and TCGA-BRCA for breast cancer subtype classification [4][15]. - HDMIL achieved an AUC of 90.88% and an accuracy of 88.61% on the Camelyon16 dataset, outperforming previous best methods by 3.13% and 3.18%, respectively [14][15]. Group 4: Innovation and Impact - The introduction of the Kolmogorov-Arnold classifier within the HDMIL framework significantly enhances classification performance [11][17]. - The research contributes to the rapid development of digital pathology, particularly in cancer diagnosis, showcasing the potential of AI in medical applications [18][21].