PURPOSE: Automated curation of breast cancer treatment data with minimal human involvement could accelerate the collection of statewide and nationwide evidence for patient management and assessing the effectiveness of treatment pathways. The primary challenges are the complexity and inconsistency of structured clinical data streams and accurate extraction of this information from free-text clinical narratives. MATERIALS AND METHODS: We proposed a hybrid two-phase information extraction framework that combined a Unified Medical Language System parser (phase-1) with a fine-tuned large language model (LLM; phase-2) to extract longitudinal treatment timelines from time-stamped clinical notes. Our framework was developed through end-to-end joint learning as a question-answering model, where the model was trained to simultaneously answer five questions, each corresponding to a specific treatment. RESULTS: We fine-tuned and internally validated the model on 26,692 patients with breast cancer (diagnosed between 2013 and 2020) receiving treatment at Mayo Clinic and externally validated the model on 162 randomly selected patients from Stanford Healthcare. Zero-shot LLM (out-of-the-box) had high specificity but low sensitivity, indicating that although these frameworks are useful for generic language understanding, they are lacking in terms of targeted clinical tasks. The proposed model achieved 0.942 average AUROC on the internal and 0.924 on the external data, demonstrating only marginal drop in performance when evaluated on external. The proposed model also achieved better trade-off between sensitivity (average: 79.2%) and specificity (average: 76.2%) compared with rule-based (average sensitivity: 70.5%, average specificity: 68.1%) and structured codes (average sensitivity: 64.1%, average specificity: 83.5%). CONCLUSION: The proposed framework can extract temporal information about cancer treatments from various time-stamped clinic notes, regardless of the setting of treatment administration (inpatient or outpatient) or time frame. To support the cancer research community for such data curation and longitudinal analysis, we have packaged the code as a docker image, which needs minimal system reconfiguration and shared with an open-source academic license.
We analyze a single server polling model with two finite buffer queues (high and low priority queues), in which high-priority queue is served with queue length dependent service intervals, while low-priority queue is served with constant service intervals. The server follows limited service discipline to serve packets from both the queues. We derive buffer occupancy distribution of high priority queue at departure instants by using embedded Markov chain approach. Simulation results confirm the validity of analytical results. This polling model can be used to model and analyze any node in a packet-switched network, which handles both real-time and non real-time traffic.
We derive the distribution of inter-departure times of a finite-buffer single-server queue with Poisson arrival process and queue-length dependent service times, where the server goes to vacation if either the queue is emptied or a limited number (R1) of packets are served, whichever occurs first, in the current busy period. We consider two types of vacation distributions: (i) deterministic and (ii) exponential. Queue-length distribution at embedded points is derived first, then, distribution and variance of inter-departure times are derived, for both types of vacations. Simulation results are in good agreement with the derived analytical results. The above framework would be useful at the receiver in modeling and analyzing jitter and waiting time of Time Division Multiplexing (TDM) Emulated Packets (TEPs) in TDM over Packet-Switched Network (TDM over PSN) technology as a function of buffer size.
TDMoIP is the emulation of Time Division Multiplexing (TDM) over Packet Switched Network (PSN), say IP. When TDM traffic is packetized and injected into a PSN for transportation, packets arrive at the destination with different inter-arrival times, due to variable delay (jitter) introduced by PSN. This network induced jitter should be minimized, because TDM traffic is highly synchronous. Problem of jitter from theoretical (competitive and statistical analysis) to more practical view point had been well studied [1], [6]–[9]. In this paper, we have proposed two on-line algorithms, algorithm-A and algorithm-B to control the rate-jitter. We have shown both analytically and by simulation that the rate-jitter achieved by algorithm-A is strictly less than the rate-jitter of on-line algorithm proposed in [1] and through simulations, it is shown that algorithm-B also achieves less rate-jitter compared with [1]. We also undertake the statistical analysis of above algorithms, in particular, we modeled jitter buffer as an M/D̃/1/B on queue and steady state queue length distribution is calculated. The correctness of the analytical results corresponding to our proposed algorithms is verified with simulation results. From the simulation results, it is also shown that the mean waiting time of a packet in the buffer is less for both the proposed algorithms compared with [1].