Magnetic resonance spectroscopy of breast cancer tissue used for tumor classification and lymph node prediction
Bibliographic record
Abstract
The treatment plan for a breast cancer patient is based on tumor size and grade, lymph node involvement and steroid hormone receptors. Lymph node status is the strongest prognostic factor for breast cancer patients. About 25% of node-negative patients experience recurrence or metastasis [ 1 ]. Additional methods might be important for better treatment strategies. Malignant cells have an altered metabolism, and metabolic mapping might become a tool in cancer diagnostics. High-resolution magic angle spinning (HR-MAS) magnetic resonance (MR) spectroscopy of tissue biopsies provides detailed information on their metabolic composition [ 2 ]. The aim of this study was to compare MR spectroscopic findings from breast cancer tissue with histological grading of tumor and patient lymph node status. Breast cancer and non-involved adjacent tissue were excised from patients with palpable breast cancer diagnosed as invasive ductal carcinoma (IDC). Tissue specimens were analyzed in D 2 O-PBS in a 50 μl MAS rotor (4 mm o.d.). HR-MAS MR spectra were recorded on a BRUKER AVANCE DRX600 spectrometer at 4°C. The samples were spun at 5 kHz. Proton MR spin echo spectra were acquired with a total echo time of 285 ms and presaturation of the water peak. A pathologist scored the relative areas of normal and neoplastic elements visually after MR analysis. Samples with less than 5% tumor content were excluded. This resulted in a final database consisting of 115 samples: 48 non-involved tissue and 69 IDCs (grade I [ n = 4], grade II [ n = 33] and grade III [ n = 32]). Of the 69 tumor samples, 37 were from patients with no spread and 32 samples were from patients with lymphatic spread of cancer cells. The spectral region 2.9–4.8 ppm was selected for principal component analysis (PCA). Two sample sets were used as PCA input: all samples ( n = 115) and tumor samples ( n = 69). Classification of IDC groups (grade I, grade II and grade III) and lymphatic spread was performed by a probabilistic neural network (PNN) strategy [ 3 ]. The 25 first principal component (PC) scores from PCA of tumor samples were used as the input in PNN. Both PCA and PNN were performed with full cross-validation. The 2D score plot of PC2 and PC3 from the PCA of all samples is shown in Fig. 1 . All samples from non-involved tissue are clearly separated from tumor samples. Tumor samples intersperse with no possibility to differentiate among the three types of grading. The PNN of spectra from tumor samples resulted in true classification of 56 of the 69 samples with respect to grading, whereas two samples were not classified. The specificity and sensitivity of classification exceeded 80% for all groups. Principal component analysis score plot of all samples ( n = 115). Labeling of samples from patient diagnosis. A PCA score plot of PC2 and PC3 for tumor samples is shown in Fig. 2 . A trend of clustering with respect to lymph node status can be seen. Classification results of node-positive and node-negative samples using PNN is presented in Table 1 . Samples from patients with spread of cancer cells to lymph nodes can be predicted with a specificity of 97% and a sensitivity of 92%. Principal component analysis score plot of tumor samples ( n = 69). Labeling of samples from patient lymph node status. PCA led to a complete separation of the non-involved and cancerous samples. The metabolism of cancerous tissue is clearly different from non-involved tissue. Samples from lymph node-positive and lymph node-negative patients could not be separated by PCA, while PNN led to classification of the two groups with misclassification of only four samples. Metabolic patterns in breast tumors from patients with lymphatic spread differ from those without lymphatic spread. These findings show that HR-MAS of breast cancer biopsies has the potential of becoming a diagnostic tool.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".